* [PATCH 01/04] Kexec / Kdump: Generic code
2006-11-22 7:10 [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502) Magnus Damm
@ 2006-11-22 7:10 ` Magnus Damm
2006-11-22 7:11 ` [PATCH 02/04] Kexec / Kdump: Code shared between x86_32 and x86_64 Magnus Damm
` (4 subsequent siblings)
5 siblings, 0 replies; 30+ messages in thread
From: Magnus Damm @ 2006-11-22 7:10 UTC (permalink / raw)
To: Keir Fraser
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, magnus.damm,
Isaku Yamahata, Magnus Damm, Horms
[PATCH 01/04] Kexec / Kdump: Generic code
This patch implements the generic portion of the Kexec / Kdump port to Xen.
Signed-Off-By: Magnus Damm <magnus@valinux.co.jp>
---
Applies on top of xen-unstable-12502.
linux-2.6-xen-sparse/drivers/xen/core/Makefile | 1
linux-2.6-xen-sparse/drivers/xen/core/machine_kexec.c | 171 ++++++++
patches/linux-2.6.16.29/kexec-generic.patch | 224 +++++++++++
patches/linux-2.6.16.29/series | 1
xen/arch/ia64/xen/Makefile | 2
xen/arch/ia64/xen/crash.c | 19
xen/arch/ia64/xen/machine_kexec.c | 34 +
xen/arch/powerpc/Makefile | 2
xen/arch/powerpc/crash.c | 19
xen/arch/powerpc/machine_kexec.c | 34 +
xen/arch/x86/Makefile | 2
xen/arch/x86/crash.c | 19
xen/arch/x86/machine_kexec.c | 34 +
xen/common/Makefile | 1
xen/common/kexec.c | 345 +++++++++++++++++
xen/common/page_alloc.c | 33 +
xen/drivers/char/console.c | 3
xen/include/asm-ia64/elf.h | 25 +
xen/include/asm-ia64/kexec.h | 25 +
xen/include/asm-powerpc/elf.h | 25 +
xen/include/asm-powerpc/kexec.h | 25 +
xen/include/asm-x86/elf.h | 25 +
xen/include/asm-x86/kexec.h | 24 +
xen/include/public/kexec.h | 131 ++++++
xen/include/xen/elf.h | 16
xen/include/xen/elfcore.h | 127 ++++++
xen/include/xen/hypercall.h | 6
xen/include/xen/kexec.h | 43 ++
xen/include/xen/mm.h | 1
29 files changed, 1394 insertions(+), 23 deletions(-)
--- 0001/linux-2.6-xen-sparse/drivers/xen/core/Makefile
+++ work/linux-2.6-xen-sparse/drivers/xen/core/Makefile
@@ -11,3 +11,4 @@ obj-$(CONFIG_XEN_SYSFS) += xen_sysfs.o
obj-$(CONFIG_XEN_SKBUFF) += skbuff.o
obj-$(CONFIG_XEN_REBOOT) += reboot.o machine_reboot.o
obj-$(CONFIG_XEN_SMPBOOT) += smpboot.o
+obj-$(CONFIG_KEXEC) += machine_kexec.o
--- /dev/null
+++ work/linux-2.6-xen-sparse/drivers/xen/core/machine_kexec.c
@@ -0,0 +1,171 @@
+/*
+ * drivers/xen/core/machine_kexec.c
+ * handle transition of Linux booting another kernel
+ */
+
+#include <linux/kexec.h>
+#include <xen/interface/kexec.h>
+#include <linux/mm.h>
+#include <linux/bootmem.h>
+#include <asm/hypercall.h>
+#include <asm/kexec-xen.h>
+
+extern void machine_kexec_setup_load_arg(xen_kexec_image_t *xki,
+ struct kimage *image);
+
+int xen_max_nr_phys_cpus;
+struct resource xen_hypervisor_res;
+struct resource *xen_phys_cpus;
+
+void xen_machine_kexec_setup_resources(void)
+{
+ xen_kexec_range_t range;
+ struct resource *res;
+ int k = 0;
+
+ /* determine maximum number of physical cpus */
+
+ while (1) {
+ memset(&range, 0, sizeof(range));
+ range.range = KEXEC_RANGE_MA_CPU;
+ range.nr = k;
+
+ if (HYPERVISOR_kexec_op(KEXEC_CMD_kexec_get_range, &range))
+ break;
+
+ k++;
+ }
+
+ xen_max_nr_phys_cpus = k;
+
+ /* allocate xen_phys_cpus */
+
+ xen_phys_cpus = alloc_bootmem_low(k * sizeof(struct resource));
+ BUG_ON(!xen_phys_cpus);
+
+ /* fill in xen_phys_cpus with per-cpu crash note information */
+
+ for (k = 0; k < xen_max_nr_phys_cpus; k++) {
+ memset(&range, 0, sizeof(range));
+ range.range = KEXEC_RANGE_MA_CPU;
+ range.nr = k;
+
+ BUG_ON(HYPERVISOR_kexec_op(KEXEC_CMD_kexec_get_range, &range));
+
+ res = xen_phys_cpus + k;
+
+ memset(res, 0, sizeof(*res));
+ res->name = "Crash note";
+ res->start = range.start;
+ res->end = range.start + range.size - 1;
+ res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
+ }
+
+ /* fill in xen_hypervisor_res with hypervisor machine address range */
+
+ memset(&range, 0, sizeof(range));
+ range.range = KEXEC_RANGE_MA_XEN;
+
+ BUG_ON(HYPERVISOR_kexec_op(KEXEC_CMD_kexec_get_range, &range));
+
+ xen_hypervisor_res.name = "Hypervisor code and data";
+ xen_hypervisor_res.start = range.start;
+ xen_hypervisor_res.end = range.start + range.size - 1;
+ xen_hypervisor_res.flags = IORESOURCE_BUSY | IORESOURCE_MEM;
+
+ /* fill in crashk_res if range is reserved by hypervisor */
+
+ memset(&range, 0, sizeof(range));
+ range.range = KEXEC_RANGE_MA_CRASH;
+
+ BUG_ON(HYPERVISOR_kexec_op(KEXEC_CMD_kexec_get_range, &range));
+
+ if (range.size) {
+ crashk_res.start = range.start;
+ crashk_res.end = range.start + range.size - 1;
+ }
+}
+
+void xen_machine_kexec_register_resources(struct resource *res)
+{
+ int k;
+
+ request_resource(res, &xen_hypervisor_res);
+
+ for (k = 0; k < xen_max_nr_phys_cpus; k++)
+ request_resource(res, xen_phys_cpus + k);
+
+}
+
+static void setup_load_arg(xen_kexec_image_t *xki, struct kimage *image)
+{
+ machine_kexec_setup_load_arg(xki, image);
+
+ xki->indirection_page = image->head;
+ xki->start_address = image->start;
+}
+
+/*
+ * Load the image into xen so xen can kdump itself
+ * This might have been done in prepare, but prepare
+ * is currently called too early. It might make sense
+ * to move prepare, but for now, just add an extra hook.
+ */
+int xen_machine_kexec_load(struct kimage *image)
+{
+ xen_kexec_load_t xkl;
+
+ memset(&xkl, 0, sizeof(xkl));
+ xkl.type = image->type;
+ setup_load_arg(&xkl.image, image);
+ return HYPERVISOR_kexec_op(KEXEC_CMD_kexec_load, &xkl);
+}
+
+/*
+ * Unload the image that was stored by machine_kexec_load()
+ * This might have been done in machine_kexec_cleanup() but it
+ * is called too late, and its possible xen could try and kdump
+ * using resources that have been freed.
+ */
+void xen_machine_kexec_unload(struct kimage *image)
+{
+ xen_kexec_load_t xkl;
+
+ memset(&xkl, 0, sizeof(xkl));
+ xkl.type = image->type;
+ HYPERVISOR_kexec_op(KEXEC_CMD_kexec_unload, &xkl);
+}
+
+/*
+ * Do not allocate memory (or fail in any way) in machine_kexec().
+ * We are past the point of no return, committed to rebooting now.
+ *
+ * This has the hypervisor move to the prefered reboot CPU,
+ * stop all CPUs and kexec. That is it combines machine_shutdown()
+ * and machine_kexec() in Linux kexec terms.
+ */
+NORET_TYPE void xen_machine_kexec(struct kimage *image)
+{
+ xen_kexec_exec_t xke;
+
+ memset(&xke, 0, sizeof(xke));
+ xke.type = image->type;
+ HYPERVISOR_kexec_op(KEXEC_CMD_kexec, &xke);
+ panic("KEXEC_CMD_kexec hypercall should not return\n");
+}
+
+void machine_shutdown(void)
+{
+ /* do nothing */
+}
+
+
+/*
+ * Local variables:
+ * c-file-style: "linux"
+ * indent-tabs-mode: t
+ * c-indent-level: 8
+ * c-basic-offset: 8
+ * tab-width: 8
+ * End:
+ */
--- /dev/null
+++ work/patches/linux-2.6.16.29/kexec-generic.patch
@@ -0,0 +1,224 @@
+--- 0001/include/linux/kexec.h
++++ work/include/linux/kexec.h
+@@ -91,6 +91,13 @@ struct kimage {
+ extern NORET_TYPE void machine_kexec(struct kimage *image) ATTRIB_NORET;
+ extern int machine_kexec_prepare(struct kimage *image);
+ extern void machine_kexec_cleanup(struct kimage *image);
++#ifdef CONFIG_XEN
++extern int xen_machine_kexec_load(struct kimage *image);
++extern void xen_machine_kexec_unload(struct kimage *image);
++extern NORET_TYPE void xen_machine_kexec(struct kimage *image) ATTRIB_NORET;
++extern void xen_machine_kexec_setup_resources(void);
++extern void xen_machine_kexec_register_resources(struct resource *res);
++#endif
+ extern asmlinkage long sys_kexec_load(unsigned long entry,
+ unsigned long nr_segments,
+ struct kexec_segment __user *segments,
+--- 0001/kernel/kexec.c
++++ work/kernel/kexec.c
+@@ -26,6 +26,9 @@
+ #include <asm/io.h>
+ #include <asm/system.h>
+ #include <asm/semaphore.h>
++#ifdef CONFIG_XEN
++#include <asm/kexec-xen.h>
++#endif
+
+ /* Per cpu memory for storing cpu states in case of system crash. */
+ note_buf_t* crash_notes;
+@@ -403,7 +406,7 @@ static struct page *kimage_alloc_normal_
+ pages = kimage_alloc_pages(GFP_KERNEL, order);
+ if (!pages)
+ break;
+- pfn = page_to_pfn(pages);
++ pfn = kexec_page_to_pfn(pages);
+ epfn = pfn + count;
+ addr = pfn << PAGE_SHIFT;
+ eaddr = epfn << PAGE_SHIFT;
+@@ -437,6 +440,7 @@ static struct page *kimage_alloc_normal_
+ return pages;
+ }
+
++#ifndef CONFIG_XEN
+ static struct page *kimage_alloc_crash_control_pages(struct kimage *image,
+ unsigned int order)
+ {
+@@ -490,7 +494,7 @@ static struct page *kimage_alloc_crash_c
+ }
+ /* If I don't overlap any segments I have found my hole! */
+ if (i == image->nr_segments) {
+- pages = pfn_to_page(hole_start >> PAGE_SHIFT);
++ pages = kexec_pfn_to_page(hole_start >> PAGE_SHIFT);
+ break;
+ }
+ }
+@@ -517,6 +521,13 @@ struct page *kimage_alloc_control_pages(
+
+ return pages;
+ }
++#else /* !CONFIG_XEN */
++struct page *kimage_alloc_control_pages(struct kimage *image,
++ unsigned int order)
++{
++ return kimage_alloc_normal_control_pages(image, order);
++}
++#endif
+
+ static int kimage_add_entry(struct kimage *image, kimage_entry_t entry)
+ {
+@@ -532,7 +543,7 @@ static int kimage_add_entry(struct kimag
+ return -ENOMEM;
+
+ ind_page = page_address(page);
+- *image->entry = virt_to_phys(ind_page) | IND_INDIRECTION;
++ *image->entry = kexec_virt_to_phys(ind_page) | IND_INDIRECTION;
+ image->entry = ind_page;
+ image->last_entry = ind_page +
+ ((PAGE_SIZE/sizeof(kimage_entry_t)) - 1);
+@@ -593,13 +604,13 @@ static int kimage_terminate(struct kimag
+ #define for_each_kimage_entry(image, ptr, entry) \
+ for (ptr = &image->head; (entry = *ptr) && !(entry & IND_DONE); \
+ ptr = (entry & IND_INDIRECTION)? \
+- phys_to_virt((entry & PAGE_MASK)): ptr +1)
++ kexec_phys_to_virt((entry & PAGE_MASK)): ptr +1)
+
+ static void kimage_free_entry(kimage_entry_t entry)
+ {
+ struct page *page;
+
+- page = pfn_to_page(entry >> PAGE_SHIFT);
++ page = kexec_pfn_to_page(entry >> PAGE_SHIFT);
+ kimage_free_pages(page);
+ }
+
+@@ -611,6 +622,10 @@ static void kimage_free(struct kimage *i
+ if (!image)
+ return;
+
++#ifdef CONFIG_XEN
++ xen_machine_kexec_unload(image);
++#endif
++
+ kimage_free_extra_pages(image);
+ for_each_kimage_entry(image, ptr, entry) {
+ if (entry & IND_INDIRECTION) {
+@@ -686,7 +701,7 @@ static struct page *kimage_alloc_page(st
+ * have a match.
+ */
+ list_for_each_entry(page, &image->dest_pages, lru) {
+- addr = page_to_pfn(page) << PAGE_SHIFT;
++ addr = kexec_page_to_pfn(page) << PAGE_SHIFT;
+ if (addr == destination) {
+ list_del(&page->lru);
+ return page;
+@@ -701,12 +716,12 @@ static struct page *kimage_alloc_page(st
+ if (!page)
+ return NULL;
+ /* If the page cannot be used file it away */
+- if (page_to_pfn(page) >
++ if (kexec_page_to_pfn(page) >
+ (KEXEC_SOURCE_MEMORY_LIMIT >> PAGE_SHIFT)) {
+ list_add(&page->lru, &image->unuseable_pages);
+ continue;
+ }
+- addr = page_to_pfn(page) << PAGE_SHIFT;
++ addr = kexec_page_to_pfn(page) << PAGE_SHIFT;
+
+ /* If it is the destination page we want use it */
+ if (addr == destination)
+@@ -729,7 +744,7 @@ static struct page *kimage_alloc_page(st
+ struct page *old_page;
+
+ old_addr = *old & PAGE_MASK;
+- old_page = pfn_to_page(old_addr >> PAGE_SHIFT);
++ old_page = kexec_pfn_to_page(old_addr >> PAGE_SHIFT);
+ copy_highpage(page, old_page);
+ *old = addr | (*old & ~PAGE_MASK);
+
+@@ -779,7 +794,7 @@ static int kimage_load_normal_segment(st
+ result = -ENOMEM;
+ goto out;
+ }
+- result = kimage_add_page(image, page_to_pfn(page)
++ result = kimage_add_page(image, kexec_page_to_pfn(page)
+ << PAGE_SHIFT);
+ if (result < 0)
+ goto out;
+@@ -811,6 +826,7 @@ out:
+ return result;
+ }
+
++#ifndef CONFIG_XEN
+ static int kimage_load_crash_segment(struct kimage *image,
+ struct kexec_segment *segment)
+ {
+@@ -833,7 +849,7 @@ static int kimage_load_crash_segment(str
+ char *ptr;
+ size_t uchunk, mchunk;
+
+- page = pfn_to_page(maddr >> PAGE_SHIFT);
++ page = kexec_pfn_to_page(maddr >> PAGE_SHIFT);
+ if (page == 0) {
+ result = -ENOMEM;
+ goto out;
+@@ -881,6 +897,13 @@ static int kimage_load_segment(struct ki
+
+ return result;
+ }
++#else /* CONFIG_XEN */
++static int kimage_load_segment(struct kimage *image,
++ struct kexec_segment *segment)
++{
++ return kimage_load_normal_segment(image, segment);
++}
++#endif
+
+ /*
+ * Exec Kernel system call: for obvious reasons only root may call it.
+@@ -991,6 +1014,11 @@ asmlinkage long sys_kexec_load(unsigned
+ if (result)
+ goto out;
+ }
++#ifdef CONFIG_XEN
++ result = xen_machine_kexec_load(image);
++ if (result)
++ goto out;
++#endif
+ /* Install the new kernel, and Uninstall the old */
+ image = xchg(dest_image, image);
+
+@@ -1045,7 +1073,6 @@ void crash_kexec(struct pt_regs *regs)
+ struct kimage *image;
+ int locked;
+
+-
+ /* Take the kexec_lock here to prevent sys_kexec_load
+ * running on one cpu from replacing the crash kernel
+ * we are using after a panic on a different cpu.
+@@ -1061,7 +1088,11 @@ void crash_kexec(struct pt_regs *regs)
+ struct pt_regs fixed_regs;
+ crash_setup_regs(&fixed_regs, regs);
+ machine_crash_shutdown(&fixed_regs);
++#ifdef CONFIG_XEN
++ xen_machine_kexec(image);
++#else
+ machine_kexec(image);
++#endif
+ }
+ xchg(&kexec_lock, 0);
+ }
+--- 0002/kernel/sys.c
++++ work/kernel/sys.c
+@@ -435,8 +435,12 @@ void kernel_kexec(void)
+ kernel_restart_prepare(NULL);
+ printk(KERN_EMERG "Starting new kernel\n");
+ machine_shutdown();
++#ifdef CONFIG_XEN
++ xen_machine_kexec(image);
++#else
+ machine_kexec(image);
+ #endif
++#endif
+ }
+ EXPORT_SYMBOL_GPL(kernel_kexec);
+
--- 0001/patches/linux-2.6.16.29/series
+++ work/patches/linux-2.6.16.29/series
@@ -1,3 +1,4 @@
+kexec-generic.patch
blktap-aio-16_03_06.patch
device_bind.patch
fix-hz-suspend.patch
--- 0001/xen/arch/ia64/xen/Makefile
+++ work/xen/arch/ia64/xen/Makefile
@@ -1,3 +1,5 @@
+obj-y += machine_kexec.o
+obj-y += crash.o
obj-y += acpi.o
obj-y += dom0_ops.o
obj-y += domain.o
--- /dev/null
+++ work/xen/arch/ia64/xen/crash.c
@@ -0,0 +1,19 @@
+#include <xen/lib.h> /* for printk() used in stub */
+#include <xen/types.h>
+#include <public/kexec.h>
+
+void machine_crash_shutdown(void)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
+
--- /dev/null
+++ work/xen/arch/ia64/xen/machine_kexec.c
@@ -0,0 +1,34 @@
+#include <xen/lib.h> /* for printk() used in stubs */
+#include <xen/types.h>
+#include <public/kexec.h>
+
+int machine_kexec_load(int type, int slot, xen_kexec_image_t *image)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ return -1;
+}
+
+void machine_kexec_unload(int type, int slot, xen_kexec_image_t *image)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+void machine_kexec(xen_kexec_image_t *image)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+void machine_shutdown(xen_kexec_image_t *image)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--- 0001/xen/arch/powerpc/Makefile
+++ work/xen/arch/powerpc/Makefile
@@ -40,6 +40,8 @@ obj-y += smp-tbsync.o
obj-y += sysctl.o
obj-y += time.o
obj-y += usercopy.o
+obj-y += machine_kexec.o
+obj-y += crash.o
obj-$(debug) += 0opt.o
obj-$(crash_debug) += gdbstub.o
--- /dev/null
+++ work/xen/arch/powerpc/crash.c
@@ -0,0 +1,19 @@
+#include <xen/lib.h> /* for printk() used in stub */
+#include <xen/types.h>
+#include <public/kexec.h>
+
+void machine_crash_shutdown(void)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
+
--- /dev/null
+++ work/xen/arch/powerpc/machine_kexec.c
@@ -0,0 +1,34 @@
+#include <xen/lib.h> /* for printk() used in stubs */
+#include <xen/types.h>
+#include <public/kexec.h>
+
+int machine_kexec_load(int type, int slot, xen_kexec_image_t *image)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ return -1;
+}
+
+void machine_kexec_unload(int type, int slot, xen_kexec_image_t *image)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+void machine_kexec(xen_kexec_image_t *image)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+void machine_shutdown(xen_kexec_image_t *image)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--- 0001/xen/arch/x86/Makefile
+++ work/xen/arch/x86/Makefile
@@ -43,6 +43,8 @@ obj-y += trampoline.o
obj-y += traps.o
obj-y += usercopy.o
obj-y += x86_emulate.o
+obj-y += machine_kexec.o
+obj-y += crash.o
obj-$(crash_debug) += gdbstub.o
--- /dev/null
+++ work/xen/arch/x86/crash.c
@@ -0,0 +1,19 @@
+#include <xen/lib.h> /* for printk() used in stub */
+#include <xen/types.h>
+#include <public/kexec.h>
+
+void machine_crash_shutdown(void)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
+
--- /dev/null
+++ work/xen/arch/x86/machine_kexec.c
@@ -0,0 +1,34 @@
+#include <xen/lib.h> /* for printk() used in stubs */
+#include <xen/types.h>
+#include <public/kexec.h>
+
+int machine_kexec_load(int type, int slot, xen_kexec_image_t *image)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ return -1;
+}
+
+void machine_kexec_unload(int type, int slot, xen_kexec_image_t *image)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+void machine_kexec(xen_kexec_image_t *image)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+void machine_shutdown(xen_kexec_image_t *image)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--- 0001/xen/common/Makefile
+++ work/xen/common/Makefile
@@ -7,6 +7,7 @@ obj-y += event_channel.o
obj-y += grant_table.o
obj-y += kernel.o
obj-y += keyhandler.o
+obj-y += kexec.o
obj-y += lib.o
obj-y += memory.o
obj-y += multicall.o
--- /dev/null
+++ work/xen/common/kexec.c
@@ -0,0 +1,345 @@
+/******************************************************************************
+ * kexec.c - Achitecture independent kexec code for Xen
+ *
+ * Xen port written by:
+ * - Simon 'Horms' Horman <horms@verge.net.au>
+ * - Magnus Damm <magnus@valinux.co.jp>
+ */
+
+#include <asm/kexec.h>
+#include <xen/lib.h>
+#include <xen/ctype.h>
+#include <xen/errno.h>
+#include <xen/guest_access.h>
+#include <xen/sched.h>
+#include <xen/types.h>
+#include <xen/kexec.h>
+#include <xen/keyhandler.h>
+#include <public/kexec.h>
+#include <xen/cpumask.h>
+#include <asm/atomic.h>
+#include <xen/spinlock.h>
+
+static char opt_crashkernel[32] = "";
+string_param("crashkernel", opt_crashkernel);
+
+DEFINE_PER_CPU (crash_note_t, crash_notes);
+cpumask_t crash_saved_cpus;
+int crashing_cpu;
+
+xen_kexec_image_t kexec_image[KEXEC_IMAGE_NR];
+
+#define KEXEC_FLAG_DEFAULT_POS (KEXEC_IMAGE_NR + 0)
+#define KEXEC_FLAG_CRASH_POS (KEXEC_IMAGE_NR + 1)
+#define KEXEC_FLAG_IN_PROGRESS (KEXEC_IMAGE_NR + 2)
+
+unsigned long kexec_flags = 0; /* the lowest bits are for KEXEC_IMAGE... */
+
+spinlock_t kexec_lock = SPIN_LOCK_UNLOCKED;
+
+static void one_cpu_only(void)
+{
+ /* Only allow the first cpu to continue - force other cpus to spin */
+ if (test_and_set_bit(KEXEC_FLAG_IN_PROGRESS, &kexec_flags))
+ {
+ while (1);
+ }
+}
+
+void machine_crash_save_cpu(void)
+{
+ int cpu = smp_processor_id();
+ crash_note_t *cntp;
+
+ if (!cpu_test_and_set(cpu, crash_saved_cpus))
+ {
+ cntp = &per_cpu(crash_notes, cpu);
+ elf_core_save_regs(&cntp->core.desc.desc.pr_reg);
+
+ /* setup crash note header */
+
+ cntp->core.note.note.note.namesz = CORE_STR_LEN;
+ cntp->core.note.note.note.descsz = sizeof(ELF_Prstatus);
+ cntp->core.note.note.note.type = NT_PRSTATUS;
+ memcpy(cntp->core.note.note.name, CORE_STR, CORE_STR_LEN);
+ }
+}
+
+void machine_crash_kexec(void)
+{
+ int pos, cpu;
+ xen_kexec_image_t *image;
+
+ one_cpu_only();
+
+ machine_crash_save_cpu();
+ crashing_cpu = smp_processor_id();
+
+ machine_crash_shutdown();
+
+ /* the memory for per-cpu crash notes are not initialized
+ * for non-present cpus. make sure the crash notes are cleared.
+ */
+
+ for (cpu = 0; cpu < NR_CPUS; cpu++)
+ {
+ if (cpu != crashing_cpu && !cpu_present(cpu))
+ memset(&per_cpu(crash_notes, cpu), 0, sizeof(crash_note_t));
+ }
+
+ pos = (test_bit(KEXEC_FLAG_CRASH_POS, &kexec_flags) != 0);
+
+ if (test_bit(KEXEC_IMAGE_CRASH_BASE + pos, &kexec_flags))
+ {
+ image = &kexec_image[KEXEC_IMAGE_CRASH_BASE + pos];
+ machine_kexec(image); /* Does not return */
+ }
+
+ while (1); /* No image available - just spin */
+}
+
+static void do_crashdump_trigger(unsigned char key)
+{
+ printk("triggering crashdump\n");
+ machine_crash_kexec();
+}
+
+static __init int register_crashdump_trigger(void)
+{
+ register_keyhandler('c', do_crashdump_trigger, "trigger a crashdump");
+ return 0;
+}
+__initcall(register_crashdump_trigger);
+
+void machine_kexec_reserved(xen_kexec_reserve_t *reservation)
+{
+ unsigned long val[2];
+ char *str = opt_crashkernel;
+ int k = 0;
+
+ memset(reservation, 0, sizeof(*reservation));
+
+ while (k < ARRAY_SIZE(val)) {
+ if (*str == '\0') {
+ break;
+ }
+ val[k] = simple_strtoul(str, &str, 0);
+ switch (toupper(*str)) {
+ case 'G': val[k] <<= 10;
+ case 'M': val[k] <<= 10;
+ case 'K': val[k] <<= 10;
+ str++;
+ }
+ if (*str == '@') {
+ str++;
+ }
+ k++;
+ }
+
+ if (k == ARRAY_SIZE(val)) {
+ reservation->size = val[0];
+ reservation->start = val[1];
+ }
+}
+
+static int kexec_get_reserve(xen_kexec_range_t *range)
+{
+ xen_kexec_reserve_t reservation;
+
+ machine_kexec_reserved(&reservation);
+
+ range->start = reservation.start;
+ range->size = reservation.size;
+ return 0;
+}
+
+extern unsigned long _text, _end;
+
+static int kexec_get_xen(xen_kexec_range_t *range, int get_ma)
+{
+ if (get_ma)
+ range->start = virt_to_maddr(&_text);
+ else
+ range->start = (unsigned long) &_text;
+
+ range->size = &_end - &_text;
+ return 0;
+}
+
+static int kexec_get_cpu(xen_kexec_range_t *range)
+{
+ if (range->nr < 0 || range->nr >= NR_CPUS)
+ return -EINVAL;
+
+ range->start = __pa((unsigned long)&per_cpu(crash_notes, range->nr));
+ range->size = sizeof(crash_note_t);
+ return 0;
+}
+
+static int kexec_get_range(XEN_GUEST_HANDLE(void) uarg)
+{
+ xen_kexec_range_t range;
+ int ret = -EINVAL;
+
+ if (unlikely(copy_from_guest(&range, uarg, 1)))
+ return -EFAULT;
+
+ switch (range.range)
+ {
+ case KEXEC_RANGE_MA_CRASH:
+ ret = kexec_get_reserve(&range);
+ break;
+ case KEXEC_RANGE_MA_XEN:
+ ret = kexec_get_xen(&range, 1);
+ break;
+ case KEXEC_RANGE_VA_XEN:
+ ret = kexec_get_xen(&range, 0);
+ break;
+ case KEXEC_RANGE_MA_CPU:
+ ret = kexec_get_cpu(&range);
+ break;
+ }
+
+ if (ret == 0 && unlikely(copy_to_guest(uarg, &range, 1)))
+ return -EFAULT;
+
+ return ret;
+}
+
+static int kexec_load_get_bits(int type, int *base, int *bit)
+{
+ switch (type)
+ {
+ case KEXEC_TYPE_DEFAULT:
+ *base = KEXEC_IMAGE_DEFAULT_BASE;
+ *bit = KEXEC_FLAG_DEFAULT_POS;
+ break;
+ case KEXEC_TYPE_CRASH:
+ *base = KEXEC_IMAGE_CRASH_BASE;
+ *bit = KEXEC_FLAG_CRASH_POS;
+ break;
+ default:
+ return -1;
+ }
+ return 0;
+}
+
+static int kexec_load_unload(unsigned long op, XEN_GUEST_HANDLE(void) uarg)
+{
+ xen_kexec_load_t load;
+ xen_kexec_image_t *image;
+ int base, bit, pos;
+ int ret = 0;
+
+ if (unlikely(copy_from_guest(&load, uarg, 1)))
+ return -EFAULT;
+
+ if (kexec_load_get_bits(load.type, &base, &bit))
+ return -EINVAL;
+
+ pos = (test_bit(bit, &kexec_flags) != 0);
+
+ /* Load the user data into an unused image */
+ if (op == KEXEC_CMD_kexec_load)
+ {
+ image = &kexec_image[base + !pos];
+
+ BUG_ON(test_bit((base + !pos), &kexec_flags)); /* must be free */
+
+ memcpy(image, &load.image, sizeof(*image));
+
+ if (!(ret = machine_kexec_load(load.type, base + !pos, image)))
+ {
+ /* Set image present bit */
+ set_bit((base + !pos), &kexec_flags);
+
+ /* Make new image the active one */
+ change_bit(bit, &kexec_flags);
+ }
+ }
+
+ /* Unload the old image if present and load successful */
+ if (ret == 0 && !test_bit(KEXEC_FLAG_IN_PROGRESS, &kexec_flags))
+ {
+ if (test_and_clear_bit((base + pos), &kexec_flags))
+ {
+ image = &kexec_image[base + pos];
+ machine_kexec_unload(load.type, base + pos, image);
+ }
+ }
+
+ return ret;
+}
+
+static int kexec_exec(XEN_GUEST_HANDLE(void) uarg)
+{
+ xen_kexec_exec_t exec;
+ xen_kexec_image_t *image;
+ int base, bit, pos;
+
+ if (unlikely(copy_from_guest(&exec, uarg, 1)))
+ return -EFAULT;
+
+ if (kexec_load_get_bits(exec.type, &base, &bit))
+ return -EINVAL;
+
+ pos = (test_bit(bit, &kexec_flags) != 0);
+
+ /* Only allow kexec/kdump into loaded images */
+ if (!test_bit(base + pos, &kexec_flags))
+ return -ENOENT;
+
+ switch (exec.type)
+ {
+ case KEXEC_TYPE_DEFAULT:
+ image = &kexec_image[base + pos];
+ one_cpu_only();
+ machine_shutdown(image); /* Does not return */
+ break;
+ case KEXEC_TYPE_CRASH:
+ machine_crash_kexec(); /* Does not return */
+ break;
+ }
+
+ return -EINVAL; /* never reached */
+}
+
+long do_kexec_op(unsigned long op, XEN_GUEST_HANDLE(void) uarg)
+{
+ unsigned long flags;
+ int ret = -EINVAL;
+
+ if ( !IS_PRIV(current->domain) )
+ return -EPERM;
+
+ switch (op)
+ {
+ case KEXEC_CMD_kexec_get_range:
+ ret = kexec_get_range(uarg);
+ break;
+ case KEXEC_CMD_kexec_load:
+ case KEXEC_CMD_kexec_unload:
+ spin_lock_irqsave(&kexec_lock, flags);
+ if (!test_bit(KEXEC_FLAG_IN_PROGRESS, &kexec_flags))
+ {
+ ret = kexec_load_unload(op, uarg);
+ }
+ spin_unlock_irqrestore(&kexec_lock, flags);
+ break;
+ case KEXEC_CMD_kexec:
+ ret = kexec_exec(uarg);
+ break;
+ }
+
+ return ret;
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--- 0001/xen/common/page_alloc.c
+++ work/xen/common/page_alloc.c
@@ -216,24 +216,35 @@ void init_boot_pages(paddr_t ps, paddr_t
}
}
+unsigned long alloc_boot_pages_at(unsigned long nr_pfns, unsigned long pfn_at)
+{
+ unsigned long i;
+
+ for ( i = 0; i < nr_pfns; i++ )
+ if ( allocated_in_map(pfn_at + i) )
+ break;
+
+ if ( i == nr_pfns )
+ {
+ map_alloc(pfn_at, nr_pfns);
+ return pfn_at;
+ }
+
+ return 0;
+}
+
unsigned long alloc_boot_pages(unsigned long nr_pfns, unsigned long pfn_align)
{
- unsigned long pg, i;
+ unsigned long pg, i = 0;
for ( pg = 0; (pg + nr_pfns) < max_page; pg += pfn_align )
{
- for ( i = 0; i < nr_pfns; i++ )
- if ( allocated_in_map(pg + i) )
- break;
-
- if ( i == nr_pfns )
- {
- map_alloc(pg, nr_pfns);
- return pg;
- }
+ i = alloc_boot_pages_at(nr_pfns, pg);
+ if (i != 0)
+ break;
}
- return 0;
+ return i;
}
--- 0001/xen/drivers/char/console.c
+++ work/xen/drivers/char/console.c
@@ -27,6 +27,7 @@
#include <xen/guest_access.h>
#include <xen/shutdown.h>
#include <xen/vga.h>
+#include <xen/kexec.h>
#include <asm/current.h>
#include <asm/debugger.h>
#include <asm/io.h>
@@ -865,6 +866,8 @@ void panic(const char *fmt, ...)
debugger_trap_immediate();
+ machine_crash_kexec();
+
if ( opt_noreboot )
{
machine_halt();
--- /dev/null
+++ work/xen/include/asm-ia64/elf.h
@@ -0,0 +1,25 @@
+#ifndef __IA64_ELF_H__
+#define __IA64_ELF_H__
+
+#include <xen/lib.h> /* for printk() used in stub */
+
+typedef struct {
+ unsigned long dummy;
+} ELF_Gregset;
+
+extern inline void elf_core_save_regs(ELF_Gregset *dst)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+#endif /* __IA64_ELF_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--- /dev/null
+++ work/xen/include/asm-ia64/kexec.h
@@ -0,0 +1,25 @@
+#ifndef __IA64_KEXEC_H__
+#define __IA64_KEXEC_H__
+
+#include <xen/lib.h> /* for printk() used in stub */
+#include <xen/types.h>
+#include <public/xen.h>
+#include <xen/kexec.h>
+
+static inline void machine_kexec(xen_kexec_image_t *image)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+#endif /* __IA64_KEXEC_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
+
--- /dev/null
+++ work/xen/include/asm-powerpc/elf.h
@@ -0,0 +1,25 @@
+#ifndef _ASM_ELF_H__
+#define _ASM_ELF_H__
+
+#include <xen/lib.h> /* for printk() used in stub */
+
+typedef struct {
+ unsigned long dummy;
+} ELF_Gregset;
+
+extern inline void elf_core_save_regs(ELF_Gregset *dst)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+#endif /* _ASM_ELF_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--- /dev/null
+++ work/xen/include/asm-powerpc/kexec.h
@@ -0,0 +1,25 @@
+#ifndef _ASM_KEXEC_H__
+#define _ASM_KEXEC_H__
+
+#include <xen/lib.h> /* for printk() used in stub */
+#include <xen/types.h>
+#include <public/xen.h>
+#include <xen/kexec.h>
+
+static inline void machine_kexec(xen_kexec_image_t *image)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+#endif /* _ASM_KEXEC_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
+
--- /dev/null
+++ work/xen/include/asm-x86/elf.h
@@ -0,0 +1,25 @@
+#ifndef __X86_ELF_H__
+#define __X86_ELF_H__
+
+#include <xen/lib.h> /* for printk() used in stub */
+
+typedef struct {
+ unsigned long dummy;
+} ELF_Gregset;
+
+extern inline void elf_core_save_regs(ELF_Gregset *dst)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+#endif /* __X86_ELF_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--- /dev/null
+++ work/xen/include/asm-x86/kexec.h
@@ -0,0 +1,24 @@
+#ifndef __X86_KEXEC_H__
+#define __X86_KEXEC_H__
+
+#include <xen/lib.h> /* for printk() used in stub */
+#include <xen/types.h>
+#include <public/xen.h>
+#include <xen/kexec.h>
+
+static inline void machine_kexec(xen_kexec_image_t *image)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+#endif /* __X86_KEXEC_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--- /dev/null
+++ work/xen/include/public/kexec.h
@@ -0,0 +1,131 @@
+/******************************************************************************
+ * kexec.h - Public portion
+ *
+ * Xen port written by:
+ * - Simon 'Horms' Horman <horms@verge.net.au>
+ * - Magnus Damm <magnus@valinux.co.jp>
+ */
+
+#ifndef _XEN_PUBLIC_KEXEC_H
+#define _XEN_PUBLIC_KEXEC_H
+
+
+/* This file describes the Kexec / Kdump hypercall interface for Xen.
+ *
+ * Kexec under vanilla Linux allows a user to reboot the physical machine
+ * into a new user-specified kernel. The Xen port extends this idea
+ * to allow rebooting of the machine from dom0. When kexec for dom0
+ * is used to reboot, both the hypervisor and the domains get replaced
+ * with some other kernel. It is possible to kexec between vanilla
+ * Linux and Xen and back again. Xen to Xen works well too.
+ *
+ * The hypercall interface for kexec can be divided into three main
+ * types of hypercall operations:
+ *
+ * 1) Range information:
+ * This is used by the dom0 kernel to ask the hypervisor about various
+ * address information. This information is needed to allow kexec-tools
+ * to fill in the ELF headers for /proc/vmcore properly.
+ *
+ * 2) Load and unload of images:
+ * There are no big surprises here, the kexec binary from kexec-tools
+ * runs in userspace in dom0. The tool loads/unloads data into the
+ * dom0 kernel such as new kernel, initramfs and hypervisor. When
+ * loaded the dom0 kernel performs a load hypercall operation, and
+ * before releasing all page references the dom0 kernel calls unload.
+ *
+ * 3) Kexec operation:
+ * This is used to start a previously loaded kernel.
+ */
+
+#include "xen.h"
+
+/*
+ * Prototype for this hypercall is:
+ * int kexec_op(int cmd, void *args)
+ * @cmd == KEXEC_CMD_...
+ * KEXEC operation to perform
+ * @args == Operation-specific extra arguments (NULL if none).
+ */
+
+/*
+ * Kexec supports two types of operation:
+ * - kexec into a regular kernel, very similar to a standard reboot
+ * - KEXEC_TYPE_DEFAULT is used to specify this type
+ * - kexec into a special "crash kernel", aka kexec-on-panic
+ * - KEXEC_TYPE_CRASH is used to specify this type
+ * - parts of our system may be broken at kexec-on-panic time
+ * - the code should be kept as simple and self-contained as possible
+ */
+
+#define KEXEC_TYPE_DEFAULT 0
+#define KEXEC_TYPE_CRASH 1
+
+
+/* The kexec implementation for Xen allows the user to load two
+ * types of kernels, KEXEC_TYPE_DEFAULT and KEXEC_TYPE_CRASH.
+ * All data needed for a kexec reboot is kept in one xen_kexec_image_t
+ * per "instance". The data mainly consists of machine address lists to pages
+ * together with destination addresses. The data in xen_kexec_image_t
+ * is passed to the "code page" which is one page of code that performs
+ * the final relocations before jumping to the new kernel.
+ */
+
+typedef struct xen_kexec_image {
+ unsigned long indirection_page;
+ unsigned long start_address;
+} xen_kexec_image_t;
+
+/*
+ * Perform kexec having previously loaded a kexec or kdump kernel
+ * as appropriate.
+ * type == KEXEC_TYPE_DEFAULT or KEXEC_TYPE_CRASH [in]
+ */
+#define KEXEC_CMD_kexec 0
+typedef struct xen_kexec_exec {
+ int type;
+} xen_kexec_exec_t;
+
+/*
+ * Load/Unload kernel image for kexec or kdump.
+ * type == KEXEC_TYPE_DEFAULT or KEXEC_TYPE_CRASH [in]
+ * image == relocation information for kexec (ignored for unload) [in]
+ */
+#define KEXEC_CMD_kexec_load 1
+#define KEXEC_CMD_kexec_unload 2
+typedef struct xen_kexec_load {
+ int type;
+ xen_kexec_image_t image;
+} xen_kexec_load_t;
+
+#define KEXEC_RANGE_MA_CRASH 0 /* machine address and size of crash area */
+#define KEXEC_RANGE_MA_XEN 1 /* machine address and size of Xen itself */
+#define KEXEC_RANGE_VA_XEN 2 /* virtual adrress and size of Xen itself */
+#define KEXEC_RANGE_MA_CPU 3 /* machine address and size of a CPU note */
+
+/*
+ * Find the address and size of certain memory areas
+ * range == KEXEC_RANGE_... [in]
+ * nr == physical CPU number (starting from 0) if KEXEC_RANGE_MA_CPU [in]
+ * size == number of bytes reserved in window [out]
+ * start == address of the first byte in the window [out]
+ */
+#define KEXEC_CMD_kexec_get_range 3
+typedef struct xen_kexec_range {
+ int range;
+ int nr;
+ unsigned long size;
+ unsigned long start;
+} xen_kexec_range_t;
+
+#endif /* _XEN_PUBLIC_KEXEC_H */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--- 0001/xen/include/xen/elf.h
+++ work/xen/include/xen/elf.h
@@ -452,18 +452,12 @@ unsigned int elf_hash(const unsigned cha
/*
* Note Definitions
*/
-typedef struct {
- Elf32_Word namesz;
- Elf32_Word descsz;
- Elf32_Word type;
-} Elf32_Note;
typedef struct {
- Elf64_Half namesz;
- Elf64_Half descsz;
- Elf64_Half type;
-} Elf64_Note;
-
+ u32 namesz;
+ u32 descsz;
+ u32 type;
+} Elf_Note; /* same format for both 32-bit and 64-bit ELF */
#if defined(ELFSIZE)
#define CONCAT(x,y) __CONCAT(x,y)
@@ -486,7 +480,6 @@ typedef struct {
#define Elf_Addr Elf32_Addr
#define Elf_Off Elf32_Off
#define Elf_Nhdr Elf32_Nhdr
-#define Elf_Note Elf32_Note
#define ELF_R_SYM ELF32_R_SYM
#define ELF_R_TYPE ELF32_R_TYPE
@@ -511,7 +504,6 @@ typedef struct {
#define Elf_Addr Elf64_Addr
#define Elf_Off Elf64_Off
#define Elf_Nhdr Elf64_Nhdr
-#define Elf_Note Elf64_Note
#define ELF_R_SYM ELF64_R_SYM
#define ELF_R_TYPE ELF64_R_TYPE
--- /dev/null
+++ work/xen/include/xen/elfcore.h
@@ -0,0 +1,127 @@
+/******************************************************************************
+ * elfcore.h
+ *
+ * Based heavily on include/linux/elfcore.h from Linux 2.6.16
+ * Naming scheeme based on include/xen/elf.h (not include/linux/elfcore.h)
+ *
+ */
+
+#ifndef __ELFCOREC_H__
+#define __ELFCOREC_H__
+
+#include <xen/types.h>
+#include <xen/elf.h>
+#include <asm/elf.h>
+#include <public/xen.h>
+
+#define NT_PRSTATUS 1
+
+typedef struct
+{
+ int signo; /* signal number */
+ int code; /* extra code */
+ int errno; /* errno */
+} ELF_Signifo;
+
+/* These seem to be the same length on all architectures on Linux */
+typedef int ELF_Pid;
+typedef struct {
+ long tv_sec;
+ long tv_usec;
+} ELF_Timeval;
+
+/*
+ * Definitions to generate Intel SVR4-like core files.
+ * These mostly have the same names as the SVR4 types with "elf_"
+ * tacked on the front to prevent clashes with linux definitions,
+ * and the typedef forms have been avoided. This is mostly like
+ * the SVR4 structure, but more Linuxy, with things that Linux does
+ * not support and which gdb doesn't really use excluded.
+ */
+typedef struct
+{
+ ELF_Signifo pr_info; /* Info associated with signal */
+ short pr_cursig; /* Current signal */
+ unsigned long pr_sigpend; /* Set of pending signals */
+ unsigned long pr_sighold; /* Set of held signals */
+ ELF_Pid pr_pid;
+ ELF_Pid pr_ppid;
+ ELF_Pid pr_pgrp;
+ ELF_Pid pr_sid;
+ ELF_Timeval pr_utime; /* User time */
+ ELF_Timeval pr_stime; /* System time */
+ ELF_Timeval pr_cutime; /* Cumulative user time */
+ ELF_Timeval pr_cstime; /* Cumulative system time */
+ ELF_Gregset pr_reg; /* GP registers - from asm header file */
+ int pr_fpvalid; /* True if math co-processor being used. */
+} ELF_Prstatus;
+
+/*
+ * The following data structures provide 64-bit ELF notes. In theory it should
+ * be possible to support both 64-bit and 32-bit ELF files, but to keep it
+ * simple we only do 64-bit.
+ *
+ * Please note that the current code aligns the 64-bit notes in the same
+ * way as Linux does. We are not following the 64-bit ELF spec, no one does.
+ *
+ * We are avoiding two problems by restricting us to 64-bit notes only:
+ * - Alignment of notes change with the word size. Ick.
+ * - We would need to tell kexec-tools which format we are using in the
+ * hypervisor to make sure the right ELF format is generated.
+ * That requires infrastructure. Let's not.
+ */
+
+#define ALIGN(x, n) ((x + ((1 << n) - 1)) / (1 << n))
+#define PAD32(x) u32 pad_data[ALIGN(x, 2)]
+
+#define TYPEDEF_NOTE(type, strlen, desctype) \
+ typedef struct { \
+ union { \
+ struct { \
+ Elf_Note note; \
+ unsigned char name[strlen]; \
+ } note; \
+ PAD32(sizeof(Elf_Note) + strlen); \
+ } note; \
+ union { \
+ desctype desc; \
+ PAD32(sizeof(desctype)); \
+ } desc; \
+ } __attribute__ ((packed)) type
+
+#define CORE_STR "CORE"
+#define CORE_STR_LEN 5 /* including terminating zero */
+
+TYPEDEF_NOTE(crash_note_core_t, CORE_STR_LEN, ELF_Prstatus);
+
+#define XEN_STR "XEN CORE"
+#define XEN_STR_LEN 9 /* including terminating zero */
+
+typedef struct {
+ unsigned long xen_major_version;
+ unsigned long xen_minor_version;
+ unsigned long xen_extra_version;
+ unsigned long xen_changeset;
+ unsigned long xen_compiler;
+ unsigned long xen_compile_date;
+ unsigned long xen_compile_time;
+} xen_crash_xen_regs_t;
+
+TYPEDEF_NOTE(crash_note_xen_t, XEN_STR_LEN, xen_crash_xen_regs_t);
+
+typedef struct {
+ crash_note_core_t core;
+ crash_note_xen_t xen;
+} __attribute__ ((packed)) crash_note_t;
+
+#endif /* __ELFCOREC_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--- 0001/xen/include/xen/hypercall.h
+++ work/xen/include/xen/hypercall.h
@@ -102,4 +102,10 @@ do_hvm_op(
unsigned long op,
XEN_GUEST_HANDLE(void) arg);
+extern long
+do_kexec_op(
+ unsigned long op,
+ int arg1,
+ XEN_GUEST_HANDLE(void) arg);
+
#endif /* __XEN_HYPERCALL_H__ */
--- /dev/null
+++ work/xen/include/xen/kexec.h
@@ -0,0 +1,43 @@
+#ifndef __XEN_KEXEC_H__
+#define __XEN_KEXEC_H__
+
+#include <public/kexec.h>
+#include <asm/percpu.h>
+#include <xen/elfcore.h>
+
+DECLARE_PER_CPU (crash_note_t, crash_notes);
+extern int crashing_cpu;
+
+typedef struct xen_kexec_reserve {
+ unsigned long size;
+ unsigned long start;
+} xen_kexec_reserve_t;
+
+/* We have space for 4 images to support atomic update
+ * of images. This is important for CRASH images since
+ * a panic can happen at any time...
+ */
+
+#define KEXEC_IMAGE_DEFAULT_BASE 0
+#define KEXEC_IMAGE_CRASH_BASE 2
+#define KEXEC_IMAGE_NR 4
+
+int machine_kexec_load(int type, int slot, xen_kexec_image_t *image);
+void machine_kexec_unload(int type, int slot, xen_kexec_image_t *image);
+void machine_kexec_reserved(xen_kexec_reserve_t *reservation);
+void machine_shutdown(xen_kexec_image_t *image);
+void machine_crash_kexec(void);
+void machine_crash_save_cpu(void);
+void machine_crash_shutdown(void);
+
+#endif /* __XEN_KEXEC_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--- 0001/xen/include/xen/mm.h
+++ work/xen/include/xen/mm.h
@@ -40,6 +40,7 @@ struct page_info;
paddr_t init_boot_allocator(paddr_t bitmap_start);
void init_boot_pages(paddr_t ps, paddr_t pe);
unsigned long alloc_boot_pages(unsigned long nr_pfns, unsigned long pfn_align);
+unsigned long alloc_boot_pages_at(unsigned long nr_pfns, unsigned long pfn_at);
void end_boot_allocator(void);
/* Generic allocator. These functions are *not* interrupt-safe. */
^ permalink raw reply [flat|nested] 30+ messages in thread* [PATCH 02/04] Kexec / Kdump: Code shared between x86_32 and x86_64
2006-11-22 7:10 [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502) Magnus Damm
2006-11-22 7:10 ` [PATCH 01/04] Kexec / Kdump: Generic code Magnus Damm
@ 2006-11-22 7:11 ` Magnus Damm
2006-11-22 7:11 ` [PATCH 03/04] Kexec / Kdump: x86_32 specific code Magnus Damm
` (3 subsequent siblings)
5 siblings, 0 replies; 30+ messages in thread
From: Magnus Damm @ 2006-11-22 7:11 UTC (permalink / raw)
To: Keir Fraser
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, magnus.damm,
Isaku Yamahata, Magnus Damm, Horms
[PATCH 02/04] Kexec / Kdump: Code shared between x86_32 and x86_64
This patch contains Kexec / Kdump code shared between x86_32 and x86_64.
Signed-Off-By: Magnus Damm <magnus@valinux.co.jp>
---
Applies on top of xen-unstable-12502.
patches/linux-2.6.16.29/git-2a...f7.patch | 62 ++++
patches/linux-2.6.16.29/git-2e...11.patch | 93 +++++++
patches/linux-2.6.16.29/series | 2
xen/arch/x86/crash.c | 130 +++++++++-
xen/arch/x86/machine_kexec.c | 83 +++++-
xen/arch/x86/setup.c | 73 ++++-
xen/arch/x86/traps.c | 2
xen/include/asm-x86/elf.h | 15 -
xen/include/asm-x86/fixmap.h | 4
xen/include/asm-x86/hypercall.h | 5
xen/include/asm-x86/kexec.h | 14 -
xen/include/asm-x86/x86_32/elf.h | 25 +
xen/include/asm-x86/x86_32/kexec.h | 24 +
xen/include/asm-x86/x86_64/elf.h | 25 +
xen/include/asm-x86/x86_64/kexec.h | 24 +
xen/include/public/kexec.h | 7
xen/include/xen/elfcore.h | 3
17 files changed, 550 insertions(+), 41 deletions(-)
--- /dev/null
+++ work/patches/linux-2.6.16.29/git-2a8a3d5b65e86ec1dfef7d268c64a909eab94af7.patch
@@ -0,0 +1,62 @@
+From: Eric W. Biederman <ebiederm@xmission.com>
+Date: Sun, 30 Jul 2006 10:03:20 +0000 (-0700)
+Subject: [PATCH] machine_kexec.c: Fix the description of segment handling
+X-Git-Tag: v2.6.18-rc4
+X-Git-Url: http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=2a8a3d5b65e86ec1dfef7d268c64a909eab94af7
+
+[PATCH] machine_kexec.c: Fix the description of segment handling
+
+One of my original comments in machine_kexec was unclear
+and this should fix it.
+
+Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
+Cc: Andi Kleen <ak@muc.de>
+Acked-by: Horms <horms@verge.net.au>
+Signed-off-by: Andrew Morton <akpm@osdl.org>
+Signed-off-by: Linus Torvalds <torvalds@osdl.org>
+---
+
+--- a/arch/i386/kernel/machine_kexec.c
++++ b/arch/i386/kernel/machine_kexec.c
+@@ -189,14 +189,11 @@ NORET_TYPE void machine_kexec(struct kim
+ memcpy((void *)reboot_code_buffer, relocate_new_kernel,
+ relocate_new_kernel_size);
+
+- /* The segment registers are funny things, they are
+- * automatically loaded from a table, in memory wherever you
+- * set them to a specific selector, but this table is never
+- * accessed again you set the segment to a different selector.
+- *
+- * The more common model is are caches where the behide
+- * the scenes work is done, but is also dropped at arbitrary
+- * times.
++ /* The segment registers are funny things, they have both a
++ * visible and an invisible part. Whenever the visible part is
++ * set to a specific selector, the invisible part is loaded
++ * with from a table in memory. At no other time is the
++ * descriptor table in memory accessed.
+ *
+ * I take advantage of this here by force loading the
+ * segments, before I zap the gdt with an invalid value.
+--- a/arch/x86_64/kernel/machine_kexec.c
++++ b/arch/x86_64/kernel/machine_kexec.c
+@@ -207,14 +207,11 @@ NORET_TYPE void machine_kexec(struct kim
+ __flush_tlb();
+
+
+- /* The segment registers are funny things, they are
+- * automatically loaded from a table, in memory wherever you
+- * set them to a specific selector, but this table is never
+- * accessed again unless you set the segment to a different selector.
+- *
+- * The more common model are caches where the behide
+- * the scenes work is done, but is also dropped at arbitrary
+- * times.
++ /* The segment registers are funny things, they have both a
++ * visible and an invisible part. Whenever the visible part is
++ * set to a specific selector, the invisible part is loaded
++ * with from a table in memory. At no other time is the
++ * descriptor table in memory accessed.
+ *
+ * I take advantage of this here by force loading the
+ * segments, before I zap the gdt with an invalid value.
--- /dev/null
+++ work/patches/linux-2.6.16.29/git-2efe55a9cec8418f0e0cde3dc3787a42fddc4411.patch
@@ -0,0 +1,93 @@
+From: Tobias Klauser <tklauser@nuerscht.ch>
+Date: Mon, 26 Jun 2006 16:57:34 +0000 (+0200)
+Subject: Storage class should be first
+X-Git-Tag: v2.6.18-rc1
+X-Git-Url: http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=2efe55a9cec8418f0e0cde3dc3787a42fddc4411
+
+Storage class should be first
+
+Storage class should be before const
+
+Signed-off-by: Tobias Klauser <tklauser@nuerscht.ch>
+Signed-off-by: Adrian Bunk <bunk@stusta.de>
+---
+
+--- a/arch/i386/kernel/machine_kexec.c
++++ b/arch/i386/kernel/machine_kexec.c
+@@ -133,9 +133,9 @@ typedef asmlinkage NORET_TYPE void (*rel
+ unsigned long start_address,
+ unsigned int has_pae) ATTRIB_NORET;
+
+-const extern unsigned char relocate_new_kernel[];
++extern const unsigned char relocate_new_kernel[];
+ extern void relocate_new_kernel_end(void);
+-const extern unsigned int relocate_new_kernel_size;
++extern const unsigned int relocate_new_kernel_size;
+
+ /*
+ * A architecture hook called to validate the
+--- a/arch/powerpc/kernel/machine_kexec_32.c
++++ b/arch/powerpc/kernel/machine_kexec_32.c
+@@ -30,8 +30,8 @@ typedef NORET_TYPE void (*relocate_new_k
+ */
+ void default_machine_kexec(struct kimage *image)
+ {
+- const extern unsigned char relocate_new_kernel[];
+- const extern unsigned int relocate_new_kernel_size;
++ extern const unsigned char relocate_new_kernel[];
++ extern const unsigned int relocate_new_kernel_size;
+ unsigned long page_list;
+ unsigned long reboot_code_buffer, reboot_code_buffer_phys;
+ relocate_new_kernel_t rnk;
+--- a/arch/ppc/kernel/machine_kexec.c
++++ b/arch/ppc/kernel/machine_kexec.c
+@@ -25,8 +25,8 @@ typedef NORET_TYPE void (*relocate_new_k
+ unsigned long reboot_code_buffer,
+ unsigned long start_address) ATTRIB_NORET;
+
+-const extern unsigned char relocate_new_kernel[];
+-const extern unsigned int relocate_new_kernel_size;
++extern const unsigned char relocate_new_kernel[];
++extern const unsigned int relocate_new_kernel_size;
+
+ void machine_shutdown(void)
+ {
+--- a/arch/s390/kernel/machine_kexec.c
++++ b/arch/s390/kernel/machine_kexec.c
+@@ -27,8 +27,8 @@ static void kexec_halt_all_cpus(void *);
+
+ typedef void (*relocate_kernel_t) (kimage_entry_t *, unsigned long);
+
+-const extern unsigned char relocate_kernel[];
+-const extern unsigned long long relocate_kernel_len;
++extern const unsigned char relocate_kernel[];
++extern const unsigned long long relocate_kernel_len;
+
+ int
+ machine_kexec_prepare(struct kimage *image)
+--- a/arch/sh/kernel/machine_kexec.c
++++ b/arch/sh/kernel/machine_kexec.c
+@@ -25,8 +25,8 @@ typedef NORET_TYPE void (*relocate_new_k
+ unsigned long start_address,
+ unsigned long vbr_reg) ATTRIB_NORET;
+
+-const extern unsigned char relocate_new_kernel[];
+-const extern unsigned int relocate_new_kernel_size;
++extern const unsigned char relocate_new_kernel[];
++extern const unsigned int relocate_new_kernel_size;
+ extern void *gdb_vbr_vector;
+
+ /*
+--- a/arch/x86_64/kernel/machine_kexec.c
++++ b/arch/x86_64/kernel/machine_kexec.c
+@@ -149,8 +149,8 @@ typedef NORET_TYPE void (*relocate_new_k
+ unsigned long start_address,
+ unsigned long pgtable) ATTRIB_NORET;
+
+-const extern unsigned char relocate_new_kernel[];
+-const extern unsigned long relocate_new_kernel_size;
++extern const unsigned char relocate_new_kernel[];
++extern const unsigned long relocate_new_kernel_size;
+
+ int machine_kexec_prepare(struct kimage *image)
+ {
--- 0003/patches/linux-2.6.16.29/series
+++ work/patches/linux-2.6.16.29/series
@@ -1,4 +1,6 @@
kexec-generic.patch
+git-2efe55a9cec8418f0e0cde3dc3787a42fddc4411.patch
+git-2a8a3d5b65e86ec1dfef7d268c64a909eab94af7.patch
blktap-aio-16_03_06.patch
device_bind.patch
fix-hz-suspend.patch
--- 0003/xen/arch/x86/crash.c
+++ work/xen/arch/x86/crash.c
@@ -1,10 +1,134 @@
-#include <xen/lib.h> /* for printk() used in stub */
+/******************************************************************************
+ * crash.c
+ *
+ * Based heavily on arch/i386/kernel/crash.c from Linux 2.6.16
+ *
+ * Xen port written by:
+ * - Simon 'Horms' Horman <horms@verge.net.au>
+ * - Magnus Damm <magnus@valinux.co.jp>
+ */
+
+#include <asm/atomic.h>
+#include <asm/elf.h>
+#include <asm/percpu.h>
+#include <asm/kexec.h>
#include <xen/types.h>
-#include <public/kexec.h>
+#include <xen/irq.h>
+#include <asm/ipi.h>
+#include <asm/nmi.h>
+#include <xen/string.h>
+#include <xen/elf.h>
+#include <xen/elfcore.h>
+#include <xen/smp.h>
+#include <xen/delay.h>
+#include <xen/perfc.h>
+#include <xen/kexec.h>
+#include <xen/sched.h>
+#include <xen/version.h>
+#include <public/xen.h>
+#include <asm/hvm/hvm.h>
+
+#ifdef CONFIG_SMP
+static atomic_t waiting_for_crash_ipi;
+
+static int crash_nmi_callback(struct cpu_user_regs *regs, int cpu)
+{
+ /* Don't do anything if this handler is invoked on crashing cpu.
+ * Otherwise, system will completely hang. Crashing cpu can get
+ * an NMI if system was initially booted with nmi_watchdog parameter.
+ */
+ if (cpu == crashing_cpu)
+ return 1;
+ local_irq_disable();
+
+ machine_crash_save_cpu();
+ disable_local_APIC();
+ atomic_dec(&waiting_for_crash_ipi);
+ hvm_disable();
+
+ for ( ; ; )
+ __asm__ __volatile__ ( "hlt" );
+
+ return 1;
+}
+
+/*
+ * By using the NMI code instead of a vector we just sneak thru the
+ * word generator coming out with just what we want. AND it does
+ * not matter if clustered_apic_mode is set or not.
+ */
+static void smp_send_nmi_allbutself(void)
+{
+ cpumask_t allbutself = cpu_online_map;
+
+ cpu_clear(smp_processor_id(), allbutself);
+ send_IPI_mask(allbutself, APIC_DM_NMI);
+}
+
+static void nmi_shootdown_cpus(void)
+{
+ unsigned long msecs;
+
+ atomic_set(&waiting_for_crash_ipi, num_online_cpus() - 1);
+ /* Would it be better to replace the trap vector here? */
+ set_nmi_callback(crash_nmi_callback);
+ /* Ensure the new callback function is set before sending
+ * out the NMI
+ */
+ wmb();
+
+ smp_send_nmi_allbutself();
+
+ msecs = 1000; /* Wait at most a second for the other cpus to stop */
+ while ((atomic_read(&waiting_for_crash_ipi) > 0) && msecs) {
+ mdelay(1);
+ msecs--;
+ }
+
+ /* Leave the nmi callback set */
+ disable_local_APIC();
+}
+#endif
+
+static void crash_save_xen_notes(void)
+{
+ int cpu = smp_processor_id();
+ crash_note_t *cntp = &per_cpu(crash_notes, cpu);
+
+ /* this code assumes that the first note has been written already */
+
+ cntp->xen.note.note.note.namesz = XEN_STR_LEN;
+ cntp->xen.note.note.note.descsz = sizeof(xen_crash_xen_regs_t);
+ cntp->xen.note.note.note.type = 0x10000001; /* NT_XEN_DOM0_CR3 */
+ memcpy(cntp->xen.note.note.name, XEN_STR, XEN_STR_LEN);
+
+ cntp->xen.desc.desc.xen_major_version = xen_major_version();
+ cntp->xen.desc.desc.xen_minor_version = xen_minor_version();
+ cntp->xen.desc.desc.xen_extra_version = __pa(xen_extra_version());
+ cntp->xen.desc.desc.xen_changeset = __pa(xen_changeset());
+ cntp->xen.desc.desc.xen_compiler = __pa(xen_compiler());
+ cntp->xen.desc.desc.xen_compile_date = __pa(xen_compile_date());
+ cntp->xen.desc.desc.xen_compile_time = __pa(xen_compile_time());
+
+ cntp->xen.desc.desc.dom0_pfn_to_mfn_frame_list_list = \
+ dom0->shared_info->arch.pfn_to_mfn_frame_list_list;
+}
void machine_crash_shutdown(void)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ printk("machine_crash_shutdown: %d\n", smp_processor_id());
+ local_irq_disable();
+
+#ifdef CONFIG_SMP
+ nmi_shootdown_cpus();
+#endif
+
+#ifdef CONFIG_X86_IO_APIC
+ disable_IO_APIC();
+#endif
+ hvm_disable();
+
+ crash_save_xen_notes();
}
/*
--- 0003/xen/arch/x86/machine_kexec.c
+++ work/xen/arch/x86/machine_kexec.c
@@ -1,26 +1,89 @@
-#include <xen/lib.h> /* for printk() used in stubs */
+/******************************************************************************
+ * machine_kexec.c
+ *
+ * Xen port written by:
+ * - Simon 'Horms' Horman <horms@verge.net.au>
+ * - Magnus Damm <magnus@valinux.co.jp>
+ */
+
+#include <xen/lib.h>
+#include <asm/irq.h>
+#include <asm/page.h>
+#include <asm/flushtlb.h>
+#include <xen/smp.h>
+#include <xen/nmi.h>
#include <xen/types.h>
-#include <public/kexec.h>
+#include <xen/console.h>
+#include <xen/kexec.h>
+#include <asm/kexec.h>
+#include <xen/domain_page.h>
+#include <asm/fixmap.h>
+#include <asm/hvm/hvm.h>
int machine_kexec_load(int type, int slot, xen_kexec_image_t *image)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
- return -1;
+ unsigned long prev_ma = 0;
+ int fix_base = FIX_KEXEC_BASE_0 + (slot * (KEXEC_XEN_NO_PAGES >> 1));
+ int k;
+
+ /* setup fixmap to point to our pages and record the virtual address
+ * in every odd index in page_list[].
+ */
+
+ for (k = 0; k < KEXEC_XEN_NO_PAGES; k++) {
+ if ((k & 1) == 0) { /* even pages: machine address */
+ prev_ma = image->page_list[k];
+ }
+ else { /* odd pages: va for previous ma */
+ set_fixmap(fix_base + (k >> 1), prev_ma);
+ image->page_list[k] = fix_to_virt(fix_base + (k >> 1));
+ }
+ }
+
+ return 0;
}
void machine_kexec_unload(int type, int slot, xen_kexec_image_t *image)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
}
-
-void machine_kexec(xen_kexec_image_t *image)
+
+static void __machine_shutdown(void *data)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
-}
+ xen_kexec_image_t *image = (xen_kexec_image_t *)data;
+ watchdog_disable();
+ console_start_sync();
+
+ smp_send_stop();
+
+#ifdef CONFIG_X86_IO_APIC
+ disable_IO_APIC();
+#endif
+ hvm_disable();
+
+ machine_kexec(image);
+}
+
void machine_shutdown(xen_kexec_image_t *image)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ int reboot_cpu_id;
+ cpumask_t reboot_cpu;
+
+ reboot_cpu_id = 0;
+
+ if (!cpu_isset(reboot_cpu_id, cpu_online_map))
+ reboot_cpu_id = smp_processor_id();
+
+ if (reboot_cpu_id != smp_processor_id()) {
+ cpus_clear(reboot_cpu);
+ cpu_set(reboot_cpu_id, reboot_cpu);
+ on_selected_cpus(reboot_cpu, __machine_shutdown, image, 1, 0);
+ for (;;)
+ ; /* nothing */
+ }
+ else
+ __machine_shutdown(image);
+ BUG();
}
/*
--- 0001/xen/arch/x86/setup.c
+++ work/xen/arch/x86/setup.c
@@ -27,6 +27,7 @@
#include <asm/shadow.h>
#include <asm/e820.h>
#include <acm/acm_hooks.h>
+#include <xen/kexec.h>
extern void dmi_scan_machine(void);
extern void generic_apic_probe(void);
@@ -273,6 +274,20 @@ static void srat_detect_node(int cpu)
printk(KERN_INFO "CPU %d APIC %d -> Node %d\n", cpu, apicid, node);
}
+void __init move_memory(unsigned long dst,
+ unsigned long src_start, unsigned long src_end)
+{
+#if defined(CONFIG_X86_32)
+ memmove((void *)dst, /* use low mapping */
+ (void *)src_start, /* use low mapping */
+ src_end - src_start);
+#elif defined(CONFIG_X86_64)
+ memmove(__va(dst),
+ __va(src_start),
+ src_end - src_start);
+#endif
+}
+
void __init __start_xen(multiboot_info_t *mbi)
{
char __cmdline[] = "", *cmdline = __cmdline;
@@ -284,6 +299,7 @@ void __init __start_xen(multiboot_info_t
unsigned long nr_pages, modules_length;
paddr_t s, e;
int i, e820_warn = 0, e820_raw_nr = 0, bytes = 0;
+ xen_kexec_reserve_t crash_area;
struct ns16550_defaults ns16550 = {
.data_bits = 8,
.parity = 'n',
@@ -415,15 +431,8 @@ void __init __start_xen(multiboot_info_t
initial_images_start = xenheap_phys_end;
initial_images_end = initial_images_start + modules_length;
-#if defined(CONFIG_X86_32)
- memmove((void *)initial_images_start, /* use low mapping */
- (void *)mod[0].mod_start, /* use low mapping */
- mod[mbi->mods_count-1].mod_end - mod[0].mod_start);
-#elif defined(CONFIG_X86_64)
- memmove(__va(initial_images_start),
- __va(mod[0].mod_start),
- mod[mbi->mods_count-1].mod_end - mod[0].mod_start);
-#endif
+ move_memory(initial_images_start,
+ mod[0].mod_start, mod[mbi->mods_count-1].mod_end);
/* Initialise boot-time allocator with all RAM situated after modules. */
xenheap_phys_start = init_boot_allocator(__pa(&_end));
@@ -471,6 +480,52 @@ void __init __start_xen(multiboot_info_t
#endif
}
+ machine_kexec_reserved(&crash_area);
+ if (crash_area.size > 0) {
+ unsigned long kdump_start, kdump_size, k;
+
+ /* mark images pages as free for now */
+
+ init_boot_pages(initial_images_start, initial_images_end);
+
+ kdump_start = crash_area.start;
+ kdump_size = crash_area.size;
+
+ printk("Kdump: %luMB (%lukB) at 0x%lx\n",
+ kdump_size >> 20,
+ kdump_size >> 10,
+ kdump_start);
+
+ if ((kdump_start & ~PAGE_MASK) || (kdump_size & ~PAGE_MASK))
+ panic("Kdump parameters not page aligned\n");
+
+ kdump_start >>= PAGE_SHIFT;
+ kdump_size >>= PAGE_SHIFT;
+
+ /* allocate pages for Kdump memory area */
+
+ k = alloc_boot_pages_at(kdump_size, kdump_start);
+
+ if (k != kdump_start)
+ panic("Unable to reserve Kdump memory\n");
+
+ /* allocate pages for relocated initial images */
+
+ k = ((initial_images_end - initial_images_start) & ~PAGE_MASK) ? 1 : 0;
+ k += (initial_images_end - initial_images_start) >> PAGE_SHIFT;
+
+ k = alloc_boot_pages(k, 1);
+
+ if (!k)
+ panic("Unable to allocate initial images memory\n");
+
+ move_memory(k << PAGE_SHIFT, initial_images_start, initial_images_end);
+
+ initial_images_end -= initial_images_start;
+ initial_images_start = k << PAGE_SHIFT;
+ initial_images_end += initial_images_start;
+ }
+
memguard_init();
percpu_guard_areas();
--- 0001/xen/arch/x86/traps.c
+++ work/xen/arch/x86/traps.c
@@ -45,6 +45,7 @@
#include <xen/iocap.h>
#include <xen/nmi.h>
#include <xen/version.h>
+#include <xen/kexec.h>
#include <asm/shadow.h>
#include <asm/system.h>
#include <asm/io.h>
@@ -1633,6 +1634,7 @@ static void unknown_nmi_error(unsigned c
printk("Uhhuh. NMI received for unknown reason %02x.\n", reason);
printk("Dazed and confused, but trying to continue\n");
printk("Do you have a strange power saving mode enabled?\n");
+ machine_crash_kexec();
}
}
--- 0003/xen/include/asm-x86/elf.h
+++ work/xen/include/asm-x86/elf.h
@@ -1,16 +1,11 @@
#ifndef __X86_ELF_H__
#define __X86_ELF_H__
-#include <xen/lib.h> /* for printk() used in stub */
-
-typedef struct {
- unsigned long dummy;
-} ELF_Gregset;
-
-extern inline void elf_core_save_regs(ELF_Gregset *dst)
-{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
-}
+#ifdef __x86_64__
+#include <asm/x86_64/elf.h>
+#else
+#include <asm/x86_32/elf.h>
+#endif
#endif /* __X86_ELF_H__ */
--- 0001/xen/include/asm-x86/fixmap.h
+++ work/xen/include/asm-x86/fixmap.h
@@ -16,6 +16,7 @@
#include <asm/apicdef.h>
#include <asm/acpi.h>
#include <asm/page.h>
+#include <xen/kexec.h>
/*
* Here we define all the compile-time 'special' virtual
@@ -36,6 +37,9 @@ enum fixed_addresses {
FIX_ACPI_END = FIX_ACPI_BEGIN + FIX_ACPI_PAGES - 1,
FIX_HPET_BASE,
FIX_CYCLONE_TIMER,
+ FIX_KEXEC_BASE_0,
+ FIX_KEXEC_BASE_END = FIX_KEXEC_BASE_0 \
+ + ((KEXEC_XEN_NO_PAGES >> 1) * KEXEC_IMAGE_NR) - 1,
__end_of_fixed_addresses
};
--- 0001/xen/include/asm-x86/hypercall.h
+++ work/xen/include/asm-x86/hypercall.h
@@ -6,6 +6,7 @@
#define __ASM_X86_HYPERCALL_H__
#include <public/physdev.h>
+#include <xen/types.h>
extern long
do_event_channel_op_compat(
@@ -87,6 +88,10 @@ extern long
arch_do_vcpu_op(
int cmd, struct vcpu *v, XEN_GUEST_HANDLE(void) arg);
+extern int
+do_kexec(
+ unsigned long op, unsigned arg1, XEN_GUEST_HANDLE(void) uarg);
+
#ifdef __x86_64__
extern long
--- 0003/xen/include/asm-x86/kexec.h
+++ work/xen/include/asm-x86/kexec.h
@@ -1,15 +1,11 @@
#ifndef __X86_KEXEC_H__
#define __X86_KEXEC_H__
-#include <xen/lib.h> /* for printk() used in stub */
-#include <xen/types.h>
-#include <public/xen.h>
-#include <xen/kexec.h>
-
-static inline void machine_kexec(xen_kexec_image_t *image)
-{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
-}
+#ifdef __x86_64__
+#include <asm/x86_64/kexec.h>
+#else
+#include <asm/x86_32/kexec.h>
+#endif
#endif /* __X86_KEXEC_H__ */
--- /dev/null
+++ work/xen/include/asm-x86/x86_32/elf.h
@@ -0,0 +1,25 @@
+#ifndef __X86_32_ELF_H__
+#define __X86_32_ELF_H__
+
+#include <xen/lib.h> /* for printk() used in stub */
+
+typedef struct {
+ unsigned long dummy;
+} ELF_Gregset;
+
+extern inline void elf_core_save_regs(ELF_Gregset *dst)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+#endif /* __X86_32_ELF_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--- /dev/null
+++ work/xen/include/asm-x86/x86_32/kexec.h
@@ -0,0 +1,24 @@
+#ifndef __X86_32_KEXEC_H__
+#define __X86_32_KEXEC_H__
+
+#include <xen/lib.h> /* for printk() used in stub */
+#include <xen/types.h>
+#include <public/xen.h>
+#include <xen/kexec.h>
+
+static inline void machine_kexec(xen_kexec_image_t *image)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+#endif /* __X86_32_KEXEC_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--- /dev/null
+++ work/xen/include/asm-x86/x86_64/elf.h
@@ -0,0 +1,25 @@
+#ifndef __X86_64_ELF_H__
+#define __X86_64_ELF_H__
+
+#include <xen/lib.h> /* for printk() used in stub */
+
+typedef struct {
+ unsigned long dummy;
+} ELF_Gregset;
+
+extern inline void elf_core_save_regs(ELF_Gregset *dst)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+#endif /* __X86_64_ELF_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--- /dev/null
+++ work/xen/include/asm-x86/x86_64/kexec.h
@@ -0,0 +1,24 @@
+#ifndef __X86_64_KEXEC_H__
+#define __X86_64_KEXEC_H__
+
+#include <xen/lib.h> /* for printk() used in stub */
+#include <xen/types.h>
+#include <public/xen.h>
+#include <xen/kexec.h>
+
+static inline void machine_kexec(xen_kexec_image_t *image)
+{
+ printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+}
+
+#endif /* __X86_64_KEXEC_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--- 0003/xen/include/public/kexec.h
+++ work/xen/include/public/kexec.h
@@ -40,6 +40,10 @@
#include "xen.h"
+#if defined(__i386__) || defined(__x86_64__)
+#define KEXEC_XEN_NO_PAGES 17
+#endif
+
/*
* Prototype for this hypercall is:
* int kexec_op(int cmd, void *args)
@@ -72,6 +76,9 @@
*/
typedef struct xen_kexec_image {
+#if defined(__i386__) || defined(__x86_64__)
+ unsigned long page_list[KEXEC_XEN_NO_PAGES];
+#endif
unsigned long indirection_page;
unsigned long start_address;
} xen_kexec_image_t;
--- 0003/xen/include/xen/elfcore.h
+++ work/xen/include/xen/elfcore.h
@@ -105,6 +105,9 @@ typedef struct {
unsigned long xen_compiler;
unsigned long xen_compile_date;
unsigned long xen_compile_time;
+#ifdef CONFIG_X86
+ unsigned long dom0_pfn_to_mfn_frame_list_list;
+#endif
} xen_crash_xen_regs_t;
TYPEDEF_NOTE(crash_note_xen_t, XEN_STR_LEN, xen_crash_xen_regs_t);
^ permalink raw reply [flat|nested] 30+ messages in thread* [PATCH 03/04] Kexec / Kdump: x86_32 specific code
2006-11-22 7:10 [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502) Magnus Damm
2006-11-22 7:10 ` [PATCH 01/04] Kexec / Kdump: Generic code Magnus Damm
2006-11-22 7:11 ` [PATCH 02/04] Kexec / Kdump: Code shared between x86_32 and x86_64 Magnus Damm
@ 2006-11-22 7:11 ` Magnus Damm
2006-11-22 7:11 ` [PATCH 04/04] Kexec / Kdump: x86_64 " Magnus Damm
` (2 subsequent siblings)
5 siblings, 0 replies; 30+ messages in thread
From: Magnus Damm @ 2006-11-22 7:11 UTC (permalink / raw)
To: Keir Fraser
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, magnus.damm,
Isaku Yamahata, Magnus Damm, Horms
[PATCH 03/04] Kexec / Kdump: x86_32 specific code
This patch contains the x86_32 implementation of Kexec / Kdump for Xen.
Signed-Off-By: Magnus Damm <magnus@valinux.co.jp>
---
Applies on top of xen-unstable-12502.
buildconfigs/linux-defconfig_xen_x86_32 | 2
linux-2.6-xen-sparse/arch/i386/Kconfig | 2
linux-2.6-xen-sparse/arch/i386/kernel/setup-xen.c | 19
linux-2.6-xen-sparse/include/asm-i386/kexec-xen.h | 30
linux-2.6-xen-sparse/include/asm-i386/mach-xen/asm/hypercall.h | 8
patches/linux-2.6.16.29/git-35...c9.patch | 401 ++++++
patches/linux-2.6.16.29/linux-2.6.16.29-crash-xen-i386.patch | 31
patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec...code-i386.patch | 169 +++
patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-xen-i386.patch | 54 +
patches/linux-2.6.16.29/series | 4
xen/arch/x86/x86_32/entry.S | 2
xen/include/asm-x86/x86_32/elf.h | 38
xen/include/asm-x86/x86_32/kexec.h | 28
13 files changed, 777 insertions(+), 11 deletions(-)
--- 0002/buildconfigs/linux-defconfig_xen_x86_32
+++ work/buildconfigs/linux-defconfig_xen_x86_32
@@ -183,6 +183,7 @@ CONFIG_MTRR=y
CONFIG_REGPARM=y
CONFIG_SECCOMP=y
CONFIG_HZ_100=y
+CONFIG_KEXEC=y
# CONFIG_HZ_250 is not set
# CONFIG_HZ_1000 is not set
CONFIG_HZ=100
@@ -1036,6 +1037,7 @@ CONFIG_DNOTIFY=y
#
CONFIG_PROC_FS=y
CONFIG_PROC_KCORE=y
+# CONFIG_PROC_VMCORE is not set
CONFIG_SYSFS=y
CONFIG_TMPFS=y
# CONFIG_HUGETLB_PAGE is not set
--- 0001/linux-2.6-xen-sparse/arch/i386/Kconfig
+++ work/linux-2.6-xen-sparse/arch/i386/Kconfig
@@ -726,7 +726,7 @@ source kernel/Kconfig.hz
config KEXEC
bool "kexec system call (EXPERIMENTAL)"
- depends on EXPERIMENTAL && !X86_XEN
+ depends on EXPERIMENTAL && !XEN_UNPRIVILEGED_GUEST
help
kexec is a system call that implements the ability to shutdown your
current kernel, and to start another kernel. It is like a reboot
--- 0001/linux-2.6-xen-sparse/arch/i386/kernel/setup-xen.c
+++ work/linux-2.6-xen-sparse/arch/i386/kernel/setup-xen.c
@@ -69,6 +69,10 @@
#include "setup_arch_pre.h"
#include <bios_ebda.h>
+#ifdef CONFIG_XEN
+#include <xen/interface/kexec.h>
+#endif
+
/* Forward Declaration. */
void __init find_max_pfn(void);
@@ -943,6 +947,7 @@ static void __init parse_cmdline_early (
* after a kernel panic.
*/
else if (!memcmp(from, "crashkernel=", 12)) {
+#ifndef CONFIG_XEN
unsigned long size, base;
size = memparse(from+12, &from);
if (*from == '@') {
@@ -953,6 +958,10 @@ static void __init parse_cmdline_early (
crashk_res.start = base;
crashk_res.end = base + size - 1;
}
+#else
+ printk("Ignoring crashkernel command line, "
+ "parameter will be supplied by xen\n");
+#endif
}
#endif
#ifdef CONFIG_PROC_VMCORE
@@ -1322,10 +1331,14 @@ void __init setup_bootmem_allocator(void
}
#endif
#ifdef CONFIG_KEXEC
+#ifdef CONFIG_XEN
+ xen_machine_kexec_setup_resources();
+#else
if (crashk_res.start != crashk_res.end)
reserve_bootmem(crashk_res.start,
crashk_res.end - crashk_res.start + 1);
#endif
+#endif
if (!xen_feature(XENFEAT_auto_translated_physmap))
phys_to_machine_mapping =
@@ -1389,7 +1402,11 @@ legacy_init_iomem_resources(struct e820e
request_resource(res, data_resource);
#endif
#ifdef CONFIG_KEXEC
- request_resource(res, &crashk_res);
+ if (crashk_res.start != crashk_res.end)
+ request_resource(res, &crashk_res);
+#ifdef CONFIG_XEN
+ xen_machine_kexec_register_resources(res);
+#endif
#endif
}
}
--- /dev/null
+++ work/linux-2.6-xen-sparse/include/asm-i386/kexec-xen.h
@@ -0,0 +1,30 @@
+#ifndef _I386_KEXEC_XEN_H
+#define _I386_KEXEC_XEN_H
+
+/* Kexec needs to know about the actual physical addresss.
+ * But in xen, on some architectures, a physical address is a
+ * pseudo-physical addresss. */
+
+#ifdef CONFIG_XEN
+#define kexec_page_to_pfn(page) pfn_to_mfn(page_to_pfn(page))
+#define kexec_pfn_to_page(pfn) pfn_to_page(mfn_to_pfn(pfn))
+#define kexec_virt_to_phys(addr) virt_to_machine(addr)
+#define kexec_phys_to_virt(addr) phys_to_virt(machine_to_phys(addr))
+#else
+#define kexec_page_to_pfn(page) page_to_pfn(page)
+#define kexec_pfn_to_page(pfn) pfn_to_page(pfn)
+#define kexec_virt_to_phys(addr) virt_to_phys(addr)
+#define kexec_phys_to_virt(addr) phys_to_virt(addr)
+#endif
+
+#endif /* _I386_KEXEC_XEN_H */
+
+/*
+ * Local variables:
+ * c-file-style: "linux"
+ * indent-tabs-mode: t
+ * c-indent-level: 8
+ * c-basic-offset: 8
+ * tab-width: 8
+ * End:
+ */
--- 0001/linux-2.6-xen-sparse/include/asm-i386/mach-xen/asm/hypercall.h
+++ work/linux-2.6-xen-sparse/include/asm-i386/mach-xen/asm/hypercall.h
@@ -395,5 +395,13 @@ HYPERVISOR_xenoprof_op(
return _hypercall2(int, xenoprof_op, op, arg);
}
+static inline int
+HYPERVISOR_kexec_op(
+ unsigned long op, void *args)
+{
+ return _hypercall2(int, kexec_op, op, args);
+}
+
+
#endif /* __HYPERCALL_H__ */
--- /dev/null
+++ work/patches/linux-2.6.16.29/git-3566561bfadffcb5dbc85d576be80c0dbf2cccc9.patch
@@ -0,0 +1,401 @@
+From: Magnus Damm <magnus@valinux.co.jp>
+Date: Tue, 26 Sep 2006 08:52:38 +0000 (+0200)
+Subject: [PATCH] i386: Avoid overwriting the current pgd (V4, i386)
+X-Git-Url: http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=3566561bfadffcb5dbc85d576be80c0dbf2cccc9
+
+[PATCH] i386: Avoid overwriting the current pgd (V4, i386)
+
+kexec: Avoid overwriting the current pgd (V4, i386)
+
+This patch upgrades the i386-specific kexec code to avoid overwriting the
+current pgd. Overwriting the current pgd is bad when CONFIG_CRASH_DUMP is used
+to start a secondary kernel that dumps the memory of the previous kernel.
+
+The code introduces a new set of page tables. These tables are used to provide
+an executable identity mapping without overwriting the current pgd.
+
+Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
+Signed-off-by: Andi Kleen <ak@suse.de>
+---
+
+--- a/arch/i386/kernel/machine_kexec.c
++++ b/arch/i386/kernel/machine_kexec.c
+@@ -21,70 +21,13 @@
+ #include <asm/system.h>
+
+ #define PAGE_ALIGNED __attribute__ ((__aligned__(PAGE_SIZE)))
+-
+-#define L0_ATTR (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY)
+-#define L1_ATTR (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY)
+-#define L2_ATTR (_PAGE_PRESENT)
+-
+-#define LEVEL0_SIZE (1UL << 12UL)
+-
+-#ifndef CONFIG_X86_PAE
+-#define LEVEL1_SIZE (1UL << 22UL)
+-static u32 pgtable_level1[1024] PAGE_ALIGNED;
+-
+-static void identity_map_page(unsigned long address)
+-{
+- unsigned long level1_index, level2_index;
+- u32 *pgtable_level2;
+-
+- /* Find the current page table */
+- pgtable_level2 = __va(read_cr3());
+-
+- /* Find the indexes of the physical address to identity map */
+- level1_index = (address % LEVEL1_SIZE)/LEVEL0_SIZE;
+- level2_index = address / LEVEL1_SIZE;
+-
+- /* Identity map the page table entry */
+- pgtable_level1[level1_index] = address | L0_ATTR;
+- pgtable_level2[level2_index] = __pa(pgtable_level1) | L1_ATTR;
+-
+- /* Flush the tlb so the new mapping takes effect.
+- * Global tlb entries are not flushed but that is not an issue.
+- */
+- load_cr3(pgtable_level2);
+-}
+-
+-#else
+-#define LEVEL1_SIZE (1UL << 21UL)
+-#define LEVEL2_SIZE (1UL << 30UL)
+-static u64 pgtable_level1[512] PAGE_ALIGNED;
+-static u64 pgtable_level2[512] PAGE_ALIGNED;
+-
+-static void identity_map_page(unsigned long address)
+-{
+- unsigned long level1_index, level2_index, level3_index;
+- u64 *pgtable_level3;
+-
+- /* Find the current page table */
+- pgtable_level3 = __va(read_cr3());
+-
+- /* Find the indexes of the physical address to identity map */
+- level1_index = (address % LEVEL1_SIZE)/LEVEL0_SIZE;
+- level2_index = (address % LEVEL2_SIZE)/LEVEL1_SIZE;
+- level3_index = address / LEVEL2_SIZE;
+-
+- /* Identity map the page table entry */
+- pgtable_level1[level1_index] = address | L0_ATTR;
+- pgtable_level2[level2_index] = __pa(pgtable_level1) | L1_ATTR;
+- set_64bit(&pgtable_level3[level3_index],
+- __pa(pgtable_level2) | L2_ATTR);
+-
+- /* Flush the tlb so the new mapping takes effect.
+- * Global tlb entries are not flushed but that is not an issue.
+- */
+- load_cr3(pgtable_level3);
+-}
++static u32 kexec_pgd[1024] PAGE_ALIGNED;
++#ifdef CONFIG_X86_PAE
++static u32 kexec_pmd0[1024] PAGE_ALIGNED;
++static u32 kexec_pmd1[1024] PAGE_ALIGNED;
+ #endif
++static u32 kexec_pte0[1024] PAGE_ALIGNED;
++static u32 kexec_pte1[1024] PAGE_ALIGNED;
+
+ static void set_idt(void *newidt, __u16 limit)
+ {
+@@ -128,16 +71,6 @@ static void load_segments(void)
+ #undef __STR
+ }
+
+-typedef asmlinkage NORET_TYPE void (*relocate_new_kernel_t)(
+- unsigned long indirection_page,
+- unsigned long reboot_code_buffer,
+- unsigned long start_address,
+- unsigned int has_pae) ATTRIB_NORET;
+-
+-extern const unsigned char relocate_new_kernel[];
+-extern void relocate_new_kernel_end(void);
+-extern const unsigned int relocate_new_kernel_size;
+-
+ /*
+ * A architecture hook called to validate the
+ * proposed image and prepare the control pages
+@@ -170,25 +103,29 @@ void machine_kexec_cleanup(struct kimage
+ */
+ NORET_TYPE void machine_kexec(struct kimage *image)
+ {
+- unsigned long page_list;
+- unsigned long reboot_code_buffer;
+-
+- relocate_new_kernel_t rnk;
++ unsigned long page_list[PAGES_NR];
++ void *control_page;
+
+ /* Interrupts aren't acceptable while we reboot */
+ local_irq_disable();
+
+- /* Compute some offsets */
+- reboot_code_buffer = page_to_pfn(image->control_code_page)
+- << PAGE_SHIFT;
+- page_list = image->head;
+-
+- /* Set up an identity mapping for the reboot_code_buffer */
+- identity_map_page(reboot_code_buffer);
+-
+- /* copy it out */
+- memcpy((void *)reboot_code_buffer, relocate_new_kernel,
+- relocate_new_kernel_size);
++ control_page = page_address(image->control_code_page);
++ memcpy(control_page, relocate_kernel, PAGE_SIZE);
++
++ page_list[PA_CONTROL_PAGE] = __pa(control_page);
++ page_list[VA_CONTROL_PAGE] = (unsigned long)relocate_kernel;
++ page_list[PA_PGD] = __pa(kexec_pgd);
++ page_list[VA_PGD] = (unsigned long)kexec_pgd;
++#ifdef CONFIG_X86_PAE
++ page_list[PA_PMD_0] = __pa(kexec_pmd0);
++ page_list[VA_PMD_0] = (unsigned long)kexec_pmd0;
++ page_list[PA_PMD_1] = __pa(kexec_pmd1);
++ page_list[VA_PMD_1] = (unsigned long)kexec_pmd1;
++#endif
++ page_list[PA_PTE_0] = __pa(kexec_pte0);
++ page_list[VA_PTE_0] = (unsigned long)kexec_pte0;
++ page_list[PA_PTE_1] = __pa(kexec_pte1);
++ page_list[VA_PTE_1] = (unsigned long)kexec_pte1;
+
+ /* The segment registers are funny things, they have both a
+ * visible and an invisible part. Whenever the visible part is
+@@ -207,8 +144,8 @@ NORET_TYPE void machine_kexec(struct kim
+ set_idt(phys_to_virt(0),0);
+
+ /* now call it */
+- rnk = (relocate_new_kernel_t) reboot_code_buffer;
+- (*rnk)(page_list, reboot_code_buffer, image->start, cpu_has_pae);
++ relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
++ image->start, cpu_has_pae);
+ }
+
+ /* crashkernel=size@addr specifies the location to reserve for
+--- a/arch/i386/kernel/relocate_kernel.S
++++ b/arch/i386/kernel/relocate_kernel.S
+@@ -7,16 +7,138 @@
+ */
+
+ #include <linux/linkage.h>
++#include <asm/page.h>
++#include <asm/kexec.h>
++
++/*
++ * Must be relocatable PIC code callable as a C function
++ */
++
++#define PTR(x) (x << 2)
++#define PAGE_ALIGNED (1 << PAGE_SHIFT)
++#define PAGE_ATTR 0x63 /* _PAGE_PRESENT|_PAGE_RW|_PAGE_ACCESSED|_PAGE_DIRTY */
++#define PAE_PGD_ATTR 0x01 /* _PAGE_PRESENT */
++
++ .text
++ .align PAGE_ALIGNED
++ .globl relocate_kernel
++relocate_kernel:
++ movl 8(%esp), %ebp /* list of pages */
++
++#ifdef CONFIG_X86_PAE
++ /* map the control page at its virtual address */
++
++ movl PTR(VA_PGD)(%ebp), %edi
++ movl PTR(VA_CONTROL_PAGE)(%ebp), %eax
++ andl $0xc0000000, %eax
++ shrl $27, %eax
++ addl %edi, %eax
++
++ movl PTR(PA_PMD_0)(%ebp), %edx
++ orl $PAE_PGD_ATTR, %edx
++ movl %edx, (%eax)
++
++ movl PTR(VA_PMD_0)(%ebp), %edi
++ movl PTR(VA_CONTROL_PAGE)(%ebp), %eax
++ andl $0x3fe00000, %eax
++ shrl $18, %eax
++ addl %edi, %eax
++
++ movl PTR(PA_PTE_0)(%ebp), %edx
++ orl $PAGE_ATTR, %edx
++ movl %edx, (%eax)
++
++ movl PTR(VA_PTE_0)(%ebp), %edi
++ movl PTR(VA_CONTROL_PAGE)(%ebp), %eax
++ andl $0x001ff000, %eax
++ shrl $9, %eax
++ addl %edi, %eax
++
++ movl PTR(PA_CONTROL_PAGE)(%ebp), %edx
++ orl $PAGE_ATTR, %edx
++ movl %edx, (%eax)
++
++ /* identity map the control page at its physical address */
++
++ movl PTR(VA_PGD)(%ebp), %edi
++ movl PTR(PA_CONTROL_PAGE)(%ebp), %eax
++ andl $0xc0000000, %eax
++ shrl $27, %eax
++ addl %edi, %eax
++
++ movl PTR(PA_PMD_1)(%ebp), %edx
++ orl $PAE_PGD_ATTR, %edx
++ movl %edx, (%eax)
++
++ movl PTR(VA_PMD_1)(%ebp), %edi
++ movl PTR(PA_CONTROL_PAGE)(%ebp), %eax
++ andl $0x3fe00000, %eax
++ shrl $18, %eax
++ addl %edi, %eax
++
++ movl PTR(PA_PTE_1)(%ebp), %edx
++ orl $PAGE_ATTR, %edx
++ movl %edx, (%eax)
++
++ movl PTR(VA_PTE_1)(%ebp), %edi
++ movl PTR(PA_CONTROL_PAGE)(%ebp), %eax
++ andl $0x001ff000, %eax
++ shrl $9, %eax
++ addl %edi, %eax
++
++ movl PTR(PA_CONTROL_PAGE)(%ebp), %edx
++ orl $PAGE_ATTR, %edx
++ movl %edx, (%eax)
++#else
++ /* map the control page at its virtual address */
++
++ movl PTR(VA_PGD)(%ebp), %edi
++ movl PTR(VA_CONTROL_PAGE)(%ebp), %eax
++ andl $0xffc00000, %eax
++ shrl $20, %eax
++ addl %edi, %eax
++
++ movl PTR(PA_PTE_0)(%ebp), %edx
++ orl $PAGE_ATTR, %edx
++ movl %edx, (%eax)
++
++ movl PTR(VA_PTE_0)(%ebp), %edi
++ movl PTR(VA_CONTROL_PAGE)(%ebp), %eax
++ andl $0x003ff000, %eax
++ shrl $10, %eax
++ addl %edi, %eax
++
++ movl PTR(PA_CONTROL_PAGE)(%ebp), %edx
++ orl $PAGE_ATTR, %edx
++ movl %edx, (%eax)
++
++ /* identity map the control page at its physical address */
++
++ movl PTR(VA_PGD)(%ebp), %edi
++ movl PTR(PA_CONTROL_PAGE)(%ebp), %eax
++ andl $0xffc00000, %eax
++ shrl $20, %eax
++ addl %edi, %eax
++
++ movl PTR(PA_PTE_1)(%ebp), %edx
++ orl $PAGE_ATTR, %edx
++ movl %edx, (%eax)
++
++ movl PTR(VA_PTE_1)(%ebp), %edi
++ movl PTR(PA_CONTROL_PAGE)(%ebp), %eax
++ andl $0x003ff000, %eax
++ shrl $10, %eax
++ addl %edi, %eax
++
++ movl PTR(PA_CONTROL_PAGE)(%ebp), %edx
++ orl $PAGE_ATTR, %edx
++ movl %edx, (%eax)
++#endif
+
+- /*
+- * Must be relocatable PIC code callable as a C function, that once
+- * it starts can not use the previous processes stack.
+- */
+- .globl relocate_new_kernel
+ relocate_new_kernel:
+ /* read the arguments and say goodbye to the stack */
+ movl 4(%esp), %ebx /* page_list */
+- movl 8(%esp), %ebp /* reboot_code_buffer */
++ movl 8(%esp), %ebp /* list of pages */
+ movl 12(%esp), %edx /* start address */
+ movl 16(%esp), %ecx /* cpu_has_pae */
+
+@@ -24,11 +146,26 @@ relocate_new_kernel:
+ pushl $0
+ popfl
+
+- /* set a new stack at the bottom of our page... */
+- lea 4096(%ebp), %esp
++ /* get physical address of control page now */
++ /* this is impossible after page table switch */
++ movl PTR(PA_CONTROL_PAGE)(%ebp), %edi
++
++ /* switch to new set of page tables */
++ movl PTR(PA_PGD)(%ebp), %eax
++ movl %eax, %cr3
++
++ /* setup a new stack at the end of the physical control page */
++ lea 4096(%edi), %esp
+
+- /* store the parameters back on the stack */
+- pushl %edx /* store the start address */
++ /* jump to identity mapped page */
++ movl %edi, %eax
++ addl $(identity_mapped - relocate_kernel), %eax
++ pushl %eax
++ ret
++
++identity_mapped:
++ /* store the start address on the stack */
++ pushl %edx
+
+ /* Set cr0 to a known state:
+ * 31 0 == Paging disabled
+@@ -113,8 +250,3 @@ relocate_new_kernel:
+ xorl %edi, %edi
+ xorl %ebp, %ebp
+ ret
+-relocate_new_kernel_end:
+-
+- .globl relocate_new_kernel_size
+-relocate_new_kernel_size:
+- .long relocate_new_kernel_end - relocate_new_kernel
+--- a/include/asm-i386/kexec.h
++++ b/include/asm-i386/kexec.h
+@@ -1,6 +1,26 @@
+ #ifndef _I386_KEXEC_H
+ #define _I386_KEXEC_H
+
++#define PA_CONTROL_PAGE 0
++#define VA_CONTROL_PAGE 1
++#define PA_PGD 2
++#define VA_PGD 3
++#define PA_PTE_0 4
++#define VA_PTE_0 5
++#define PA_PTE_1 6
++#define VA_PTE_1 7
++#ifdef CONFIG_X86_PAE
++#define PA_PMD_0 8
++#define VA_PMD_0 9
++#define PA_PMD_1 10
++#define VA_PMD_1 11
++#define PAGES_NR 12
++#else
++#define PAGES_NR 8
++#endif
++
++#ifndef __ASSEMBLY__
++
+ #include <asm/fixmap.h>
+ #include <asm/ptrace.h>
+ #include <asm/string.h>
+@@ -72,5 +92,12 @@ static inline void crash_setup_regs(stru
+ newregs->eip = (unsigned long)current_text_addr();
+ }
+ }
++asmlinkage NORET_TYPE void
++relocate_kernel(unsigned long indirection_page,
++ unsigned long control_page,
++ unsigned long start_address,
++ unsigned int has_pae) ATTRIB_NORET;
++
++#endif /* __ASSEMBLY__ */
+
+ #endif /* _I386_KEXEC_H */
--- /dev/null
+++ work/patches/linux-2.6.16.29/linux-2.6.16.29-crash-xen-i386.patch
@@ -0,0 +1,31 @@
+--- 0001/arch/i386/kernel/crash.c
++++ work/arch/i386/kernel/crash.c 2006-10-26 15:45:35.000000000 +0900
+@@ -90,6 +90,7 @@ static void crash_save_self(struct pt_re
+ crash_save_this_cpu(regs, cpu);
+ }
+
++#ifndef CONFIG_XEN
+ #ifdef CONFIG_SMP
+ static atomic_t waiting_for_crash_ipi;
+
+@@ -158,6 +159,7 @@ static void nmi_shootdown_cpus(void)
+ /* There are no cpus to shootdown */
+ }
+ #endif
++#endif /* CONFIG_XEN */
+
+ void machine_crash_shutdown(struct pt_regs *regs)
+ {
+@@ -174,10 +176,12 @@ void machine_crash_shutdown(struct pt_re
+
+ /* Make a note of crashing cpu. Will be used in NMI callback.*/
+ crashing_cpu = smp_processor_id();
++#ifndef CONFIG_XEN
+ nmi_shootdown_cpus();
+ lapic_shutdown();
+ #if defined(CONFIG_X86_IO_APIC)
+ disable_IO_APIC();
+ #endif
++#endif /* CONFIG_XEN */
+ crash_save_self(regs);
+ }
--- /dev/null
+++ work/patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-move_segment_code-i386.patch
@@ -0,0 +1,169 @@
+kexec: Move asm segment handling code to the assembly file (i386)
+
+This patch moves the idt, gdt, and segment handling code from machine_kexec.c
+to relocate_kernel.S. The main reason behind this move is to avoid code
+duplication in the Xen hypervisor. With this patch all code required to kexec
+is put on the control page.
+
+On top of that this patch also counts as a cleanup - I think it is much
+nicer to write assembly directly in assembly files than wrap inline assembly
+in C functions for no apparent reason.
+
+Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
+---
+
+ Applies to 2.6.19-rc1.
+
+ machine_kexec.c | 59 -----------------------------------------------------
+ relocate_kernel.S | 58 +++++++++++++++++++++++++++++++++++++++++++++++-----
+ 2 files changed, 53 insertions(+), 64 deletions(-)
+
+--- 0002/arch/i386/kernel/machine_kexec.c
++++ work/arch/i386/kernel/machine_kexec.c 2006-10-05 15:49:08.000000000 +0900
+@@ -29,48 +29,6 @@ static u32 kexec_pmd1[1024] PAGE_ALIGNED
+ static u32 kexec_pte0[1024] PAGE_ALIGNED;
+ static u32 kexec_pte1[1024] PAGE_ALIGNED;
+
+-static void set_idt(void *newidt, __u16 limit)
+-{
+- struct Xgt_desc_struct curidt;
+-
+- /* ia32 supports unaliged loads & stores */
+- curidt.size = limit;
+- curidt.address = (unsigned long)newidt;
+-
+- load_idt(&curidt);
+-};
+-
+-
+-static void set_gdt(void *newgdt, __u16 limit)
+-{
+- struct Xgt_desc_struct curgdt;
+-
+- /* ia32 supports unaligned loads & stores */
+- curgdt.size = limit;
+- curgdt.address = (unsigned long)newgdt;
+-
+- load_gdt(&curgdt);
+-};
+-
+-static void load_segments(void)
+-{
+-#define __STR(X) #X
+-#define STR(X) __STR(X)
+-
+- __asm__ __volatile__ (
+- "\tljmp $"STR(__KERNEL_CS)",$1f\n"
+- "\t1:\n"
+- "\tmovl $"STR(__KERNEL_DS)",%%eax\n"
+- "\tmovl %%eax,%%ds\n"
+- "\tmovl %%eax,%%es\n"
+- "\tmovl %%eax,%%fs\n"
+- "\tmovl %%eax,%%gs\n"
+- "\tmovl %%eax,%%ss\n"
+- ::: "eax", "memory");
+-#undef STR
+-#undef __STR
+-}
+-
+ /*
+ * A architecture hook called to validate the
+ * proposed image and prepare the control pages
+@@ -127,23 +85,6 @@ NORET_TYPE void machine_kexec(struct kim
+ page_list[PA_PTE_1] = __pa(kexec_pte1);
+ page_list[VA_PTE_1] = (unsigned long)kexec_pte1;
+
+- /* The segment registers are funny things, they have both a
+- * visible and an invisible part. Whenever the visible part is
+- * set to a specific selector, the invisible part is loaded
+- * with from a table in memory. At no other time is the
+- * descriptor table in memory accessed.
+- *
+- * I take advantage of this here by force loading the
+- * segments, before I zap the gdt with an invalid value.
+- */
+- load_segments();
+- /* The gdt & idt are now invalid.
+- * If you want to load them you must set up your own idt & gdt.
+- */
+- set_gdt(phys_to_virt(0),0);
+- set_idt(phys_to_virt(0),0);
+-
+- /* now call it */
+ relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
+ image->start, cpu_has_pae);
+ }
+--- 0002/arch/i386/kernel/relocate_kernel.S
++++ work/arch/i386/kernel/relocate_kernel.S 2006-10-05 16:03:21.000000000 +0900
+@@ -154,14 +154,45 @@ relocate_new_kernel:
+ movl PTR(PA_PGD)(%ebp), %eax
+ movl %eax, %cr3
+
++ /* setup idt */
++ movl %edi, %eax
++ addl $(idt_48 - relocate_kernel), %eax
++ lidtl (%eax)
++
++ /* setup gdt */
++ movl %edi, %eax
++ addl $(gdt - relocate_kernel), %eax
++ movl %edi, %esi
++ addl $((gdt_48 - relocate_kernel) + 2), %esi
++ movl %eax, (%esi)
++
++ movl %edi, %eax
++ addl $(gdt_48 - relocate_kernel), %eax
++ lgdtl (%eax)
++
++ /* setup data segment registers */
++ mov $(gdt_ds - gdt), %eax
++ mov %eax, %ds
++ mov %eax, %es
++ mov %eax, %fs
++ mov %eax, %gs
++ mov %eax, %ss
++
+ /* setup a new stack at the end of the physical control page */
+ lea 4096(%edi), %esp
+
+- /* jump to identity mapped page */
+- movl %edi, %eax
+- addl $(identity_mapped - relocate_kernel), %eax
+- pushl %eax
+- ret
++ /* load new code segment and jump to identity mapped page */
++ movl %edi, %esi
++ xorl %eax, %eax
++ pushl %eax
++ pushl %esi
++ pushl %eax
++ movl $(gdt_cs - gdt), %eax
++ pushl %eax
++ movl %edi, %eax
++ addl $(identity_mapped - relocate_kernel),%eax
++ pushl %eax
++ iretl
+
+ identity_mapped:
+ /* store the start address on the stack */
+@@ -250,3 +281,20 @@ identity_mapped:
+ xorl %edi, %edi
+ xorl %ebp, %ebp
+ ret
++
++ .align 16
++gdt:
++ .quad 0x0000000000000000 /* NULL descriptor */
++gdt_cs:
++ .quad 0x00cf9a000000ffff /* kernel 4GB code at 0x00000000 */
++gdt_ds:
++ .quad 0x00cf92000000ffff /* kernel 4GB data at 0x00000000 */
++gdt_end:
++
++gdt_48:
++ .word gdt_end - gdt - 1 /* limit */
++ .long 0 /* base - filled in by code above */
++
++idt_48:
++ .word 0 /* limit */
++ .long 0 /* base */
--- /dev/null
+++ work/patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-xen-i386.patch
@@ -0,0 +1,54 @@
+--- 0004/arch/i386/kernel/machine_kexec.c
++++ work/arch/i386/kernel/machine_kexec.c 2006-10-11 18:34:06.000000000 +0900
+@@ -20,6 +20,10 @@
+ #include <asm/desc.h>
+ #include <asm/system.h>
+
++#ifdef CONFIG_XEN
++#include <xen/interface/kexec.h>
++#endif
++
+ #define PAGE_ALIGNED __attribute__ ((__aligned__(PAGE_SIZE)))
+ static u32 kexec_pgd[1024] PAGE_ALIGNED;
+ #ifdef CONFIG_X86_PAE
+@@ -29,6 +33,40 @@ static u32 kexec_pmd1[1024] PAGE_ALIGNED
+ static u32 kexec_pte0[1024] PAGE_ALIGNED;
+ static u32 kexec_pte1[1024] PAGE_ALIGNED;
+
++#ifdef CONFIG_XEN
++
++#define __ma(x) (pfn_to_mfn(__pa((x)) >> PAGE_SHIFT) << PAGE_SHIFT)
++
++#if PAGES_NR > KEXEC_XEN_NO_PAGES
++#error PAGES_NR is greater than KEXEC_XEN_NO_PAGES - Xen support will break
++#endif
++
++#if PA_CONTROL_PAGE != 0
++#error PA_CONTROL_PAGE is non zero - Xen support will break
++#endif
++
++void machine_kexec_setup_load_arg(xen_kexec_image_t *xki, struct kimage *image)
++{
++ void *control_page;
++
++ memset(xki->page_list, 0, sizeof(xki->page_list));
++
++ control_page = page_address(image->control_code_page);
++ memcpy(control_page, relocate_kernel, PAGE_SIZE);
++
++ xki->page_list[PA_CONTROL_PAGE] = __ma(control_page);
++ xki->page_list[PA_PGD] = __ma(kexec_pgd);
++#ifdef CONFIG_X86_PAE
++ xki->page_list[PA_PMD_0] = __ma(kexec_pmd0);
++ xki->page_list[PA_PMD_1] = __ma(kexec_pmd1);
++#endif
++ xki->page_list[PA_PTE_0] = __ma(kexec_pte0);
++ xki->page_list[PA_PTE_1] = __ma(kexec_pte1);
++
++}
++
++#endif /* CONFIG_XEN */
++
+ /*
+ * A architecture hook called to validate the
+ * proposed image and prepare the control pages
--- 0004/patches/linux-2.6.16.29/series
+++ work/patches/linux-2.6.16.29/series
@@ -1,6 +1,10 @@
kexec-generic.patch
git-2efe55a9cec8418f0e0cde3dc3787a42fddc4411.patch
git-2a8a3d5b65e86ec1dfef7d268c64a909eab94af7.patch
+git-3566561bfadffcb5dbc85d576be80c0dbf2cccc9.patch
+linux-2.6.19-rc1-kexec-move_segment_code-i386.patch
+linux-2.6.19-rc1-kexec-xen-i386.patch
+linux-2.6.16.29-crash-xen-i386.patch
blktap-aio-16_03_06.patch
device_bind.patch
fix-hz-suspend.patch
--- 0001/xen/arch/x86/x86_32/entry.S
+++ work/xen/arch/x86/x86_32/entry.S
@@ -659,6 +659,7 @@ ENTRY(hypercall_table)
.long do_hvm_op
.long do_sysctl /* 35 */
.long do_domctl
+ .long do_kexec_op
.rept NR_hypercalls-((.-hypercall_table)/4)
.long do_ni_hypercall
.endr
@@ -701,6 +702,7 @@ ENTRY(hypercall_args_table)
.byte 2 /* do_hvm_op */
.byte 1 /* do_sysctl */ /* 35 */
.byte 1 /* do_domctl */
+ .byte 2 /* do_kexec_op */
.rept NR_hypercalls-(.-hypercall_args_table)
.byte 0 /* do_ni_hypercall */
.endr
--- 0004/xen/include/asm-x86/x86_32/elf.h
+++ work/xen/include/asm-x86/x86_32/elf.h
@@ -1,15 +1,47 @@
#ifndef __X86_32_ELF_H__
#define __X86_32_ELF_H__
-#include <xen/lib.h> /* for printk() used in stub */
+#include <asm/processor.h>
typedef struct {
- unsigned long dummy;
+ unsigned long ebx;
+ unsigned long ecx;
+ unsigned long edx;
+ unsigned long esi;
+ unsigned long edi;
+ unsigned long ebp;
+ unsigned long eax;
+ unsigned long ds;
+ unsigned long es;
+ unsigned long fs;
+ unsigned long gs;
+ unsigned long orig_eax;
+ unsigned long eip;
+ unsigned long cs;
+ unsigned long eflags;
+ unsigned long esp;
+ unsigned long ss;
} ELF_Gregset;
extern inline void elf_core_save_regs(ELF_Gregset *dst)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ asm volatile("movl %%ebx,%0" : "=m"(dst->ebx));
+ asm volatile("movl %%ecx,%0" : "=m"(dst->ecx));
+ asm volatile("movl %%edx,%0" : "=m"(dst->edx));
+ asm volatile("movl %%esi,%0" : "=m"(dst->esi));
+ asm volatile("movl %%edi,%0" : "=m"(dst->edi));
+ asm volatile("movl %%ebp,%0" : "=m"(dst->ebp));
+ asm volatile("movl %%eax,%0" : "=m"(dst->eax));
+ asm volatile("movw %%ds, %%ax;" :"=a"(dst->ds));
+ asm volatile("movw %%es, %%ax;" :"=a"(dst->es));
+ asm volatile("movw %%fs, %%ax;" :"=a"(dst->fs));
+ asm volatile("movw %%gs, %%ax;" :"=a"(dst->gs));
+ /* orig_eax not filled in for now */
+ dst->eip = (unsigned long)current_text_addr();
+ asm volatile("movw %%cs, %%ax;" :"=a"(dst->cs));
+ asm volatile("pushfl; popl %0" :"=m"(dst->eflags));
+ asm volatile("movl %%esp,%0" : "=m"(dst->esp));
+ asm volatile("movw %%ss, %%ax;" :"=a"(dst->ss));
}
#endif /* __X86_32_ELF_H__ */
--- 0004/xen/include/asm-x86/x86_32/kexec.h
+++ work/xen/include/asm-x86/x86_32/kexec.h
@@ -1,17 +1,33 @@
-#ifndef __X86_32_KEXEC_H__
-#define __X86_32_KEXEC_H__
+/******************************************************************************
+ * kexec.h
+ *
+ * Based heavily on machine_kexec.c and kexec.h from Linux 2.6.19-rc1
+ *
+ */
+
+#ifndef __X86_KEXEC_X86_32_H__
+#define __X86_KEXEC_X86_32_H__
-#include <xen/lib.h> /* for printk() used in stub */
#include <xen/types.h>
-#include <public/xen.h>
#include <xen/kexec.h>
+#include <asm/fixmap.h>
+
+typedef asmlinkage void (*relocate_new_kernel_t)(
+ unsigned long indirection_page,
+ unsigned long page_list,
+ unsigned long start_address,
+ unsigned int has_pae);
static inline void machine_kexec(xen_kexec_image_t *image)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ relocate_new_kernel_t rnk;
+
+ rnk = (relocate_new_kernel_t) image->page_list[1];
+ (*rnk)(image->indirection_page, (unsigned long)image->page_list,
+ image->start_address, (unsigned long)cpu_has_pae);
}
-#endif /* __X86_32_KEXEC_H__ */
+#endif /* __X86_KEXEC_X86_32_H__ */
/*
* Local variables:
^ permalink raw reply [flat|nested] 30+ messages in thread* [PATCH 04/04] Kexec / Kdump: x86_64 specific code
2006-11-22 7:10 [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502) Magnus Damm
` (2 preceding siblings ...)
2006-11-22 7:11 ` [PATCH 03/04] Kexec / Kdump: x86_32 specific code Magnus Damm
@ 2006-11-22 7:11 ` Magnus Damm
2006-11-22 18:24 ` [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502) Ian Campbell
2006-11-28 18:24 ` Ian Campbell
5 siblings, 0 replies; 30+ messages in thread
From: Magnus Damm @ 2006-11-22 7:11 UTC (permalink / raw)
To: Keir Fraser
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, magnus.damm,
Isaku Yamahata, Magnus Damm, Horms
[PATCH 04/04] Kexec / Kdump: x86_64 specific code
This patch contains the x86_64 implementation of Kexec / Kdump for Xen.
Signed-Off-By: Magnus Damm <magnus@valinux.co.jp>
---
Applies on top of xen-unstable-12502.
buildconfigs/linux-defconfig_xen_x86_64 | 1
linux-2.6-xen-sparse/arch/x86_64/Kconfig | 2
linux-2.6-xen-sparse/arch/x86_64/kernel/e820-xen.c | 6
linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c | 13
linux-2.6-xen-sparse/include/asm-x86_64/kexec-xen.h | 30
linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h | 7
linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h | 2
patches/linux-2.6.16.29/git-4b...1f.patch | 375 +++
patches/linux-2.6.16.29/linux-2.6.16.29-crash-xen-x86_64.patch | 35
patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec...code-x86_64.patch | 161 +
patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-xen-x86_64.patch | 162 +
patches/linux-2.6.16.29/series | 4
xen/arch/x86/x86_64/entry.S | 2
xen/include/asm-x86/x86_64/elf.h | 58 +
xen/include/asm-x86/x86_64/kexec.h | 23
15 files changed, 872 insertions(+), 9 deletions(-)
--- 0002/buildconfigs/linux-defconfig_xen_x86_64
+++ work/buildconfigs/linux-defconfig_xen_x86_64
@@ -138,6 +138,7 @@ CONFIG_SWIOTLB=y
CONFIG_PHYSICAL_START=0x100000
CONFIG_SECCOMP=y
CONFIG_HZ_100=y
+CONFIG_KEXEC=y
# CONFIG_HZ_250 is not set
# CONFIG_HZ_1000 is not set
CONFIG_HZ=100
--- 0001/linux-2.6-xen-sparse/arch/x86_64/Kconfig
+++ work/linux-2.6-xen-sparse/arch/x86_64/Kconfig
@@ -435,7 +435,7 @@ config X86_MCE_AMD
config KEXEC
bool "kexec system call (EXPERIMENTAL)"
- depends on EXPERIMENTAL && !X86_64_XEN
+ depends on EXPERIMENTAL && !XEN_UNPRIVILEGED_GUEST
help
kexec is a system call that implements the ability to shutdown your
current kernel, and to start another kernel. It is like a reboot
--- 0001/linux-2.6-xen-sparse/arch/x86_64/kernel/e820-xen.c
+++ work/linux-2.6-xen-sparse/arch/x86_64/kernel/e820-xen.c
@@ -260,7 +260,11 @@ void __init e820_reserve_resources(struc
request_resource(res, &data_resource);
#endif
#ifdef CONFIG_KEXEC
- request_resource(res, &crashk_res);
+ if (crashk_res.start != crashk_res.end)
+ request_resource(res, &crashk_res);
+#ifdef CONFIG_XEN
+ xen_machine_kexec_register_resources(res);
+#endif
#endif
}
}
--- 0001/linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c
+++ work/linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c
@@ -80,6 +80,10 @@
#include <asm/mach-xen/setup_arch_post.h>
#include <xen/interface/memory.h>
+#ifdef CONFIG_XEN
+#include <xen/interface/kexec.h>
+#endif
+
extern unsigned long start_pfn;
extern struct edid_info edid_info;
@@ -450,6 +454,7 @@ static __init void parse_cmdline_early (
* after a kernel panic.
*/
else if (!memcmp(from, "crashkernel=", 12)) {
+#ifndef CONFIG_XEN
unsigned long size, base;
size = memparse(from+12, &from);
if (*from == '@') {
@@ -460,6 +465,10 @@ static __init void parse_cmdline_early (
crashk_res.start = base;
crashk_res.end = base + size - 1;
}
+#else
+ printk("Ignoring crashkernel command line, "
+ "parameter will be supplied by xen\n");
+#endif
}
#endif
@@ -812,11 +821,15 @@ void __init setup_arch(char **cmdline_p)
#endif
#endif /* !CONFIG_XEN */
#ifdef CONFIG_KEXEC
+#ifdef CONFIG_XEN
+ xen_machine_kexec_setup_resources();
+#else
if (crashk_res.start != crashk_res.end) {
reserve_bootmem(crashk_res.start,
crashk_res.end - crashk_res.start + 1);
}
#endif
+#endif
paging_init();
#ifdef CONFIG_X86_LOCAL_APIC
--- /dev/null
+++ work/linux-2.6-xen-sparse/include/asm-x86_64/kexec-xen.h
@@ -0,0 +1,30 @@
+#ifndef _X86_64_KEXEC_XEN_H
+#define _X86_64_KEXEC_XEN_H
+
+/* Kexec needs to know about the actual physical addresss.
+ * But in xen, on some architectures, a physical address is a
+ * pseudo-physical addresss. */
+
+#ifdef CONFIG_XEN
+#define kexec_page_to_pfn(page) pfn_to_mfn(page_to_pfn(page))
+#define kexec_pfn_to_page(pfn) pfn_to_page(mfn_to_pfn(pfn))
+#define kexec_virt_to_phys(addr) virt_to_machine(addr)
+#define kexec_phys_to_virt(addr) phys_to_virt(machine_to_phys(addr))
+#else
+#define kexec_page_to_pfn(page) page_to_pfn(page)
+#define kexec_pfn_to_page(pfn) pfn_to_page(pfn)
+#define kexec_virt_to_phys(addr) virt_to_phys(addr)
+#define kexec_phys_to_virt(addr) phys_to_virt(addr)
+#endif
+
+#endif /* _X86_64_KEXEC_XEN_H */
+
+/*
+ * Local variables:
+ * c-file-style: "linux"
+ * indent-tabs-mode: t
+ * c-indent-level: 8
+ * c-basic-offset: 8
+ * tab-width: 8
+ * End:
+ */
--- 0001/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h
+++ work/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h
@@ -396,4 +396,11 @@ HYPERVISOR_xenoprof_op(
return _hypercall2(int, xenoprof_op, op, arg);
}
+static inline int
+HYPERVISOR_kexec_op(
+ unsigned long op, void *args)
+{
+ return _hypercall2(int, kexec_op, op, args);
+}
+
#endif /* __HYPERCALL_H__ */
--- 0001/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h
+++ work/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h
@@ -90,6 +90,8 @@ extern unsigned long profile_pc(struct p
#define profile_pc(regs) instruction_pointer(regs)
#endif
+#include <linux/compiler.h>
+
void signal_fault(struct pt_regs *regs, void __user *frame, char *where);
struct task_struct;
--- /dev/null
+++ work/patches/linux-2.6.16.29/git-4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f.patch
@@ -0,0 +1,375 @@
+From: Magnus Damm <magnus@valinux.co.jp>
+Date: Tue, 26 Sep 2006 08:52:38 +0000 (+0200)
+Subject: [PATCH] Avoid overwriting the current pgd (V4, x86_64)
+X-Git-Tag: v2.6.19-rc1
+X-Git-Url: http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f
+
+[PATCH] Avoid overwriting the current pgd (V4, x86_64)
+
+kexec: Avoid overwriting the current pgd (V4, x86_64)
+
+This patch upgrades the x86_64-specific kexec code to avoid overwriting the
+current pgd. Overwriting the current pgd is bad when CONFIG_CRASH_DUMP is used
+to start a secondary kernel that dumps the memory of the previous kernel.
+
+The code introduces a new set of page tables. These tables are used to provide
+an executable identity mapping without overwriting the current pgd.
+
+Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
+Signed-off-by: Andi Kleen <ak@suse.de>
+---
+
+--- a/arch/x86_64/kernel/machine_kexec.c
++++ b/arch/x86_64/kernel/machine_kexec.c
+@@ -15,6 +15,15 @@
+ #include <asm/mmu_context.h>
+ #include <asm/io.h>
+
++#define PAGE_ALIGNED __attribute__ ((__aligned__(PAGE_SIZE)))
++static u64 kexec_pgd[512] PAGE_ALIGNED;
++static u64 kexec_pud0[512] PAGE_ALIGNED;
++static u64 kexec_pmd0[512] PAGE_ALIGNED;
++static u64 kexec_pte0[512] PAGE_ALIGNED;
++static u64 kexec_pud1[512] PAGE_ALIGNED;
++static u64 kexec_pmd1[512] PAGE_ALIGNED;
++static u64 kexec_pte1[512] PAGE_ALIGNED;
++
+ static void init_level2_page(pmd_t *level2p, unsigned long addr)
+ {
+ unsigned long end_addr;
+@@ -144,32 +153,19 @@ static void load_segments(void)
+ );
+ }
+
+-typedef NORET_TYPE void (*relocate_new_kernel_t)(unsigned long indirection_page,
+- unsigned long control_code_buffer,
+- unsigned long start_address,
+- unsigned long pgtable) ATTRIB_NORET;
+-
+-extern const unsigned char relocate_new_kernel[];
+-extern const unsigned long relocate_new_kernel_size;
+-
+ int machine_kexec_prepare(struct kimage *image)
+ {
+- unsigned long start_pgtable, control_code_buffer;
++ unsigned long start_pgtable;
+ int result;
+
+ /* Calculate the offsets */
+ start_pgtable = page_to_pfn(image->control_code_page) << PAGE_SHIFT;
+- control_code_buffer = start_pgtable + PAGE_SIZE;
+
+ /* Setup the identity mapped 64bit page table */
+ result = init_pgtable(image, start_pgtable);
+ if (result)
+ return result;
+
+- /* Place the code in the reboot code buffer */
+- memcpy(__va(control_code_buffer), relocate_new_kernel,
+- relocate_new_kernel_size);
+-
+ return 0;
+ }
+
+@@ -184,28 +180,34 @@ void machine_kexec_cleanup(struct kimage
+ */
+ NORET_TYPE void machine_kexec(struct kimage *image)
+ {
+- unsigned long page_list;
+- unsigned long control_code_buffer;
+- unsigned long start_pgtable;
+- relocate_new_kernel_t rnk;
++ unsigned long page_list[PAGES_NR];
++ void *control_page;
+
+ /* Interrupts aren't acceptable while we reboot */
+ local_irq_disable();
+
+- /* Calculate the offsets */
+- page_list = image->head;
+- start_pgtable = page_to_pfn(image->control_code_page) << PAGE_SHIFT;
+- control_code_buffer = start_pgtable + PAGE_SIZE;
++ control_page = page_address(image->control_code_page) + PAGE_SIZE;
++ memcpy(control_page, relocate_kernel, PAGE_SIZE);
+
+- /* Set the low half of the page table to my identity mapped
+- * page table for kexec. Leave the high half pointing at the
+- * kernel pages. Don't bother to flush the global pages
+- * as that will happen when I fully switch to my identity mapped
+- * page table anyway.
+- */
+- memcpy(__va(read_cr3()), __va(start_pgtable), PAGE_SIZE/2);
+- __flush_tlb();
++ page_list[PA_CONTROL_PAGE] = __pa(control_page);
++ page_list[VA_CONTROL_PAGE] = (unsigned long)relocate_kernel;
++ page_list[PA_PGD] = __pa(kexec_pgd);
++ page_list[VA_PGD] = (unsigned long)kexec_pgd;
++ page_list[PA_PUD_0] = __pa(kexec_pud0);
++ page_list[VA_PUD_0] = (unsigned long)kexec_pud0;
++ page_list[PA_PMD_0] = __pa(kexec_pmd0);
++ page_list[VA_PMD_0] = (unsigned long)kexec_pmd0;
++ page_list[PA_PTE_0] = __pa(kexec_pte0);
++ page_list[VA_PTE_0] = (unsigned long)kexec_pte0;
++ page_list[PA_PUD_1] = __pa(kexec_pud1);
++ page_list[VA_PUD_1] = (unsigned long)kexec_pud1;
++ page_list[PA_PMD_1] = __pa(kexec_pmd1);
++ page_list[VA_PMD_1] = (unsigned long)kexec_pmd1;
++ page_list[PA_PTE_1] = __pa(kexec_pte1);
++ page_list[VA_PTE_1] = (unsigned long)kexec_pte1;
+
++ page_list[PA_TABLE_PAGE] =
++ (unsigned long)__pa(page_address(image->control_code_page));
+
+ /* The segment registers are funny things, they have both a
+ * visible and an invisible part. Whenever the visible part is
+@@ -222,9 +224,10 @@ NORET_TYPE void machine_kexec(struct kim
+ */
+ set_gdt(phys_to_virt(0),0);
+ set_idt(phys_to_virt(0),0);
++
+ /* now call it */
+- rnk = (relocate_new_kernel_t) control_code_buffer;
+- (*rnk)(page_list, control_code_buffer, image->start, start_pgtable);
++ relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
++ image->start);
+ }
+
+ /* crashkernel=size@addr specifies the location to reserve for
+--- a/arch/x86_64/kernel/relocate_kernel.S
++++ b/arch/x86_64/kernel/relocate_kernel.S
+@@ -7,31 +7,169 @@
+ */
+
+ #include <linux/linkage.h>
++#include <asm/page.h>
++#include <asm/kexec.h>
+
+- /*
+- * Must be relocatable PIC code callable as a C function, that once
+- * it starts can not use the previous processes stack.
+- */
+- .globl relocate_new_kernel
++/*
++ * Must be relocatable PIC code callable as a C function
++ */
++
++#define PTR(x) (x << 3)
++#define PAGE_ALIGNED (1 << PAGE_SHIFT)
++#define PAGE_ATTR 0x63 /* _PAGE_PRESENT|_PAGE_RW|_PAGE_ACCESSED|_PAGE_DIRTY */
++
++ .text
++ .align PAGE_ALIGNED
+ .code64
++ .globl relocate_kernel
++relocate_kernel:
++ /* %rdi indirection_page
++ * %rsi page_list
++ * %rdx start address
++ */
++
++ /* map the control page at its virtual address */
++
++ movq $0x0000ff8000000000, %r10 /* mask */
++ mov $(39 - 3), %cl /* bits to shift */
++ movq PTR(VA_CONTROL_PAGE)(%rsi), %r11 /* address to map */
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PGD)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PUD_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PUD_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PMD_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PMD_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PTE_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PTE_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ /* identity map the control page at its physical address */
++
++ movq $0x0000ff8000000000, %r10 /* mask */
++ mov $(39 - 3), %cl /* bits to shift */
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r11 /* address to map */
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PGD)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PUD_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PUD_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PMD_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PMD_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PTE_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PTE_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
+ relocate_new_kernel:
+- /* %rdi page_list
+- * %rsi reboot_code_buffer
++ /* %rdi indirection_page
++ * %rsi page_list
+ * %rdx start address
+- * %rcx page_table
+- * %r8 arg5
+- * %r9 arg6
+ */
+
+ /* zero out flags, and disable interrupts */
+ pushq $0
+ popfq
+
+- /* set a new stack at the bottom of our page... */
+- lea 4096(%rsi), %rsp
++ /* get physical address of control page now */
++ /* this is impossible after page table switch */
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++
++ /* get physical address of page table now too */
++ movq PTR(PA_TABLE_PAGE)(%rsi), %rcx
++
++ /* switch to new set of page tables */
++ movq PTR(PA_PGD)(%rsi), %r9
++ movq %r9, %cr3
++
++ /* setup a new stack at the end of the physical control page */
++ lea 4096(%r8), %rsp
++
++ /* jump to identity mapped page */
++ addq $(identity_mapped - relocate_kernel), %r8
++ pushq %r8
++ ret
+
+- /* store the parameters back on the stack */
+- pushq %rdx /* store the start address */
++identity_mapped:
++ /* store the start address on the stack */
++ pushq %rdx
+
+ /* Set cr0 to a known state:
+ * 31 1 == Paging enabled
+@@ -136,8 +274,3 @@ relocate_new_kernel:
+ xorq %r15, %r15
+
+ ret
+-relocate_new_kernel_end:
+-
+- .globl relocate_new_kernel_size
+-relocate_new_kernel_size:
+- .quad relocate_new_kernel_end - relocate_new_kernel
+--- a/include/asm-x86_64/kexec.h
++++ b/include/asm-x86_64/kexec.h
+@@ -1,6 +1,27 @@
+ #ifndef _X86_64_KEXEC_H
+ #define _X86_64_KEXEC_H
+
++#define PA_CONTROL_PAGE 0
++#define VA_CONTROL_PAGE 1
++#define PA_PGD 2
++#define VA_PGD 3
++#define PA_PUD_0 4
++#define VA_PUD_0 5
++#define PA_PMD_0 6
++#define VA_PMD_0 7
++#define PA_PTE_0 8
++#define VA_PTE_0 9
++#define PA_PUD_1 10
++#define VA_PUD_1 11
++#define PA_PMD_1 12
++#define VA_PMD_1 13
++#define PA_PTE_1 14
++#define VA_PTE_1 15
++#define PA_TABLE_PAGE 16
++#define PAGES_NR 17
++
++#ifndef __ASSEMBLY__
++
+ #include <linux/string.h>
+
+ #include <asm/page.h>
+@@ -64,4 +85,12 @@ static inline void crash_setup_regs(stru
+ newregs->rip = (unsigned long)current_text_addr();
+ }
+ }
++
++NORET_TYPE void
++relocate_kernel(unsigned long indirection_page,
++ unsigned long page_list,
++ unsigned long start_address) ATTRIB_NORET;
++
++#endif /* __ASSEMBLY__ */
++
+ #endif /* _X86_64_KEXEC_H */
--- /dev/null
+++ work/patches/linux-2.6.16.29/linux-2.6.16.29-crash-xen-x86_64.patch
@@ -0,0 +1,35 @@
+--- 0001/arch/x86_64/kernel/crash.c
++++ work/arch/x86_64/kernel/crash.c 2006-10-26 15:44:27.000000000 +0900
+@@ -92,6 +92,7 @@ static void crash_save_self(struct pt_re
+ crash_save_this_cpu(regs, cpu);
+ }
+
++#ifndef CONFIG_XEN
+ #ifdef CONFIG_SMP
+ static atomic_t waiting_for_crash_ipi;
+
+@@ -156,6 +157,7 @@ static void nmi_shootdown_cpus(void)
+ /* There are no cpus to shootdown */
+ }
+ #endif
++#endif /* CONFIG_XEN */
+
+ void machine_crash_shutdown(struct pt_regs *regs)
+ {
+@@ -173,6 +175,8 @@ void machine_crash_shutdown(struct pt_re
+
+ /* Make a note of crashing cpu. Will be used in NMI callback.*/
+ crashing_cpu = smp_processor_id();
++
++#ifndef CONFIG_XEN
+ nmi_shootdown_cpus();
+
+ if(cpu_has_apic)
+@@ -181,6 +185,6 @@ void machine_crash_shutdown(struct pt_re
+ #if defined(CONFIG_X86_IO_APIC)
+ disable_IO_APIC();
+ #endif
+-
++#endif /* CONFIG_XEN */
+ crash_save_self(regs);
+ }
--- /dev/null
+++ work/patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-move_segment_code-x86_64.patch
@@ -0,0 +1,161 @@
+kexec: Move asm segment handling code to the assembly file (x86_64)
+
+This patch moves the idt, gdt, and segment handling code from machine_kexec.c
+to relocate_kernel.S. The main reason behind this move is to avoid code
+duplication in the Xen hypervisor. With this patch all code required to kexec
+is put on the control page.
+
+On top of that this patch also counts as a cleanup - I think it is much
+nicer to write assembly directly in assembly files than wrap inline assembly
+in C functions for no apparent reason.
+
+Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
+---
+
+ Applies to 2.6.19-rc1.
+
+ machine_kexec.c | 58 -----------------------------------------------------
+ relocate_kernel.S | 50 +++++++++++++++++++++++++++++++++++++++++----
+ 2 files changed, 45 insertions(+), 63 deletions(-)
+
+--- 0002/arch/x86_64/kernel/machine_kexec.c
++++ work/arch/x86_64/kernel/machine_kexec.c 2006-10-05 16:15:49.000000000 +0900
+@@ -112,47 +112,6 @@ static int init_pgtable(struct kimage *i
+ return init_level4_page(image, level4p, 0, end_pfn << PAGE_SHIFT);
+ }
+
+-static void set_idt(void *newidt, u16 limit)
+-{
+- struct desc_ptr curidt;
+-
+- /* x86-64 supports unaliged loads & stores */
+- curidt.size = limit;
+- curidt.address = (unsigned long)newidt;
+-
+- __asm__ __volatile__ (
+- "lidtq %0\n"
+- : : "m" (curidt)
+- );
+-};
+-
+-
+-static void set_gdt(void *newgdt, u16 limit)
+-{
+- struct desc_ptr curgdt;
+-
+- /* x86-64 supports unaligned loads & stores */
+- curgdt.size = limit;
+- curgdt.address = (unsigned long)newgdt;
+-
+- __asm__ __volatile__ (
+- "lgdtq %0\n"
+- : : "m" (curgdt)
+- );
+-};
+-
+-static void load_segments(void)
+-{
+- __asm__ __volatile__ (
+- "\tmovl %0,%%ds\n"
+- "\tmovl %0,%%es\n"
+- "\tmovl %0,%%ss\n"
+- "\tmovl %0,%%fs\n"
+- "\tmovl %0,%%gs\n"
+- : : "a" (__KERNEL_DS) : "memory"
+- );
+-}
+-
+ int machine_kexec_prepare(struct kimage *image)
+ {
+ unsigned long start_pgtable;
+@@ -209,23 +168,6 @@ NORET_TYPE void machine_kexec(struct kim
+ page_list[PA_TABLE_PAGE] =
+ (unsigned long)__pa(page_address(image->control_code_page));
+
+- /* The segment registers are funny things, they have both a
+- * visible and an invisible part. Whenever the visible part is
+- * set to a specific selector, the invisible part is loaded
+- * with from a table in memory. At no other time is the
+- * descriptor table in memory accessed.
+- *
+- * I take advantage of this here by force loading the
+- * segments, before I zap the gdt with an invalid value.
+- */
+- load_segments();
+- /* The gdt & idt are now invalid.
+- * If you want to load them you must set up your own idt & gdt.
+- */
+- set_gdt(phys_to_virt(0),0);
+- set_idt(phys_to_virt(0),0);
+-
+- /* now call it */
+ relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
+ image->start);
+ }
+--- 0002/arch/x86_64/kernel/relocate_kernel.S
++++ work/arch/x86_64/kernel/relocate_kernel.S 2006-10-05 16:18:07.000000000 +0900
+@@ -159,13 +159,39 @@ relocate_new_kernel:
+ movq PTR(PA_PGD)(%rsi), %r9
+ movq %r9, %cr3
+
++ /* setup idt */
++ movq %r8, %rax
++ addq $(idt_80 - relocate_kernel), %rax
++ lidtq (%rax)
++
++ /* setup gdt */
++ movq %r8, %rax
++ addq $(gdt - relocate_kernel), %rax
++ movq %r8, %r9
++ addq $((gdt_80 - relocate_kernel) + 2), %r9
++ movq %rax, (%r9)
++
++ movq %r8, %rax
++ addq $(gdt_80 - relocate_kernel), %rax
++ lgdtq (%rax)
++
++ /* setup data segment registers */
++ xorl %eax, %eax
++ movl %eax, %ds
++ movl %eax, %es
++ movl %eax, %fs
++ movl %eax, %gs
++ movl %eax, %ss
++
+ /* setup a new stack at the end of the physical control page */
+ lea 4096(%r8), %rsp
+
+- /* jump to identity mapped page */
+- addq $(identity_mapped - relocate_kernel), %r8
+- pushq %r8
+- ret
++ /* load new code segment and jump to identity mapped page */
++ movq %r8, %rax
++ addq $(identity_mapped - relocate_kernel), %rax
++ pushq $(gdt_cs - gdt)
++ pushq %rax
++ lretq
+
+ identity_mapped:
+ /* store the start address on the stack */
+@@ -272,5 +298,19 @@ identity_mapped:
+ xorq %r13, %r13
+ xorq %r14, %r14
+ xorq %r15, %r15
+-
+ ret
++
++ .align 16
++gdt:
++ .quad 0x0000000000000000 /* NULL descriptor */
++gdt_cs:
++ .quad 0x00af9a000000ffff
++gdt_end:
++
++gdt_80:
++ .word gdt_end - gdt - 1 /* limit */
++ .quad 0 /* base - filled in by code above */
++
++idt_80:
++ .word 0 /* limit */
++ .quad 0 /* base */
--- /dev/null
+++ work/patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-xen-x86_64.patch
@@ -0,0 +1,162 @@
+--- 0006/arch/x86_64/kernel/machine_kexec.c
++++ work/arch/x86_64/kernel/machine_kexec.c 2006-10-06 15:36:16.000000000 +0900
+@@ -24,6 +24,104 @@ static u64 kexec_pud1[512] PAGE_ALIGNED;
+ static u64 kexec_pmd1[512] PAGE_ALIGNED;
+ static u64 kexec_pte1[512] PAGE_ALIGNED;
+
++#ifdef CONFIG_XEN
++
++/* In the case of Xen, override hypervisor functions to be able to create
++ * a regular identity mapping page table...
++ */
++
++#include <xen/interface/kexec.h>
++#include <xen/interface/memory.h>
++
++#define x__pmd(x) ((pmd_t) { (x) } )
++#define x__pud(x) ((pud_t) { (x) } )
++#define x__pgd(x) ((pgd_t) { (x) } )
++
++#define x_pmd_val(x) ((x).pmd)
++#define x_pud_val(x) ((x).pud)
++#define x_pgd_val(x) ((x).pgd)
++
++static inline void x_set_pmd(pmd_t *dst, pmd_t val)
++{
++ x_pmd_val(*dst) = x_pmd_val(val);
++}
++
++static inline void x_set_pud(pud_t *dst, pud_t val)
++{
++ x_pud_val(*dst) = phys_to_machine(x_pud_val(val));
++}
++
++static inline void x_pud_clear (pud_t *pud)
++{
++ x_pud_val(*pud) = 0;
++}
++
++static inline void x_set_pgd(pgd_t *dst, pgd_t val)
++{
++ x_pgd_val(*dst) = phys_to_machine(x_pgd_val(val));
++}
++
++static inline void x_pgd_clear (pgd_t * pgd)
++{
++ x_pgd_val(*pgd) = 0;
++}
++
++#define X__PAGE_KERNEL_LARGE_EXEC \
++ _PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_PSE
++#define X_KERNPG_TABLE _PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY
++
++#define __ma(x) (pfn_to_mfn(__pa((x)) >> PAGE_SHIFT) << PAGE_SHIFT)
++
++#if PAGES_NR > KEXEC_XEN_NO_PAGES
++#error PAGES_NR is greater than KEXEC_XEN_NO_PAGES - Xen support will break
++#endif
++
++#if PA_CONTROL_PAGE != 0
++#error PA_CONTROL_PAGE is non zero - Xen support will break
++#endif
++
++void machine_kexec_setup_load_arg(xen_kexec_image_t *xki, struct kimage *image)
++{
++ void *control_page;
++ void *table_page;
++
++ memset(xki->page_list, 0, sizeof(xki->page_list));
++
++ control_page = page_address(image->control_code_page) + PAGE_SIZE;
++ memcpy(control_page, relocate_kernel, PAGE_SIZE);
++
++ table_page = page_address(image->control_code_page);
++
++ xki->page_list[PA_CONTROL_PAGE] = __ma(control_page);
++ xki->page_list[PA_TABLE_PAGE] = __ma(table_page);
++
++ xki->page_list[PA_PGD] = __ma(kexec_pgd);
++ xki->page_list[PA_PUD_0] = __ma(kexec_pud0);
++ xki->page_list[PA_PUD_1] = __ma(kexec_pud1);
++ xki->page_list[PA_PMD_0] = __ma(kexec_pmd0);
++ xki->page_list[PA_PMD_1] = __ma(kexec_pmd1);
++ xki->page_list[PA_PTE_0] = __ma(kexec_pte0);
++ xki->page_list[PA_PTE_1] = __ma(kexec_pte1);
++}
++
++#else /* CONFIG_XEN */
++
++#define x__pmd(x) __pmd(x)
++#define x__pud(x) __pud(x)
++#define x__pgd(x) __pgd(x)
++
++#define x_set_pmd(x, y) set_pmd(x, y)
++#define x_set_pud(x, y) set_pud(x, y)
++#define x_set_pgd(x, y) set_pgd(x, y)
++
++#define x_pud_clear(x) pud_clear(x)
++#define x_pgd_clear(x) pgd_clear(x)
++
++#define X__PAGE_KERNEL_LARGE_EXEC __PAGE_KERNEL_LARGE_EXEC
++#define X_KERNPG_TABLE _KERNPG_TABLE
++
++#endif /* CONFIG_XEN */
++
+ static void init_level2_page(pmd_t *level2p, unsigned long addr)
+ {
+ unsigned long end_addr;
+@@ -31,7 +129,7 @@ static void init_level2_page(pmd_t *leve
+ addr &= PAGE_MASK;
+ end_addr = addr + PUD_SIZE;
+ while (addr < end_addr) {
+- set_pmd(level2p++, __pmd(addr | __PAGE_KERNEL_LARGE_EXEC));
++ x_set_pmd(level2p++, x__pmd(addr | X__PAGE_KERNEL_LARGE_EXEC));
+ addr += PMD_SIZE;
+ }
+ }
+@@ -56,12 +154,12 @@ static int init_level3_page(struct kimag
+ }
+ level2p = (pmd_t *)page_address(page);
+ init_level2_page(level2p, addr);
+- set_pud(level3p++, __pud(__pa(level2p) | _KERNPG_TABLE));
++ x_set_pud(level3p++, x__pud(__pa(level2p) | X_KERNPG_TABLE));
+ addr += PUD_SIZE;
+ }
+ /* clear the unused entries */
+ while (addr < end_addr) {
+- pud_clear(level3p++);
++ x_pud_clear(level3p++);
+ addr += PUD_SIZE;
+ }
+ out:
+@@ -92,12 +190,12 @@ static int init_level4_page(struct kimag
+ if (result) {
+ goto out;
+ }
+- set_pgd(level4p++, __pgd(__pa(level3p) | _KERNPG_TABLE));
++ x_set_pgd(level4p++, x__pgd(__pa(level3p) | X_KERNPG_TABLE));
+ addr += PGDIR_SIZE;
+ }
+ /* clear the unused entries */
+ while (addr < end_addr) {
+- pgd_clear(level4p++);
++ x_pgd_clear(level4p++);
+ addr += PGDIR_SIZE;
+ }
+ out:
+@@ -108,8 +206,14 @@ out:
+ static int init_pgtable(struct kimage *image, unsigned long start_pgtable)
+ {
+ pgd_t *level4p;
++ unsigned long x_end_pfn = end_pfn;
++
++#ifdef CONFIG_XEN
++ x_end_pfn = HYPERVISOR_memory_op(XENMEM_maximum_ram_page, NULL);
++#endif
++
+ level4p = (pgd_t *)__va(start_pgtable);
+- return init_level4_page(image, level4p, 0, end_pfn << PAGE_SHIFT);
++ return init_level4_page(image, level4p, 0, x_end_pfn << PAGE_SHIFT);
+ }
+
+ int machine_kexec_prepare(struct kimage *image)
--- 0005/patches/linux-2.6.16.29/series
+++ work/patches/linux-2.6.16.29/series
@@ -5,6 +5,10 @@ git-3566561bfadffcb5dbc85d576be80c0dbf2c
linux-2.6.19-rc1-kexec-move_segment_code-i386.patch
linux-2.6.19-rc1-kexec-xen-i386.patch
linux-2.6.16.29-crash-xen-i386.patch
+git-4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f.patch
+linux-2.6.19-rc1-kexec-move_segment_code-x86_64.patch
+linux-2.6.19-rc1-kexec-xen-x86_64.patch
+linux-2.6.16.29-crash-xen-x86_64.patch
blktap-aio-16_03_06.patch
device_bind.patch
fix-hz-suspend.patch
--- 0001/xen/arch/x86/x86_64/entry.S
+++ work/xen/arch/x86/x86_64/entry.S
@@ -559,6 +559,7 @@ ENTRY(hypercall_table)
.quad do_hvm_op
.quad do_sysctl /* 35 */
.quad do_domctl
+ .quad do_kexec_op
.rept NR_hypercalls-((.-hypercall_table)/8)
.quad do_ni_hypercall
.endr
@@ -601,6 +602,7 @@ ENTRY(hypercall_args_table)
.byte 2 /* do_hvm_op */
.byte 1 /* do_sysctl */ /* 35 */
.byte 1 /* do_domctl */
+ .byte 2 /* do_kexec */
.rept NR_hypercalls-(.-hypercall_args_table)
.byte 0 /* do_ni_hypercall */
.endr
--- 0004/xen/include/asm-x86/x86_64/elf.h
+++ work/xen/include/asm-x86/x86_64/elf.h
@@ -1,15 +1,67 @@
#ifndef __X86_64_ELF_H__
#define __X86_64_ELF_H__
-#include <xen/lib.h> /* for printk() used in stub */
+#include <asm/processor.h>
typedef struct {
- unsigned long dummy;
+ unsigned long r15;
+ unsigned long r14;
+ unsigned long r13;
+ unsigned long r12;
+ unsigned long rbp;
+ unsigned long rbx;
+ unsigned long r11;
+ unsigned long r10;
+ unsigned long r9;
+ unsigned long r8;
+ unsigned long rax;
+ unsigned long rcx;
+ unsigned long rdx;
+ unsigned long rsi;
+ unsigned long rdi;
+ unsigned long orig_rax;
+ unsigned long rip;
+ unsigned long cs;
+ unsigned long eflags;
+ unsigned long rsp;
+ unsigned long ss;
+ unsigned long thread_fs;
+ unsigned long thread_gs;
+ unsigned long ds;
+ unsigned long es;
+ unsigned long fs;
+ unsigned long gs;
} ELF_Gregset;
extern inline void elf_core_save_regs(ELF_Gregset *dst)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ asm volatile("movq %%r15,%0" : "=m"(dst->r15));
+ asm volatile("movq %%r14,%0" : "=m"(dst->r14));
+ asm volatile("movq %%r13,%0" : "=m"(dst->r13));
+ asm volatile("movq %%r12,%0" : "=m"(dst->r12));
+ asm volatile("movq %%rbp,%0" : "=m"(dst->rbp));
+ asm volatile("movq %%rbx,%0" : "=m"(dst->rbx));
+ asm volatile("movq %%r11,%0" : "=m"(dst->r11));
+ asm volatile("movq %%r10,%0" : "=m"(dst->r10));
+ asm volatile("movq %%r9,%0" : "=m"(dst->r9));
+ asm volatile("movq %%r8,%0" : "=m"(dst->r8));
+ asm volatile("movq %%rax,%0" : "=m"(dst->rax));
+ asm volatile("movq %%rcx,%0" : "=m"(dst->rcx));
+ asm volatile("movq %%rdx,%0" : "=m"(dst->rdx));
+ asm volatile("movq %%rsi,%0" : "=m"(dst->rsi));
+ asm volatile("movq %%rdi,%0" : "=m"(dst->rdi));
+ /* orig_rax not filled in for now */
+ dst->rip = (unsigned long)current_text_addr();
+ asm volatile("movl %%cs, %%eax;" :"=a"(dst->cs));
+ asm volatile("pushfq; popq %0" :"=m"(dst->eflags));
+ asm volatile("movq %%rsp,%0" : "=m"(dst->rsp));
+ asm volatile("movl %%ss, %%eax;" :"=a"(dst->ss));
+ /* thread_fs not filled in for now */
+ /* thread_gs not filled in for now */
+ asm volatile("movl %%ds, %%eax;" :"=a"(dst->ds));
+ asm volatile("movl %%es, %%eax;" :"=a"(dst->es));
+ asm volatile("movl %%fs, %%eax;" :"=a"(dst->fs));
+ asm volatile("movl %%gs, %%eax;" :"=a"(dst->gs));
}
#endif /* __X86_64_ELF_H__ */
--- 0004/xen/include/asm-x86/x86_64/kexec.h
+++ work/xen/include/asm-x86/x86_64/kexec.h
@@ -1,14 +1,29 @@
+/******************************************************************************
+ * kexec.h
+ *
+ * Based heavily on machine_kexec.c and kexec.h from Linux 2.6.19-rc1
+ *
+ */
+
#ifndef __X86_64_KEXEC_H__
#define __X86_64_KEXEC_H__
-
-#include <xen/lib.h> /* for printk() used in stub */
+
#include <xen/types.h>
-#include <public/xen.h>
#include <xen/kexec.h>
+#include <asm/fixmap.h>
+
+typedef void (*relocate_new_kernel_t)(
+ unsigned long indirection_page,
+ unsigned long page_list,
+ unsigned long start_address);
static inline void machine_kexec(xen_kexec_image_t *image)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ relocate_new_kernel_t rnk;
+
+ rnk = (relocate_new_kernel_t) image->page_list[1];
+ (*rnk)(image->indirection_page, (unsigned long)image->page_list,
+ image->start_address);
}
#endif /* __X86_64_KEXEC_H__ */
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-22 7:10 [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502) Magnus Damm
` (3 preceding siblings ...)
2006-11-22 7:11 ` [PATCH 04/04] Kexec / Kdump: x86_64 " Magnus Damm
@ 2006-11-22 18:24 ` Ian Campbell
2006-11-27 9:19 ` Magnus Damm
2006-11-28 18:24 ` Ian Campbell
5 siblings, 1 reply; 30+ messages in thread
From: Ian Campbell @ 2006-11-22 18:24 UTC (permalink / raw)
To: Magnus Damm
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, magnus.damm, Akio Takebe,
Isaku Yamahata, Horms
Hi Magnus,
On Wed, 2006-11-22 at 16:10 +0900, Magnus Damm wrote:
> Here comes a new version of the Kexec / Kdump patches for x86 Xen. Not much
> has changed since last release, just a minor fix for kdump on x86_64.
>
> Patches to make kexec-tools Xen aware have recently been sent to the fastboot
> list. These patches will be merged in the kexec-tools-testing tree in the
> near future.
I've taken these patches out for a spin. They look pretty good. I've got
a couple of comments.
Firstly the patches break native kernel compile. You add usages of
kexec_page_to_pfn() and friends to kernel/kexec.c but only include
kexec-xen.h ifdef CONFIG_XEN. I fixed it with by removing the ifdef but
the preferred way would be to move the native definitions of kexec_*
into include/asm-i386/kexec.h and make a xen specific copy in
include/asm-i386/mach-xen/asm/kexec.h with the xen versions patched in.
Alternatively you could just merge kexec-xen.h into kexec.h.
My second comment is WRT to the ELF notes which you add to the kdump.
You include a standard PRSTATUS core ELF note per physical CPU but there
is some useful physical processor state which is not included in this
structure -- most importantly CR3.
Since the amount of physical CPU state which is not already included in
PRSTATUS is pretty small I think you could just include the whole lot in
a Xen specific note per PCPU. I'd basically include anything which is in
a Xen panic/oops message but not in PRSTATUS, that's C[0,2,3,4].
Including the debug registers might be handy too. If there was some
standard extended PRSTATUS note format for these extra things we could
use that would be even better but I don't know of one (but then I don't
really know about these things ;-)).
You also store dom0's pfn_to_mfn_frame_list_list in a Xen specific note.
What is that used for? Given a Xen symbol table it should be possible to
locate the shared info for any domain via the xen mappings and hence
find the p2m table that way. m2p is at a known virtual address already.
The contents of the h/v taint bitmap would be another interesting thing
to include in the Xen note.
Cheers,
Ian.
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-22 18:24 ` [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502) Ian Campbell
@ 2006-11-27 9:19 ` Magnus Damm
2006-11-27 12:09 ` Ian Campbell
2006-11-27 15:27 ` Dave Anderson
0 siblings, 2 replies; 30+ messages in thread
From: Magnus Damm @ 2006-11-27 9:19 UTC (permalink / raw)
To: Ian Campbell
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, Isaku Yamahata,
Magnus Damm, Horms, Dave Anderson
Hi Ian,
[Added Dave Anderson to the CC list]
Thanks for the comments!
On 11/23/06, Ian Campbell <Ian.Campbell@xensource.com> wrote:
> Hi Magnus,
>
> On Wed, 2006-11-22 at 16:10 +0900, Magnus Damm wrote:
> > Here comes a new version of the Kexec / Kdump patches for x86 Xen. Not much
> > has changed since last release, just a minor fix for kdump on x86_64.
> >
> > Patches to make kexec-tools Xen aware have recently been sent to the fastboot
> > list. These patches will be merged in the kexec-tools-testing tree in the
> > near future.
>
> I've taken these patches out for a spin. They look pretty good. I've got
> a couple of comments.
>
> Firstly the patches break native kernel compile. You add usages of
> kexec_page_to_pfn() and friends to kernel/kexec.c but only include
> kexec-xen.h ifdef CONFIG_XEN. I fixed it with by removing the ifdef but
> the preferred way would be to move the native definitions of kexec_*
> into include/asm-i386/kexec.h and make a xen specific copy in
> include/asm-i386/mach-xen/asm/kexec.h with the xen versions patched in.
> Alternatively you could just merge kexec-xen.h into kexec.h.
Sorry about the breakage. I'd like to stay away from duplicating files
so I think merging kexec-xen.h into kexec.h sounds good. I plan to add
the common non-xen version of the page macros to include/linux/kexec.h
and add the xen-specific macros to the per-architecture
include/asm/kexec.h.
I'll include a fix in the patchset that I'll send later this week.
> My second comment is WRT to the ELF notes which you add to the kdump.
> You include a standard PRSTATUS core ELF note per physical CPU but there
> is some useful physical processor state which is not included in this
> structure -- most importantly CR3.
This structure is used both for regular Linux kdumps and core dumps so
it felt natural to extend it to the Xen case as well. I do however
agree with you that it is strange that only certain registers are
saved and many system-level processor registers are unsaved.
I've discussed this with Dave Anderson, and he needed CR3 to be able
to locate certain mapping tables used for converting between
pseudo-physical and machine addresses.
http://www.mail-archive.com/crash-utility@redhat.com/msg00201.html
> Since the amount of physical CPU state which is not already included in
> PRSTATUS is pretty small I think you could just include the whole lot in
> a Xen specific note per PCPU. I'd basically include anything which is in
> a Xen panic/oops message but not in PRSTATUS, that's C[0,2,3,4].
> Including the debug registers might be handy too. If there was some
> standard extended PRSTATUS note format for these extra things we could
> use that would be even better but I don't know of one (but then I don't
> really know about these things ;-)).
I'm unfortunately not aware of any standard format.
The current Xen specific note is only written once and it is used to
give system-wide information, ie not per cpu information. So maybe it
makes sense to create a new per-cpu note for system-level register
information?
> You also store dom0's pfn_to_mfn_frame_list_list in a Xen specific note.
> What is that used for? Given a Xen symbol table it should be possible to
> locate the shared info for any domain via the xen mappings and hence
> find the p2m table that way. m2p is at a known virtual address already.
This is because Dave wanted to be able to parse dom0 kernels easily.
I'm not sure if that is the case still with the new xencrash code?
Dave, are you listening?
I thought that pointing out pfn_to_mfn_frame_list_list for dom0 was a
better, more portable way to provide Dave with this info than just
handing out CR3.
> The contents of the h/v taint bitmap would be another interesting thing
> to include in the Xen note.
This sounds like system-wide information, not per-cpu right?
Thanks,
/ magnus
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-27 9:19 ` Magnus Damm
@ 2006-11-27 12:09 ` Ian Campbell
2006-11-28 8:28 ` Magnus Damm
2006-11-27 15:27 ` Dave Anderson
1 sibling, 1 reply; 30+ messages in thread
From: Ian Campbell @ 2006-11-27 12:09 UTC (permalink / raw)
To: Magnus Damm
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, Isaku Yamahata,
Magnus Damm, Horms, Dave Anderson
On Mon, 2006-11-27 at 18:19 +0900, Magnus Damm wrote:
> On 11/23/06, Ian Campbell <Ian.Campbell@xensource.com> wrote:
> > Firstly the patches break native kernel compile. [snip]
>
> Sorry about the breakage. I'd like to stay away from duplicating files
> so I think merging kexec-xen.h into kexec.h sounds good. I plan to add
> the common non-xen version of the page macros to include/linux/kexec.h
> and add the xen-specific macros to the per-architecture
> include/asm/kexec.h.
>
> I'll include a fix in the patchset that I'll send later this week.
Great, thanks.
> > My second comment is WRT to the ELF notes which you add to the kdump.
> > You include a standard PRSTATUS core ELF note per physical CPU but there
> > is some useful physical processor state which is not included in this
> > structure -- most importantly CR3.
>
> This structure is used both for regular Linux kdumps and core dumps so
> it felt natural to extend it to the Xen case as well.
Definitely.
> I do however
> agree with you that it is strange that only certain registers are
> saved and many system-level processor registers are unsaved.
I can see why they were not included since they aren't really useful in
the normal core dump case. Perhaps it is worth talking to the native
kdump people about defining a new core note type which includes the
extended processor state that isn't in the regular core note?
We can go with a Xen specific one for now and transition to a common one
later if necessary.
> The current Xen specific note is only written once and it is used to
> give system-wide information, ie not per cpu information. So maybe it
> makes sense to create a new per-cpu note for system-level register
> information?
That makes sense to me.
Could you also #define the note types in a header somewhere. Perhaps
xen/include/public/kexec.h or xen/include/public/elfnote.h?
> > You also store dom0's pfn_to_mfn_frame_list_list in a Xen specific note.
> > What is that used for? Given a Xen symbol table it should be possible to
> > locate the shared info for any domain via the xen mappings and hence
> > find the p2m table that way. m2p is at a known virtual address already.
>
> This is because Dave wanted to be able to parse dom0 kernels easily.
> I'm not sure if that is the case still with the new xencrash code?
> Dave, are you listening?
>
> I thought that pointing out pfn_to_mfn_frame_list_list for dom0 was a
> better, more portable way to provide Dave with this info than just
> handing out CR3.
Unless you provide this list for all domains[*] the CR3 method will have
to be implemented anyway so domains != 0 can be examined. In particular
it could be useful to examine the domain which made the hypercall which
led to a crash and that might not necessarily be dom0 (although I
suppose it is most likely).
[*]If I understand correctly saving per domain information is not
possible because the notes need to be created when the kdump kernel is
loaded and the number of domains is unknown at that time.
> > The contents of the h/v taint bitmap would be another interesting thing
> > to include in the Xen note.
>
> This sounds like system-wide information, not per-cpu right?
Yep.
Cheers,
Ian.
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-27 12:09 ` Ian Campbell
@ 2006-11-28 8:28 ` Magnus Damm
2006-11-28 9:26 ` Ian Campbell
0 siblings, 1 reply; 30+ messages in thread
From: Magnus Damm @ 2006-11-28 8:28 UTC (permalink / raw)
To: Ian Campbell
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, Isaku Yamahata,
Magnus Damm, Horms, Dave Anderson
On 11/27/06, Ian Campbell <Ian.Campbell@xensource.com> wrote:
> On Mon, 2006-11-27 at 18:19 +0900, Magnus Damm wrote:
> > I do however
> > agree with you that it is strange that only certain registers are
> > saved and many system-level processor registers are unsaved.
>
> I can see why they were not included since they aren't really useful in
> the normal core dump case. Perhaps it is worth talking to the native
> kdump people about defining a new core note type which includes the
> extended processor state that isn't in the regular core note?
That sounds like a good idea, at least in theory. The number of
patches-per-year accepted for the kexec kernel code is unfortunately
pretty low...
> We can go with a Xen specific one for now and transition to a common one
> later if necessary.
I think that is better solution, at least for now.
> > The current Xen specific note is only written once and it is used to
> > give system-wide information, ie not per cpu information. So maybe it
> > makes sense to create a new per-cpu note for system-level register
> > information?
>
> That makes sense to me.
>
> Could you also #define the note types in a header somewhere. Perhaps
> xen/include/public/kexec.h or xen/include/public/elfnote.h?
Do you mean the data structures or the type value used in the elf note
header? Part of the data structures are currently dragging in
architecture-specific stuff, so I'm not that tempted... The tools
that use the data structures duplicate them anyhow, but maybe it's a
good idea.
> > > You also store dom0's pfn_to_mfn_frame_list_list in a Xen specific note.
> > > What is that used for? Given a Xen symbol table it should be possible to
> > > locate the shared info for any domain via the xen mappings and hence
> > > find the p2m table that way. m2p is at a known virtual address already.
> >
> > This is because Dave wanted to be able to parse dom0 kernels easily.
> > I'm not sure if that is the case still with the new xencrash code?
> > Dave, are you listening?
> >
> > I thought that pointing out pfn_to_mfn_frame_list_list for dom0 was a
> > better, more portable way to provide Dave with this info than just
> > handing out CR3.
>
> Unless you provide this list for all domains[*] the CR3 method will have
> to be implemented anyway so domains != 0 can be examined. In particular
> it could be useful to examine the domain which made the hypercall which
> led to a crash and that might not necessarily be dom0 (although I
> suppose it is most likely).
This was just to make it easy to support dom0 only. Extracting other
domains are done through backtracking of symbols and data structures
which is independent of elf notes.
> [*]If I understand correctly saving per domain information is not
> possible because the notes need to be created when the kdump kernel is
> loaded and the number of domains is unknown at that time.
Correct!
> > > The contents of the h/v taint bitmap would be another interesting thing
> > > to include in the Xen note.
> >
> > This sounds like system-wide information, not per-cpu right?
I've added that one now. Thanks!
/ magnus
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-28 8:28 ` Magnus Damm
@ 2006-11-28 9:26 ` Ian Campbell
0 siblings, 0 replies; 30+ messages in thread
From: Ian Campbell @ 2006-11-28 9:26 UTC (permalink / raw)
To: Magnus Damm
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, Isaku Yamahata,
Magnus Damm, Horms, Dave Anderson
On Tue, 2006-11-28 at 17:28 +0900, Magnus Damm wrote:
> On 11/27/06, Ian Campbell <Ian.Campbell@xensource.com> wrote:
> > On Mon, 2006-11-27 at 18:19 +0900, Magnus Damm wrote:
> > > I do however
> > > agree with you that it is strange that only certain registers are
> > > saved and many system-level processor registers are unsaved.
> >
> > I can see why they were not included since they aren't really useful in
> > the normal core dump case. Perhaps it is worth talking to the native
> > kdump people about defining a new core note type which includes the
> > extended processor state that isn't in the regular core note?
>
> That sounds like a good idea, at least in theory. The number of
> patches-per-year accepted for the kexec kernel code is unfortunately
> pretty low...
>
> > We can go with a Xen specific one for now and transition to a common one
> > later if necessary.
>
> I think that is better solution, at least for now.
OK.
> > > The current Xen specific note is only written once and it is used to
> > > give system-wide information, ie not per cpu information. So maybe it
> > > makes sense to create a new per-cpu note for system-level register
> > > information?
> >
> > That makes sense to me.
> >
> > Could you also #define the note types in a header somewhere. Perhaps
> > xen/include/public/kexec.h or xen/include/public/elfnote.h?
>
> Do you mean the data structures or the type value used in the elf note
> header? Part of the data structures are currently dragging in
> architecture-specific stuff, so I'm not that tempted... The tools
> that use the data structures duplicate them anyhow, but maybe it's a
> good idea.
I was talking about the note type values but now that you mention it the
data structures would be good too ;-) I think it's useful to have shared
types & data structures defined in a shared location to avoid confusion.
It's something we can deal with some other time though.
> > > I thought that pointing out pfn_to_mfn_frame_list_list for dom0 was a
> > > better, more portable way to provide Dave with this info than just
> > > handing out CR3.
> >
> > Unless you provide this list for all domains[*] the CR3 method will have
> > to be implemented anyway so domains != 0 can be examined. In particular
> > it could be useful to examine the domain which made the hypercall which
> > led to a crash and that might not necessarily be dom0 (although I
> > suppose it is most likely).
>
> This was just to make it easy to support dom0 only. Extracting other
> domains are done through backtracking of symbols and data structures
> which is independent of elf notes.
It seems to me that once backtracking code exists (it already does?) it
can be used for dom0 too and the field becomes unnecessary.
Anyway, I don't feel that strongly about it and it's only one extra
field so leave it in if you think it is useful.
> > > > The contents of the h/v taint bitmap would be another interesting thing
> > > > to include in the Xen note.
> > >
> > > This sounds like system-wide information, not per-cpu right?
>
> I've added that one now. Thanks!
Cheers.
Ian.
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-27 9:19 ` Magnus Damm
2006-11-27 12:09 ` Ian Campbell
@ 2006-11-27 15:27 ` Dave Anderson
2006-11-28 8:30 ` Magnus Damm
1 sibling, 1 reply; 30+ messages in thread
From: Dave Anderson @ 2006-11-27 15:27 UTC (permalink / raw)
To: Magnus Damm
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Ian Campbell, Akio Takebe,
Isaku Yamahata, Magnus Damm, Horms
Magnus Damm wrote:
> Hi Ian,
>
> [Added Dave Anderson to the CC list]
>
> Thanks for the comments!
>
> On 11/23/06, Ian Campbell <Ian.Campbell@xensource.com> wrote:
> > Hi Magnus,
> >
> > On Wed, 2006-11-22 at 16:10 +0900, Magnus Damm wrote:
> > > Here comes a new version of the Kexec / Kdump patches for x86 Xen. Not much
> > > has changed since last release, just a minor fix for kdump on x86_64.
> > >
> > > Patches to make kexec-tools Xen aware have recently been sent to the fastboot
> > > list. These patches will be merged in the kexec-tools-testing tree in the
> > > near future.
> >
> > I've taken these patches out for a spin. They look pretty good. I've got
> > a couple of comments.
> >
> > Firstly the patches break native kernel compile. You add usages of
> > kexec_page_to_pfn() and friends to kernel/kexec.c but only include
> > kexec-xen.h ifdef CONFIG_XEN. I fixed it with by removing the ifdef but
> > the preferred way would be to move the native definitions of kexec_*
> > into include/asm-i386/kexec.h and make a xen specific copy in
> > include/asm-i386/mach-xen/asm/kexec.h with the xen versions patched in.
> > Alternatively you could just merge kexec-xen.h into kexec.h.
>
> Sorry about the breakage. I'd like to stay away from duplicating files
> so I think merging kexec-xen.h into kexec.h sounds good. I plan to add
> the common non-xen version of the page macros to include/linux/kexec.h
> and add the xen-specific macros to the per-architecture
> include/asm/kexec.h.
>
> I'll include a fix in the patchset that I'll send later this week.
>
> > My second comment is WRT to the ELF notes which you add to the kdump.
> > You include a standard PRSTATUS core ELF note per physical CPU but there
> > is some useful physical processor state which is not included in this
> > structure -- most importantly CR3.
>
> This structure is used both for regular Linux kdumps and core dumps so
> it felt natural to extend it to the Xen case as well. I do however
> agree with you that it is strange that only certain registers are
> saved and many system-level processor registers are unsaved.
>
> I've discussed this with Dave Anderson, and he needed CR3 to be able
> to locate certain mapping tables used for converting between
> pseudo-physical and machine addresses.
>
> http://www.mail-archive.com/crash-utility@redhat.com/msg00201.html
>
The discussion boiled down to this:
If we want to be able to analyze the vmcore file from the dom0's kernel
perspective with the crash utility, we need some manner of being able to
recreate the dom0 kernel's phys_to_machine_mapping array. In an older
prototype, a dom0 CR3 value was passed in an ELF note, hence the essentially
obsolete note name "NT_XEN_KDUMP_CR3". In subsequent discussions,
it was agreed that the pfn_to_mfn_frame_list_list value was preferable
to a dom0 CR3 value.
>
> > Since the amount of physical CPU state which is not already included in
> > PRSTATUS is pretty small I think you could just include the whole lot in
> > a Xen specific note per PCPU. I'd basically include anything which is in
> > a Xen panic/oops message but not in PRSTATUS, that's C[0,2,3,4].
> > Including the debug registers might be handy too. If there was some
> > standard extended PRSTATUS note format for these extra things we could
> > use that would be even better but I don't know of one (but then I don't
> > really know about these things ;-)).
>
> I'm unfortunately not aware of any standard format.
>
> The current Xen specific note is only written once and it is used to
> give system-wide information, ie not per cpu information. So maybe it
> makes sense to create a new per-cpu note for system-level register
> information?
>
> > You also store dom0's pfn_to_mfn_frame_list_list in a Xen specific note.
> > What is that used for? Given a Xen symbol table it should be possible to
> > locate the shared info for any domain via the xen mappings and hence
> > find the p2m table that way. m2p is at a known virtual address already.
>
> This is because Dave wanted to be able to parse dom0 kernels easily.
> I'm not sure if that is the case still with the new xencrash code?
> Dave, are you listening?
>
> I thought that pointing out pfn_to_mfn_frame_list_list for dom0 was a
> better, more portable way to provide Dave with this info than just
> handing out CR3.
>
That's correct. With this simple additional note, it will be possible
to do any of the following:
$ crash vmlinux vmcore (with the NT_XEN_KDUMP_CR3 note)
$ crash xen-syms vmcore (with the new "xencrash" patch)
$ gdb xen-syms vmcore
I had originally suggested an array of pfn_to_mfn_frame_list_list values,
one for each guest domain, in which case you could execute crash
sessions for dom0 or any of the guest domains.
And now with the new xencrash patch, the crash utility can also
be run against the xen-syms binary, which means that the session
will be run from the perspective of the hypervisor binary, i.e., with its
own set of hypervisor-specific commands.
Anyway, we compromised on just the dom0 pfn_to_mfn_frame_list_list
value, given that in the majority of dom0/hypervisor crashes, the cause
of the crash will most likely be in the dom0 kernel code.
Dave
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-27 15:27 ` Dave Anderson
@ 2006-11-28 8:30 ` Magnus Damm
2006-11-28 14:01 ` Dave Anderson
0 siblings, 1 reply; 30+ messages in thread
From: Magnus Damm @ 2006-11-28 8:30 UTC (permalink / raw)
To: Dave Anderson
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Ian Campbell, Akio Takebe,
Isaku Yamahata, Magnus Damm, Horms
On 11/28/06, Dave Anderson <anderson@redhat.com> wrote:
> Magnus Damm wrote:
> > I thought that pointing out pfn_to_mfn_frame_list_list for dom0 was a
> > better, more portable way to provide Dave with this info than just
> > handing out CR3.
> >
>
> That's correct. With this simple additional note, it will be possible
> to do any of the following:
>
> $ crash vmlinux vmcore (with the NT_XEN_KDUMP_CR3 note)
> $ crash xen-syms vmcore (with the new "xencrash" patch)
> $ gdb xen-syms vmcore
>
> I had originally suggested an array of pfn_to_mfn_frame_list_list values,
> one for each guest domain, in which case you could execute crash
> sessions for dom0 or any of the guest domains.
>
> And now with the new xencrash patch, the crash utility can also
> be run against the xen-syms binary, which means that the session
> will be run from the perspective of the hypervisor binary, i.e., with its
> own set of hypervisor-specific commands.
>
> Anyway, we compromised on just the dom0 pfn_to_mfn_frame_list_list
> value, given that in the majority of dom0/hypervisor crashes, the cause
> of the crash will most likely be in the dom0 kernel code.
Thanks for the comments! The value for NT_XEN_KDUMP_CR3, do you have
any strong feelings to keep that? We have an unique string now, so I'm
tempted to just set it to 0...
/ magnus
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-28 8:30 ` Magnus Damm
@ 2006-11-28 14:01 ` Dave Anderson
2006-11-29 2:35 ` Magnus Damm
0 siblings, 1 reply; 30+ messages in thread
From: Dave Anderson @ 2006-11-28 14:01 UTC (permalink / raw)
To: Magnus Damm
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Ian Campbell, Akio Takebe,
Isaku Yamahata, Magnus Damm, Horms
Magnus Damm wrote:
> On 11/28/06, Dave Anderson <anderson@redhat.com> wrote:
> > Magnus Damm wrote:
> > > I thought that pointing out pfn_to_mfn_frame_list_list for dom0 was a
> > > better, more portable way to provide Dave with this info than just
> > > handing out CR3.
> > >
> >
> > That's correct. With this simple additional note, it will be possible
> > to do any of the following:
> >
> > $ crash vmlinux vmcore (with the NT_XEN_KDUMP_CR3 note)
> > $ crash xen-syms vmcore (with the new "xencrash" patch)
> > $ gdb xen-syms vmcore
> >
> > I had originally suggested an array of pfn_to_mfn_frame_list_list values,
> > one for each guest domain, in which case you could execute crash
> > sessions for dom0 or any of the guest domains.
> >
> > And now with the new xencrash patch, the crash utility can also
> > be run against the xen-syms binary, which means that the session
> > will be run from the perspective of the hypervisor binary, i.e., with its
> > own set of hypervisor-specific commands.
> >
> > Anyway, we compromised on just the dom0 pfn_to_mfn_frame_list_list
> > value, given that in the majority of dom0/hypervisor crashes, the cause
> > of the crash will most likely be in the dom0 kernel code.
>
> Thanks for the comments! The value for NT_XEN_KDUMP_CR3, do you have
> any strong feelings to keep that? We have an unique string now, so I'm
> tempted to just set it to 0...
>
Obviously as long as it doesn't clash with the NT_XXX type #define's in elf.h,
it's OK. For diskdump vmcores, they created an un-clashable NT_DISKDUMP
type value of 0x70000001, and so the use of 0x10000001 for NT_XEN_KDUMP_CR3
apparently followed that model. But, since you're asking, I don't particularly
like the use of 0; it just seems too ambiguous.
Dave
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-28 14:01 ` Dave Anderson
@ 2006-11-29 2:35 ` Magnus Damm
2006-11-29 9:36 ` Ian Campbell
0 siblings, 1 reply; 30+ messages in thread
From: Magnus Damm @ 2006-11-29 2:35 UTC (permalink / raw)
To: Dave Anderson
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Ian Campbell, Akio Takebe,
Isaku Yamahata, Magnus Damm, Horms
On 11/28/06, Dave Anderson <anderson@redhat.com> wrote:
> Magnus Damm wrote:
>
> > On 11/28/06, Dave Anderson <anderson@redhat.com> wrote:
> > > Magnus Damm wrote:
> > > > I thought that pointing out pfn_to_mfn_frame_list_list for dom0 was a
> > > > better, more portable way to provide Dave with this info than just
> > > > handing out CR3.
> > > >
> > >
> > > That's correct. With this simple additional note, it will be possible
> > > to do any of the following:
> > >
> > > $ crash vmlinux vmcore (with the NT_XEN_KDUMP_CR3 note)
> > > $ crash xen-syms vmcore (with the new "xencrash" patch)
> > > $ gdb xen-syms vmcore
> > >
> > > I had originally suggested an array of pfn_to_mfn_frame_list_list values,
> > > one for each guest domain, in which case you could execute crash
> > > sessions for dom0 or any of the guest domains.
> > >
> > > And now with the new xencrash patch, the crash utility can also
> > > be run against the xen-syms binary, which means that the session
> > > will be run from the perspective of the hypervisor binary, i.e., with its
> > > own set of hypervisor-specific commands.
> > >
> > > Anyway, we compromised on just the dom0 pfn_to_mfn_frame_list_list
> > > value, given that in the majority of dom0/hypervisor crashes, the cause
> > > of the crash will most likely be in the dom0 kernel code.
> >
> > Thanks for the comments! The value for NT_XEN_KDUMP_CR3, do you have
> > any strong feelings to keep that? We have an unique string now, so I'm
> > tempted to just set it to 0...
> >
>
> Obviously as long as it doesn't clash with the NT_XXX type #define's in elf.h,
> it's OK. For diskdump vmcores, they created an un-clashable NT_DISKDUMP
> type value of 0x70000001, and so the use of 0x10000001 for NT_XEN_KDUMP_CR3
> apparently followed that model. But, since you're asking, I don't particularly
> like the use of 0; it just seems too ambiguous.
Hm... But doesn't both the string and the type together form an unique
identifier?
I think the "CORE" string together with the NT_PRSTATUS type point out
a structure of type ELF_Prstatus. In the case of our Xen specific
notes we use "XEN CORE" as string and therefore are free to use any
type because we are already "unique enough" with our string. So I was
kind of questioning having both an unique string and on top of that an
unique type.
Thanks,
/ magnus
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-29 2:35 ` Magnus Damm
@ 2006-11-29 9:36 ` Ian Campbell
2006-11-29 10:59 ` Magnus Damm
0 siblings, 1 reply; 30+ messages in thread
From: Ian Campbell @ 2006-11-29 9:36 UTC (permalink / raw)
To: Magnus Damm
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, Isaku Yamahata,
Magnus Damm, Horms, Dave Anderson
On Wed, 2006-11-29 at 11:35 +0900, Magnus Damm wrote:
> On 11/28/06, Dave Anderson <anderson@redhat.com> wrote:
> > Magnus Damm wrote:
> >
> > > On 11/28/06, Dave Anderson <anderson@redhat.com> wrote:
> > > > Magnus Damm wrote:
> > > > > I thought that pointing out pfn_to_mfn_frame_list_list for dom0 was a
> > > > > better, more portable way to provide Dave with this info than just
> > > > > handing out CR3.
> > > > >
> > > >
> > > > That's correct. With this simple additional note, it will be possible
> > > > to do any of the following:
> > > >
> > > > $ crash vmlinux vmcore (with the NT_XEN_KDUMP_CR3 note)
> > > > $ crash xen-syms vmcore (with the new "xencrash" patch)
> > > > $ gdb xen-syms vmcore
> > > >
> > > > I had originally suggested an array of pfn_to_mfn_frame_list_list values,
> > > > one for each guest domain, in which case you could execute crash
> > > > sessions for dom0 or any of the guest domains.
> > > >
> > > > And now with the new xencrash patch, the crash utility can also
> > > > be run against the xen-syms binary, which means that the session
> > > > will be run from the perspective of the hypervisor binary, i.e., with its
> > > > own set of hypervisor-specific commands.
> > > >
> > > > Anyway, we compromised on just the dom0 pfn_to_mfn_frame_list_list
> > > > value, given that in the majority of dom0/hypervisor crashes, the cause
> > > > of the crash will most likely be in the dom0 kernel code.
> > >
> > > Thanks for the comments! The value for NT_XEN_KDUMP_CR3, do you have
> > > any strong feelings to keep that? We have an unique string now, so I'm
> > > tempted to just set it to 0...
> > >
> >
> > Obviously as long as it doesn't clash with the NT_XXX type #define's in elf.h,
> > it's OK. For diskdump vmcores, they created an un-clashable NT_DISKDUMP
> > type value of 0x70000001, and so the use of 0x10000001 for NT_XEN_KDUMP_CR3
> > apparently followed that model. But, since you're asking, I don't particularly
> > like the use of 0; it just seems too ambiguous.
>
> Hm... But doesn't both the string and the type together form an unique
> identifier?
That's correct. The type of a note only has to be unique vs notes with
the same name field. So since the Xen notes have name "XEN CORE" they
can't clash with the notes in elf.h which have name "CORE".
If you wanted to you could use a bit of the existing "Xen" namespace. It
is currently only used by notes which are consumed by the domain
builder. They are defined in xen/include/public/elfnote.h.
Ian.
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-29 9:36 ` Ian Campbell
@ 2006-11-29 10:59 ` Magnus Damm
2006-11-29 11:06 ` Ian Campbell
0 siblings, 1 reply; 30+ messages in thread
From: Magnus Damm @ 2006-11-29 10:59 UTC (permalink / raw)
To: Ian Campbell
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, Isaku Yamahata,
Magnus Damm, Horms, Dave Anderson
On 11/29/06, Ian Campbell <Ian.Campbell@xensource.com> wrote:
> On Wed, 2006-11-29 at 11:35 +0900, Magnus Damm wrote:
> > On 11/28/06, Dave Anderson <anderson@redhat.com> wrote:
> > > Magnus Damm wrote:
> > >
> > > > On 11/28/06, Dave Anderson <anderson@redhat.com> wrote:
> > > > > Magnus Damm wrote:
> > > > > > I thought that pointing out pfn_to_mfn_frame_list_list for dom0 was a
> > > > > > better, more portable way to provide Dave with this info than just
> > > > > > handing out CR3.
> > > > > >
> > > > >
> > > > > That's correct. With this simple additional note, it will be possible
> > > > > to do any of the following:
> > > > >
> > > > > $ crash vmlinux vmcore (with the NT_XEN_KDUMP_CR3 note)
> > > > > $ crash xen-syms vmcore (with the new "xencrash" patch)
> > > > > $ gdb xen-syms vmcore
> > > > >
> > > > > I had originally suggested an array of pfn_to_mfn_frame_list_list values,
> > > > > one for each guest domain, in which case you could execute crash
> > > > > sessions for dom0 or any of the guest domains.
> > > > >
> > > > > And now with the new xencrash patch, the crash utility can also
> > > > > be run against the xen-syms binary, which means that the session
> > > > > will be run from the perspective of the hypervisor binary, i.e., with its
> > > > > own set of hypervisor-specific commands.
> > > > >
> > > > > Anyway, we compromised on just the dom0 pfn_to_mfn_frame_list_list
> > > > > value, given that in the majority of dom0/hypervisor crashes, the cause
> > > > > of the crash will most likely be in the dom0 kernel code.
> > > >
> > > > Thanks for the comments! The value for NT_XEN_KDUMP_CR3, do you have
> > > > any strong feelings to keep that? We have an unique string now, so I'm
> > > > tempted to just set it to 0...
> > > >
> > >
> > > Obviously as long as it doesn't clash with the NT_XXX type #define's in elf.h,
> > > it's OK. For diskdump vmcores, they created an un-clashable NT_DISKDUMP
> > > type value of 0x70000001, and so the use of 0x10000001 for NT_XEN_KDUMP_CR3
> > > apparently followed that model. But, since you're asking, I don't particularly
> > > like the use of 0; it just seems too ambiguous.
> >
> > Hm... But doesn't both the string and the type together form an unique
> > identifier?
>
> That's correct. The type of a note only has to be unique vs notes with
> the same name field. So since the Xen notes have name "XEN CORE" they
> can't clash with the notes in elf.h which have name "CORE".
>
> If you wanted to you could use a bit of the existing "Xen" namespace. It
> is currently only used by notes which are consumed by the domain
> builder. They are defined in xen/include/public/elfnote.h.
Sounds like a good idea. What about:
#define XEN_ELFNOTE_CRASH_INFO 13 /* system information */
#define XEN_ELFNOTE_CRASH_REGS 14 /* per-cpu system registers */
The types will map to data structures instead of single numeric values
or strings though, but maybe that is ok.
Thanks,
/ magnus
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-29 10:59 ` Magnus Damm
@ 2006-11-29 11:06 ` Ian Campbell
0 siblings, 0 replies; 30+ messages in thread
From: Ian Campbell @ 2006-11-29 11:06 UTC (permalink / raw)
To: Magnus Damm
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, Isaku Yamahata,
Magnus Damm, Horms, Dave Anderson
On Wed, 2006-11-29 at 19:59 +0900, Magnus Damm wrote:
> > If you wanted to you could use a bit of the existing "Xen" namespace. It
> > is currently only used by notes which are consumed by the domain
> > builder. They are defined in xen/include/public/elfnote.h.
>
> Sounds like a good idea. What about:
>
> #define XEN_ELFNOTE_CRASH_INFO 13 /* system information */
> #define XEN_ELFNOTE_CRASH_REGS 14 /* per-cpu system registers */
Perhaps use 0x1000001 and 0x1000002 to distinguish the types of notes?
(we could rewrite the existing notes as 0x0000000x to make this clearer)
> The types will map to data structures instead of single numeric values
> or strings though, but maybe that is ok.
That's fine. Ultimately I'd like to get the data structures in there too
but don't worry about that for now.
Ian.
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-22 7:10 [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502) Magnus Damm
` (4 preceding siblings ...)
2006-11-22 18:24 ` [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502) Ian Campbell
@ 2006-11-28 18:24 ` Ian Campbell
2006-11-28 18:50 ` Keir Fraser
2006-11-29 4:30 ` Magnus Damm
5 siblings, 2 replies; 30+ messages in thread
From: Ian Campbell @ 2006-11-28 18:24 UTC (permalink / raw)
To: Magnus Damm
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, magnus.damm, Akio Takebe,
Isaku Yamahata, Horms
Hi Magnus,
On Wed, 2006-11-22 at 16:10 +0900, Magnus Damm wrote:
> [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
I've been playing a bit more and found a problem.
You define a per CPU variable crash_notes and on crash you loop over
NR_CPUS and clear the notes for CPUS which don't exist. Unfortunately
the percpu regions for CPUs which aren't physically present is returned
to the heap on boot (see percpu_free_unused_areas) -- this means that
you zero out heap pages on crash :-(
You need to use num_{possible,present,online}_cpus() in
machine_crash_kexec() and kexec_get_cpu() instead of NR_CPUS.
Cheers,
Ian.
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-28 18:24 ` Ian Campbell
@ 2006-11-28 18:50 ` Keir Fraser
2006-11-29 4:30 ` Magnus Damm
1 sibling, 0 replies; 30+ messages in thread
From: Keir Fraser @ 2006-11-28 18:50 UTC (permalink / raw)
To: Ian Campbell, Magnus Damm
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, magnus.damm, Akio Takebe,
Isaku Yamahata, Horms
On 28/11/06 6:24 pm, "Ian Campbell" <Ian.Campbell@XenSource.com> wrote:
> You define a per CPU variable crash_notes and on crash you loop over
> NR_CPUS and clear the notes for CPUS which don't exist. Unfortunately
> the percpu regions for CPUs which aren't physically present is returned
> to the heap on boot (see percpu_free_unused_areas) -- this means that
> you zero out heap pages on crash :-(
>
> You need to use num_{possible,present,online}_cpus() in
> machine_crash_kexec() and kexec_get_cpu() instead of NR_CPUS.
You should use for_each_cpu, for_each_present_cpu, or for_each_online_cpu.
Probably the last of those unless you care about dormant CPUs or hotplugged
CPU sockets. :-)
-- Keir
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-28 18:24 ` Ian Campbell
2006-11-28 18:50 ` Keir Fraser
@ 2006-11-29 4:30 ` Magnus Damm
2006-11-29 7:54 ` Keir Fraser
1 sibling, 1 reply; 30+ messages in thread
From: Magnus Damm @ 2006-11-29 4:30 UTC (permalink / raw)
To: Ian Campbell
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, Isaku Yamahata,
Magnus Damm, Horms
Hi Ian,
On 11/29/06, Ian Campbell <Ian.Campbell@xensource.com> wrote:
> On Wed, 2006-11-22 at 16:10 +0900, Magnus Damm wrote:
> > [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
>
> I've been playing a bit more and found a problem.
>
> You define a per CPU variable crash_notes and on crash you loop over
> NR_CPUS and clear the notes for CPUS which don't exist. Unfortunately
> the percpu regions for CPUs which aren't physically present is returned
> to the heap on boot (see percpu_free_unused_areas) -- this means that
> you zero out heap pages on crash :-(
Ouch. Let's not do that then. =) I wondered why the data areas were
something else than just zero...
We need unique data areas for each cpu, online or not, and these areas
should be zero if the cpus don't exist. This is because we export the
machine address and size of each note to dom0 which in turn exports
the ranges through /proc/iomem to user space.
In user space the kexec-tool then builds an elf header which points
out where the notes are located in machine address space (using
/proc/iomem). This header is then passed on to the secondary crash
kernel which (for some reason) compacts all per-cpu PT_NOTE program
headers into one which will be present in the final vmcore image. At
this compacting stage we need to have data present for _all_ cpus, and
the data for cpus that are not present should contain just zeros.
This scheme should work for cpu hotplug as well.
> You need to use num_{possible,present,online}_cpus() in
> machine_crash_kexec() and kexec_get_cpu() instead of NR_CPUS.
But wouldn't that leave us with machine address in /proc/iomem that
point out heap data instead of notes? I think that using and array in
bss is a simple and good solution:
crash_note_t crash_notes[NR_CPUS];
How does that sound? WIth a comment why we are not using per-cpu data of course.
Thanks for reviewing!
/ magnus
^ permalink raw reply [flat|nested] 30+ messages in thread* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-29 4:30 ` Magnus Damm
@ 2006-11-29 7:54 ` Keir Fraser
2006-11-29 8:17 ` Magnus Damm
0 siblings, 1 reply; 30+ messages in thread
From: Keir Fraser @ 2006-11-29 7:54 UTC (permalink / raw)
To: Magnus Damm, Ian Campbell
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, Isaku Yamahata,
Magnus Damm, Horms
On 29/11/06 4:30 am, "Magnus Damm" <magnus.damm@gmail.com> wrote:
> In user space the kexec-tool then builds an elf header which points
> out where the notes are located in machine address space (using
> /proc/iomem). This header is then passed on to the secondary crash
> kernel which (for some reason) compacts all per-cpu PT_NOTE program
> headers into one which will be present in the final vmcore image. At
> this compacting stage we need to have data present for _all_ cpus, and
> the data for cpus that are not present should contain just zeros.
How does kexec discover how many cpus is 'all cpus'?
-- Keir
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-29 7:54 ` Keir Fraser
@ 2006-11-29 8:17 ` Magnus Damm
2006-11-29 9:35 ` Ian Campbell
0 siblings, 1 reply; 30+ messages in thread
From: Magnus Damm @ 2006-11-29 8:17 UTC (permalink / raw)
To: Keir Fraser
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Ian Campbell, Akio Takebe,
Magnus Damm, Isaku Yamahata, Horms
On Wed, 2006-11-29 at 07:54 +0000, Keir Fraser wrote:
> On 29/11/06 4:30 am, "Magnus Damm" <magnus.damm@gmail.com> wrote:
> > In user space the kexec-tool then builds an elf header which points
> > out where the notes are located in machine address space (using
> > /proc/iomem). This header is then passed on to the secondary crash
> > kernel which (for some reason) compacts all per-cpu PT_NOTE program
> > headers into one which will be present in the final vmcore image. At
> > this compacting stage we need to have data present for _all_ cpus, and
> > the data for cpus that are not present should contain just zeros.
>
> How does kexec discover how many cpus is 'all cpus'?
The kexec tool creates (at load time) one PT_NOTE program header per
note exported through /proc/iomem. The number of PT_NOTE program headers
is the same as the NR_CPUS constant in the hypervisor.
The seconday kernel is not aware of how many cpus that was used by the
first kernel, it only expects to find valid data pointed out by the
PT_NOTE program headers. Valid data in this case is either ELF note
formatted data or zeroed memory which happens to be valid end-of-notes
ELF note data as well.
So giving the secondary kernel a pointer to heap does not work very
well. =)
/ magnus
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-29 8:17 ` Magnus Damm
@ 2006-11-29 9:35 ` Ian Campbell
2006-11-29 11:13 ` Magnus Damm
0 siblings, 1 reply; 30+ messages in thread
From: Ian Campbell @ 2006-11-29 9:35 UTC (permalink / raw)
To: Magnus Damm
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, Magnus Damm,
Isaku Yamahata, Horms
On Wed, 2006-11-29 at 17:17 +0900, Magnus Damm wrote:
> On Wed, 2006-11-29 at 07:54 +0000, Keir Fraser wrote:
> > On 29/11/06 4:30 am, "Magnus Damm" <magnus.damm@gmail.com> wrote:
> > > In user space the kexec-tool then builds an elf header which points
> > > out where the notes are located in machine address space (using
> > > /proc/iomem). This header is then passed on to the secondary crash
> > > kernel which (for some reason) compacts all per-cpu PT_NOTE program
> > > headers into one which will be present in the final vmcore image. At
> > > this compacting stage we need to have data present for _all_ cpus, and
> > > the data for cpus that are not present should contain just zeros.
> >
> > How does kexec discover how many cpus is 'all cpus'?
>
> The kexec tool creates (at load time) one PT_NOTE program header per
> note exported through /proc/iomem. The number of PT_NOTE program headers
> is the same as the NR_CPUS constant in the hypervisor.
The guest kernel creates entries in /proc/iomem by calling
kexec_get_cpu(cpu) until it returns EINVAL. This currently happens when
cpu>NR_CPUS.
I think this function should return EINVAL for cpu>num_present_cpus()
instead. Xen doesn't currently do PCPU hotplug and this wouldn't be the
only thing that would need fixing if it ever does (percpu data would be
another one I think ;-)).
This would cause the tools to create notes only for CPUs which really
exist. That would make the loop in machine_crash_kexec() unnecessary.
Cheers,
Ian.
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-29 9:35 ` Ian Campbell
@ 2006-11-29 11:13 ` Magnus Damm
2006-11-29 11:42 ` Ian Campbell
0 siblings, 1 reply; 30+ messages in thread
From: Magnus Damm @ 2006-11-29 11:13 UTC (permalink / raw)
To: Ian Campbell
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, Isaku Yamahata,
Magnus Damm, Horms
On 11/29/06, Ian Campbell <Ian.Campbell@xensource.com> wrote:
> On Wed, 2006-11-29 at 17:17 +0900, Magnus Damm wrote:
> >
> > The kexec tool creates (at load time) one PT_NOTE program header per
> > note exported through /proc/iomem. The number of PT_NOTE program headers
> > is the same as the NR_CPUS constant in the hypervisor.
>
> The guest kernel creates entries in /proc/iomem by calling
> kexec_get_cpu(cpu) until it returns EINVAL. This currently happens when
> cpu>NR_CPUS.
>
> I think this function should return EINVAL for cpu>num_present_cpus()
> instead. Xen doesn't currently do PCPU hotplug and this wouldn't be the
> only thing that would need fixing if it ever does (percpu data would be
> another one I think ;-)).
>
> This would cause the tools to create notes only for CPUs which really
> exist. That would make the loop in machine_crash_kexec() unnecessary.
I feel that using bss instead of per-cpu data is more robust and will
make future cpu hotplug support a breeze to implement - at least in
the case of kexec. Using bss also makes the loop in
machine_crash_kexec() unnecessary.
Using num_present_cpus() will of course work as well, but I'd like to
avoid adding code that likely needs to be rewritten in the near
future.. But you know the future better than I do, so what do you
think? =)
Thanks!
/ magnus
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [PATCH 00/04] Kexec / Kdump: Release 20061122 (xen-unstable-12502)
2006-11-29 11:13 ` Magnus Damm
@ 2006-11-29 11:42 ` Ian Campbell
0 siblings, 0 replies; 30+ messages in thread
From: Ian Campbell @ 2006-11-29 11:42 UTC (permalink / raw)
To: Magnus Damm
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, Isaku Yamahata,
Magnus Damm, Horms
On Wed, 2006-11-29 at 20:13 +0900, Magnus Damm wrote:
> On 11/29/06, Ian Campbell <Ian.Campbell@xensource.com> wrote:
> > On Wed, 2006-11-29 at 17:17 +0900, Magnus Damm wrote:
> > >
> > > The kexec tool creates (at load time) one PT_NOTE program header per
> > > note exported through /proc/iomem. The number of PT_NOTE program headers
> > > is the same as the NR_CPUS constant in the hypervisor.
> >
> > The guest kernel creates entries in /proc/iomem by calling
> > kexec_get_cpu(cpu) until it returns EINVAL. This currently happens when
> > cpu>NR_CPUS.
> >
> > I think this function should return EINVAL for cpu>num_present_cpus()
> > instead. Xen doesn't currently do PCPU hotplug and this wouldn't be the
> > only thing that would need fixing if it ever does (percpu data would be
> > another one I think ;-)).
> >
> > This would cause the tools to create notes only for CPUs which really
> > exist. That would make the loop in machine_crash_kexec() unnecessary.
>
> I feel that using bss instead of per-cpu data is more robust and will
> make future cpu hotplug support a breeze to implement - at least in
> the case of kexec. Using bss also makes the loop in
> machine_crash_kexec() unnecessary.
>
> Using num_present_cpus() will of course work as well, but I'd like to
> avoid adding code that likely needs to be rewritten in the near
> future.. But you know the future better than I do, so what do you
> think? =)
I don't think anyone is planning PCPU hotplug anytime soon (I've not
even heard rumours about the distant future ;-)). I do think that using
the existing infrastructure is the right way to go though rather than
open coding a different per cpu mechanism to solve a problem which
doesn't currently exist. If someone implements PCPU hotplug they will no
doubt need to update the per cpu infrastructure and if kdump is using it
then it can be taken into consideration at that time.
Cheers,
Ian.
^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH 04/04] Kexec / Kdump: x86_64 specific code
2006-11-30 7:58 [PATCH 00/04] Kexec / Kdump: Release 20061130 (xen-unstable-12621) Magnus Damm
@ 2006-11-30 7:58 ` Magnus Damm
0 siblings, 0 replies; 30+ messages in thread
From: Magnus Damm @ 2006-11-30 7:58 UTC (permalink / raw)
To: Ian Campbell
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, magnus.damm,
Isaku Yamahata, Magnus Damm, Horms, Dave Anderson, Itsuro ODA
[PATCH 04/04] Kexec / Kdump: x86_64 specific code
This patch contains the x86_64 implementation of Kexec / Kdump for Xen.
Signed-Off-By: Magnus Damm <magnus@valinux.co.jp>
Signed-Off-By: Simon Horman <horms@verge.net.au>
---
Applies on top of xen-unstable-12621.
buildconfigs/linux-defconfig_xen_x86_64 | 1
linux-2.6-xen-sparse/arch/x86_64/Kconfig | 2
linux-2.6-xen-sparse/arch/x86_64/kernel/e820-xen.c | 6
linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c | 13
linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h | 7
linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h | 2
patches/linux-2.6.16.33/git-4b...1f.patch | 375 ++++
patches/linux-2.6.16.33/linux-2.6.19-rc1-kexec...code-x86_64.patch | 161 ++
patches/linux-2.6.16.33/linux-2.6.19-rc1-kexec-xen-x86_64.patch | 219 +++
patches/linux-2.6.16.33/series | 3
xen/arch/x86/x86_64/entry.S | 2
xen/include/asm-x86/x86_64/elf.h | 72 +
xen/include/asm-x86/x86_64/kexec.h | 23
13 files changed, 877 insertions(+), 9 deletions(-)
--- 0002/buildconfigs/linux-defconfig_xen_x86_64
+++ work/buildconfigs/linux-defconfig_xen_x86_64
@@ -138,6 +138,7 @@ CONFIG_SWIOTLB=y
CONFIG_PHYSICAL_START=0x100000
CONFIG_SECCOMP=y
CONFIG_HZ_100=y
+CONFIG_KEXEC=y
# CONFIG_HZ_250 is not set
# CONFIG_HZ_1000 is not set
CONFIG_HZ=100
--- 0001/linux-2.6-xen-sparse/arch/x86_64/Kconfig
+++ work/linux-2.6-xen-sparse/arch/x86_64/Kconfig
@@ -435,7 +435,7 @@ config X86_MCE_AMD
config KEXEC
bool "kexec system call (EXPERIMENTAL)"
- depends on EXPERIMENTAL && !X86_64_XEN
+ depends on EXPERIMENTAL && !XEN_UNPRIVILEGED_GUEST
help
kexec is a system call that implements the ability to shutdown your
current kernel, and to start another kernel. It is like a reboot
--- 0001/linux-2.6-xen-sparse/arch/x86_64/kernel/e820-xen.c
+++ work/linux-2.6-xen-sparse/arch/x86_64/kernel/e820-xen.c
@@ -260,7 +260,11 @@ void __init e820_reserve_resources(struc
request_resource(res, &data_resource);
#endif
#ifdef CONFIG_KEXEC
- request_resource(res, &crashk_res);
+ if (crashk_res.start != crashk_res.end)
+ request_resource(res, &crashk_res);
+#ifdef CONFIG_XEN
+ xen_machine_kexec_register_resources(res);
+#endif
#endif
}
}
--- 0001/linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c
+++ work/linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c
@@ -80,6 +80,10 @@
#include <asm/mach-xen/setup_arch_post.h>
#include <xen/interface/memory.h>
+#ifdef CONFIG_XEN
+#include <xen/interface/kexec.h>
+#endif
+
extern unsigned long start_pfn;
extern struct edid_info edid_info;
@@ -450,6 +454,7 @@ static __init void parse_cmdline_early (
* after a kernel panic.
*/
else if (!memcmp(from, "crashkernel=", 12)) {
+#ifndef CONFIG_XEN
unsigned long size, base;
size = memparse(from+12, &from);
if (*from == '@') {
@@ -460,6 +465,10 @@ static __init void parse_cmdline_early (
crashk_res.start = base;
crashk_res.end = base + size - 1;
}
+#else
+ printk("Ignoring crashkernel command line, "
+ "parameter will be supplied by xen\n");
+#endif
}
#endif
@@ -812,11 +821,15 @@ void __init setup_arch(char **cmdline_p)
#endif
#endif /* !CONFIG_XEN */
#ifdef CONFIG_KEXEC
+#ifdef CONFIG_XEN
+ xen_machine_kexec_setup_resources();
+#else
if (crashk_res.start != crashk_res.end) {
reserve_bootmem(crashk_res.start,
crashk_res.end - crashk_res.start + 1);
}
#endif
+#endif
paging_init();
#ifdef CONFIG_X86_LOCAL_APIC
--- 0001/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h
+++ work/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h
@@ -396,4 +396,11 @@ HYPERVISOR_xenoprof_op(
return _hypercall2(int, xenoprof_op, op, arg);
}
+static inline int
+HYPERVISOR_kexec_op(
+ unsigned long op, void *args)
+{
+ return _hypercall2(int, kexec_op, op, args);
+}
+
#endif /* __HYPERCALL_H__ */
--- 0001/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h
+++ work/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h
@@ -90,6 +90,8 @@ extern unsigned long profile_pc(struct p
#define profile_pc(regs) instruction_pointer(regs)
#endif
+#include <linux/compiler.h>
+
void signal_fault(struct pt_regs *regs, void __user *frame, char *where);
struct task_struct;
--- /dev/null
+++ work/patches/linux-2.6.16.33/git-4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f.patch
@@ -0,0 +1,375 @@
+From: Magnus Damm <magnus@valinux.co.jp>
+Date: Tue, 26 Sep 2006 08:52:38 +0000 (+0200)
+Subject: [PATCH] Avoid overwriting the current pgd (V4, x86_64)
+X-Git-Tag: v2.6.19-rc1
+X-Git-Url: http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f
+
+[PATCH] Avoid overwriting the current pgd (V4, x86_64)
+
+kexec: Avoid overwriting the current pgd (V4, x86_64)
+
+This patch upgrades the x86_64-specific kexec code to avoid overwriting the
+current pgd. Overwriting the current pgd is bad when CONFIG_CRASH_DUMP is used
+to start a secondary kernel that dumps the memory of the previous kernel.
+
+The code introduces a new set of page tables. These tables are used to provide
+an executable identity mapping without overwriting the current pgd.
+
+Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
+Signed-off-by: Andi Kleen <ak@suse.de>
+---
+
+--- a/arch/x86_64/kernel/machine_kexec.c
++++ b/arch/x86_64/kernel/machine_kexec.c
+@@ -15,6 +15,15 @@
+ #include <asm/mmu_context.h>
+ #include <asm/io.h>
+
++#define PAGE_ALIGNED __attribute__ ((__aligned__(PAGE_SIZE)))
++static u64 kexec_pgd[512] PAGE_ALIGNED;
++static u64 kexec_pud0[512] PAGE_ALIGNED;
++static u64 kexec_pmd0[512] PAGE_ALIGNED;
++static u64 kexec_pte0[512] PAGE_ALIGNED;
++static u64 kexec_pud1[512] PAGE_ALIGNED;
++static u64 kexec_pmd1[512] PAGE_ALIGNED;
++static u64 kexec_pte1[512] PAGE_ALIGNED;
++
+ static void init_level2_page(pmd_t *level2p, unsigned long addr)
+ {
+ unsigned long end_addr;
+@@ -144,32 +153,19 @@ static void load_segments(void)
+ );
+ }
+
+-typedef NORET_TYPE void (*relocate_new_kernel_t)(unsigned long indirection_page,
+- unsigned long control_code_buffer,
+- unsigned long start_address,
+- unsigned long pgtable) ATTRIB_NORET;
+-
+-extern const unsigned char relocate_new_kernel[];
+-extern const unsigned long relocate_new_kernel_size;
+-
+ int machine_kexec_prepare(struct kimage *image)
+ {
+- unsigned long start_pgtable, control_code_buffer;
++ unsigned long start_pgtable;
+ int result;
+
+ /* Calculate the offsets */
+ start_pgtable = page_to_pfn(image->control_code_page) << PAGE_SHIFT;
+- control_code_buffer = start_pgtable + PAGE_SIZE;
+
+ /* Setup the identity mapped 64bit page table */
+ result = init_pgtable(image, start_pgtable);
+ if (result)
+ return result;
+
+- /* Place the code in the reboot code buffer */
+- memcpy(__va(control_code_buffer), relocate_new_kernel,
+- relocate_new_kernel_size);
+-
+ return 0;
+ }
+
+@@ -184,28 +180,34 @@ void machine_kexec_cleanup(struct kimage
+ */
+ NORET_TYPE void machine_kexec(struct kimage *image)
+ {
+- unsigned long page_list;
+- unsigned long control_code_buffer;
+- unsigned long start_pgtable;
+- relocate_new_kernel_t rnk;
++ unsigned long page_list[PAGES_NR];
++ void *control_page;
+
+ /* Interrupts aren't acceptable while we reboot */
+ local_irq_disable();
+
+- /* Calculate the offsets */
+- page_list = image->head;
+- start_pgtable = page_to_pfn(image->control_code_page) << PAGE_SHIFT;
+- control_code_buffer = start_pgtable + PAGE_SIZE;
++ control_page = page_address(image->control_code_page) + PAGE_SIZE;
++ memcpy(control_page, relocate_kernel, PAGE_SIZE);
+
+- /* Set the low half of the page table to my identity mapped
+- * page table for kexec. Leave the high half pointing at the
+- * kernel pages. Don't bother to flush the global pages
+- * as that will happen when I fully switch to my identity mapped
+- * page table anyway.
+- */
+- memcpy(__va(read_cr3()), __va(start_pgtable), PAGE_SIZE/2);
+- __flush_tlb();
++ page_list[PA_CONTROL_PAGE] = __pa(control_page);
++ page_list[VA_CONTROL_PAGE] = (unsigned long)relocate_kernel;
++ page_list[PA_PGD] = __pa(kexec_pgd);
++ page_list[VA_PGD] = (unsigned long)kexec_pgd;
++ page_list[PA_PUD_0] = __pa(kexec_pud0);
++ page_list[VA_PUD_0] = (unsigned long)kexec_pud0;
++ page_list[PA_PMD_0] = __pa(kexec_pmd0);
++ page_list[VA_PMD_0] = (unsigned long)kexec_pmd0;
++ page_list[PA_PTE_0] = __pa(kexec_pte0);
++ page_list[VA_PTE_0] = (unsigned long)kexec_pte0;
++ page_list[PA_PUD_1] = __pa(kexec_pud1);
++ page_list[VA_PUD_1] = (unsigned long)kexec_pud1;
++ page_list[PA_PMD_1] = __pa(kexec_pmd1);
++ page_list[VA_PMD_1] = (unsigned long)kexec_pmd1;
++ page_list[PA_PTE_1] = __pa(kexec_pte1);
++ page_list[VA_PTE_1] = (unsigned long)kexec_pte1;
+
++ page_list[PA_TABLE_PAGE] =
++ (unsigned long)__pa(page_address(image->control_code_page));
+
+ /* The segment registers are funny things, they have both a
+ * visible and an invisible part. Whenever the visible part is
+@@ -222,9 +224,10 @@ NORET_TYPE void machine_kexec(struct kim
+ */
+ set_gdt(phys_to_virt(0),0);
+ set_idt(phys_to_virt(0),0);
++
+ /* now call it */
+- rnk = (relocate_new_kernel_t) control_code_buffer;
+- (*rnk)(page_list, control_code_buffer, image->start, start_pgtable);
++ relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
++ image->start);
+ }
+
+ /* crashkernel=size@addr specifies the location to reserve for
+--- a/arch/x86_64/kernel/relocate_kernel.S
++++ b/arch/x86_64/kernel/relocate_kernel.S
+@@ -7,31 +7,169 @@
+ */
+
+ #include <linux/linkage.h>
++#include <asm/page.h>
++#include <asm/kexec.h>
+
+- /*
+- * Must be relocatable PIC code callable as a C function, that once
+- * it starts can not use the previous processes stack.
+- */
+- .globl relocate_new_kernel
++/*
++ * Must be relocatable PIC code callable as a C function
++ */
++
++#define PTR(x) (x << 3)
++#define PAGE_ALIGNED (1 << PAGE_SHIFT)
++#define PAGE_ATTR 0x63 /* _PAGE_PRESENT|_PAGE_RW|_PAGE_ACCESSED|_PAGE_DIRTY */
++
++ .text
++ .align PAGE_ALIGNED
+ .code64
++ .globl relocate_kernel
++relocate_kernel:
++ /* %rdi indirection_page
++ * %rsi page_list
++ * %rdx start address
++ */
++
++ /* map the control page at its virtual address */
++
++ movq $0x0000ff8000000000, %r10 /* mask */
++ mov $(39 - 3), %cl /* bits to shift */
++ movq PTR(VA_CONTROL_PAGE)(%rsi), %r11 /* address to map */
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PGD)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PUD_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PUD_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PMD_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PMD_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PTE_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PTE_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ /* identity map the control page at its physical address */
++
++ movq $0x0000ff8000000000, %r10 /* mask */
++ mov $(39 - 3), %cl /* bits to shift */
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r11 /* address to map */
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PGD)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PUD_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PUD_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PMD_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PMD_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PTE_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PTE_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
+ relocate_new_kernel:
+- /* %rdi page_list
+- * %rsi reboot_code_buffer
++ /* %rdi indirection_page
++ * %rsi page_list
+ * %rdx start address
+- * %rcx page_table
+- * %r8 arg5
+- * %r9 arg6
+ */
+
+ /* zero out flags, and disable interrupts */
+ pushq $0
+ popfq
+
+- /* set a new stack at the bottom of our page... */
+- lea 4096(%rsi), %rsp
++ /* get physical address of control page now */
++ /* this is impossible after page table switch */
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++
++ /* get physical address of page table now too */
++ movq PTR(PA_TABLE_PAGE)(%rsi), %rcx
++
++ /* switch to new set of page tables */
++ movq PTR(PA_PGD)(%rsi), %r9
++ movq %r9, %cr3
++
++ /* setup a new stack at the end of the physical control page */
++ lea 4096(%r8), %rsp
++
++ /* jump to identity mapped page */
++ addq $(identity_mapped - relocate_kernel), %r8
++ pushq %r8
++ ret
+
+- /* store the parameters back on the stack */
+- pushq %rdx /* store the start address */
++identity_mapped:
++ /* store the start address on the stack */
++ pushq %rdx
+
+ /* Set cr0 to a known state:
+ * 31 1 == Paging enabled
+@@ -136,8 +274,3 @@ relocate_new_kernel:
+ xorq %r15, %r15
+
+ ret
+-relocate_new_kernel_end:
+-
+- .globl relocate_new_kernel_size
+-relocate_new_kernel_size:
+- .quad relocate_new_kernel_end - relocate_new_kernel
+--- a/include/asm-x86_64/kexec.h
++++ b/include/asm-x86_64/kexec.h
+@@ -1,6 +1,27 @@
+ #ifndef _X86_64_KEXEC_H
+ #define _X86_64_KEXEC_H
+
++#define PA_CONTROL_PAGE 0
++#define VA_CONTROL_PAGE 1
++#define PA_PGD 2
++#define VA_PGD 3
++#define PA_PUD_0 4
++#define VA_PUD_0 5
++#define PA_PMD_0 6
++#define VA_PMD_0 7
++#define PA_PTE_0 8
++#define VA_PTE_0 9
++#define PA_PUD_1 10
++#define VA_PUD_1 11
++#define PA_PMD_1 12
++#define VA_PMD_1 13
++#define PA_PTE_1 14
++#define VA_PTE_1 15
++#define PA_TABLE_PAGE 16
++#define PAGES_NR 17
++
++#ifndef __ASSEMBLY__
++
+ #include <linux/string.h>
+
+ #include <asm/page.h>
+@@ -64,4 +85,12 @@ static inline void crash_setup_regs(stru
+ newregs->rip = (unsigned long)current_text_addr();
+ }
+ }
++
++NORET_TYPE void
++relocate_kernel(unsigned long indirection_page,
++ unsigned long page_list,
++ unsigned long start_address) ATTRIB_NORET;
++
++#endif /* __ASSEMBLY__ */
++
+ #endif /* _X86_64_KEXEC_H */
--- /dev/null
+++ work/patches/linux-2.6.16.33/linux-2.6.19-rc1-kexec-move_segment_code-x86_64.patch
@@ -0,0 +1,161 @@
+kexec: Move asm segment handling code to the assembly file (x86_64)
+
+This patch moves the idt, gdt, and segment handling code from machine_kexec.c
+to relocate_kernel.S. The main reason behind this move is to avoid code
+duplication in the Xen hypervisor. With this patch all code required to kexec
+is put on the control page.
+
+On top of that this patch also counts as a cleanup - I think it is much
+nicer to write assembly directly in assembly files than wrap inline assembly
+in C functions for no apparent reason.
+
+Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
+---
+
+ Applies to 2.6.19-rc1.
+
+ machine_kexec.c | 58 -----------------------------------------------------
+ relocate_kernel.S | 50 +++++++++++++++++++++++++++++++++++++++++----
+ 2 files changed, 45 insertions(+), 63 deletions(-)
+
+--- 0002/arch/x86_64/kernel/machine_kexec.c
++++ work/arch/x86_64/kernel/machine_kexec.c 2006-10-05 16:15:49.000000000 +0900
+@@ -112,47 +112,6 @@ static int init_pgtable(struct kimage *i
+ return init_level4_page(image, level4p, 0, end_pfn << PAGE_SHIFT);
+ }
+
+-static void set_idt(void *newidt, u16 limit)
+-{
+- struct desc_ptr curidt;
+-
+- /* x86-64 supports unaliged loads & stores */
+- curidt.size = limit;
+- curidt.address = (unsigned long)newidt;
+-
+- __asm__ __volatile__ (
+- "lidtq %0\n"
+- : : "m" (curidt)
+- );
+-};
+-
+-
+-static void set_gdt(void *newgdt, u16 limit)
+-{
+- struct desc_ptr curgdt;
+-
+- /* x86-64 supports unaligned loads & stores */
+- curgdt.size = limit;
+- curgdt.address = (unsigned long)newgdt;
+-
+- __asm__ __volatile__ (
+- "lgdtq %0\n"
+- : : "m" (curgdt)
+- );
+-};
+-
+-static void load_segments(void)
+-{
+- __asm__ __volatile__ (
+- "\tmovl %0,%%ds\n"
+- "\tmovl %0,%%es\n"
+- "\tmovl %0,%%ss\n"
+- "\tmovl %0,%%fs\n"
+- "\tmovl %0,%%gs\n"
+- : : "a" (__KERNEL_DS) : "memory"
+- );
+-}
+-
+ int machine_kexec_prepare(struct kimage *image)
+ {
+ unsigned long start_pgtable;
+@@ -209,23 +168,6 @@ NORET_TYPE void machine_kexec(struct kim
+ page_list[PA_TABLE_PAGE] =
+ (unsigned long)__pa(page_address(image->control_code_page));
+
+- /* The segment registers are funny things, they have both a
+- * visible and an invisible part. Whenever the visible part is
+- * set to a specific selector, the invisible part is loaded
+- * with from a table in memory. At no other time is the
+- * descriptor table in memory accessed.
+- *
+- * I take advantage of this here by force loading the
+- * segments, before I zap the gdt with an invalid value.
+- */
+- load_segments();
+- /* The gdt & idt are now invalid.
+- * If you want to load them you must set up your own idt & gdt.
+- */
+- set_gdt(phys_to_virt(0),0);
+- set_idt(phys_to_virt(0),0);
+-
+- /* now call it */
+ relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
+ image->start);
+ }
+--- 0002/arch/x86_64/kernel/relocate_kernel.S
++++ work/arch/x86_64/kernel/relocate_kernel.S 2006-10-05 16:18:07.000000000 +0900
+@@ -159,13 +159,39 @@ relocate_new_kernel:
+ movq PTR(PA_PGD)(%rsi), %r9
+ movq %r9, %cr3
+
++ /* setup idt */
++ movq %r8, %rax
++ addq $(idt_80 - relocate_kernel), %rax
++ lidtq (%rax)
++
++ /* setup gdt */
++ movq %r8, %rax
++ addq $(gdt - relocate_kernel), %rax
++ movq %r8, %r9
++ addq $((gdt_80 - relocate_kernel) + 2), %r9
++ movq %rax, (%r9)
++
++ movq %r8, %rax
++ addq $(gdt_80 - relocate_kernel), %rax
++ lgdtq (%rax)
++
++ /* setup data segment registers */
++ xorl %eax, %eax
++ movl %eax, %ds
++ movl %eax, %es
++ movl %eax, %fs
++ movl %eax, %gs
++ movl %eax, %ss
++
+ /* setup a new stack at the end of the physical control page */
+ lea 4096(%r8), %rsp
+
+- /* jump to identity mapped page */
+- addq $(identity_mapped - relocate_kernel), %r8
+- pushq %r8
+- ret
++ /* load new code segment and jump to identity mapped page */
++ movq %r8, %rax
++ addq $(identity_mapped - relocate_kernel), %rax
++ pushq $(gdt_cs - gdt)
++ pushq %rax
++ lretq
+
+ identity_mapped:
+ /* store the start address on the stack */
+@@ -272,5 +298,19 @@ identity_mapped:
+ xorq %r13, %r13
+ xorq %r14, %r14
+ xorq %r15, %r15
+-
+ ret
++
++ .align 16
++gdt:
++ .quad 0x0000000000000000 /* NULL descriptor */
++gdt_cs:
++ .quad 0x00af9a000000ffff
++gdt_end:
++
++gdt_80:
++ .word gdt_end - gdt - 1 /* limit */
++ .quad 0 /* base - filled in by code above */
++
++idt_80:
++ .word 0 /* limit */
++ .quad 0 /* base */
--- /dev/null
+++ work/patches/linux-2.6.16.33/linux-2.6.19-rc1-kexec-xen-x86_64.patch
@@ -0,0 +1,219 @@
+--- 0001/arch/x86_64/kernel/crash.c
++++ work/arch/x86_64/kernel/crash.c
+@@ -92,6 +92,7 @@ static void crash_save_self(struct pt_re
+ crash_save_this_cpu(regs, cpu);
+ }
+
++#ifndef CONFIG_XEN
+ #ifdef CONFIG_SMP
+ static atomic_t waiting_for_crash_ipi;
+
+@@ -156,6 +157,7 @@ static void nmi_shootdown_cpus(void)
+ /* There are no cpus to shootdown */
+ }
+ #endif
++#endif /* CONFIG_XEN */
+
+ void machine_crash_shutdown(struct pt_regs *regs)
+ {
+@@ -173,6 +175,8 @@ void machine_crash_shutdown(struct pt_re
+
+ /* Make a note of crashing cpu. Will be used in NMI callback.*/
+ crashing_cpu = smp_processor_id();
++
++#ifndef CONFIG_XEN
+ nmi_shootdown_cpus();
+
+ if(cpu_has_apic)
+@@ -181,6 +185,6 @@ void machine_crash_shutdown(struct pt_re
+ #if defined(CONFIG_X86_IO_APIC)
+ disable_IO_APIC();
+ #endif
+-
++#endif /* CONFIG_XEN */
+ crash_save_self(regs);
+ }
+--- 0010/arch/x86_64/kernel/machine_kexec.c
++++ work/arch/x86_64/kernel/machine_kexec.c
+@@ -24,6 +24,104 @@ static u64 kexec_pud1[512] PAGE_ALIGNED;
+ static u64 kexec_pmd1[512] PAGE_ALIGNED;
+ static u64 kexec_pte1[512] PAGE_ALIGNED;
+
++#ifdef CONFIG_XEN
++
++/* In the case of Xen, override hypervisor functions to be able to create
++ * a regular identity mapping page table...
++ */
++
++#include <xen/interface/kexec.h>
++#include <xen/interface/memory.h>
++
++#define x__pmd(x) ((pmd_t) { (x) } )
++#define x__pud(x) ((pud_t) { (x) } )
++#define x__pgd(x) ((pgd_t) { (x) } )
++
++#define x_pmd_val(x) ((x).pmd)
++#define x_pud_val(x) ((x).pud)
++#define x_pgd_val(x) ((x).pgd)
++
++static inline void x_set_pmd(pmd_t *dst, pmd_t val)
++{
++ x_pmd_val(*dst) = x_pmd_val(val);
++}
++
++static inline void x_set_pud(pud_t *dst, pud_t val)
++{
++ x_pud_val(*dst) = phys_to_machine(x_pud_val(val));
++}
++
++static inline void x_pud_clear (pud_t *pud)
++{
++ x_pud_val(*pud) = 0;
++}
++
++static inline void x_set_pgd(pgd_t *dst, pgd_t val)
++{
++ x_pgd_val(*dst) = phys_to_machine(x_pgd_val(val));
++}
++
++static inline void x_pgd_clear (pgd_t * pgd)
++{
++ x_pgd_val(*pgd) = 0;
++}
++
++#define X__PAGE_KERNEL_LARGE_EXEC \
++ _PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_PSE
++#define X_KERNPG_TABLE _PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY
++
++#define __ma(x) (pfn_to_mfn(__pa((x)) >> PAGE_SHIFT) << PAGE_SHIFT)
++
++#if PAGES_NR > KEXEC_XEN_NO_PAGES
++#error PAGES_NR is greater than KEXEC_XEN_NO_PAGES - Xen support will break
++#endif
++
++#if PA_CONTROL_PAGE != 0
++#error PA_CONTROL_PAGE is non zero - Xen support will break
++#endif
++
++void machine_kexec_setup_load_arg(xen_kexec_image_t *xki, struct kimage *image)
++{
++ void *control_page;
++ void *table_page;
++
++ memset(xki->page_list, 0, sizeof(xki->page_list));
++
++ control_page = page_address(image->control_code_page) + PAGE_SIZE;
++ memcpy(control_page, relocate_kernel, PAGE_SIZE);
++
++ table_page = page_address(image->control_code_page);
++
++ xki->page_list[PA_CONTROL_PAGE] = __ma(control_page);
++ xki->page_list[PA_TABLE_PAGE] = __ma(table_page);
++
++ xki->page_list[PA_PGD] = __ma(kexec_pgd);
++ xki->page_list[PA_PUD_0] = __ma(kexec_pud0);
++ xki->page_list[PA_PUD_1] = __ma(kexec_pud1);
++ xki->page_list[PA_PMD_0] = __ma(kexec_pmd0);
++ xki->page_list[PA_PMD_1] = __ma(kexec_pmd1);
++ xki->page_list[PA_PTE_0] = __ma(kexec_pte0);
++ xki->page_list[PA_PTE_1] = __ma(kexec_pte1);
++}
++
++#else /* CONFIG_XEN */
++
++#define x__pmd(x) __pmd(x)
++#define x__pud(x) __pud(x)
++#define x__pgd(x) __pgd(x)
++
++#define x_set_pmd(x, y) set_pmd(x, y)
++#define x_set_pud(x, y) set_pud(x, y)
++#define x_set_pgd(x, y) set_pgd(x, y)
++
++#define x_pud_clear(x) pud_clear(x)
++#define x_pgd_clear(x) pgd_clear(x)
++
++#define X__PAGE_KERNEL_LARGE_EXEC __PAGE_KERNEL_LARGE_EXEC
++#define X_KERNPG_TABLE _KERNPG_TABLE
++
++#endif /* CONFIG_XEN */
++
+ static void init_level2_page(pmd_t *level2p, unsigned long addr)
+ {
+ unsigned long end_addr;
+@@ -31,7 +129,7 @@ static void init_level2_page(pmd_t *leve
+ addr &= PAGE_MASK;
+ end_addr = addr + PUD_SIZE;
+ while (addr < end_addr) {
+- set_pmd(level2p++, __pmd(addr | __PAGE_KERNEL_LARGE_EXEC));
++ x_set_pmd(level2p++, x__pmd(addr | X__PAGE_KERNEL_LARGE_EXEC));
+ addr += PMD_SIZE;
+ }
+ }
+@@ -56,12 +154,12 @@ static int init_level3_page(struct kimag
+ }
+ level2p = (pmd_t *)page_address(page);
+ init_level2_page(level2p, addr);
+- set_pud(level3p++, __pud(__pa(level2p) | _KERNPG_TABLE));
++ x_set_pud(level3p++, x__pud(__pa(level2p) | X_KERNPG_TABLE));
+ addr += PUD_SIZE;
+ }
+ /* clear the unused entries */
+ while (addr < end_addr) {
+- pud_clear(level3p++);
++ x_pud_clear(level3p++);
+ addr += PUD_SIZE;
+ }
+ out:
+@@ -92,12 +190,12 @@ static int init_level4_page(struct kimag
+ if (result) {
+ goto out;
+ }
+- set_pgd(level4p++, __pgd(__pa(level3p) | _KERNPG_TABLE));
++ x_set_pgd(level4p++, x__pgd(__pa(level3p) | X_KERNPG_TABLE));
+ addr += PGDIR_SIZE;
+ }
+ /* clear the unused entries */
+ while (addr < end_addr) {
+- pgd_clear(level4p++);
++ x_pgd_clear(level4p++);
+ addr += PGDIR_SIZE;
+ }
+ out:
+@@ -108,8 +206,14 @@ out:
+ static int init_pgtable(struct kimage *image, unsigned long start_pgtable)
+ {
+ pgd_t *level4p;
++ unsigned long x_end_pfn = end_pfn;
++
++#ifdef CONFIG_XEN
++ x_end_pfn = HYPERVISOR_memory_op(XENMEM_maximum_ram_page, NULL);
++#endif
++
+ level4p = (pgd_t *)__va(start_pgtable);
+- return init_level4_page(image, level4p, 0, end_pfn << PAGE_SHIFT);
++ return init_level4_page(image, level4p, 0, x_end_pfn << PAGE_SHIFT);
+ }
+
+ int machine_kexec_prepare(struct kimage *image)
+--- 0009/include/asm-x86_64/kexec.h
++++ work/include/asm-x86_64/kexec.h
+@@ -91,6 +91,19 @@ relocate_kernel(unsigned long indirectio
+ unsigned long page_list,
+ unsigned long start_address) ATTRIB_NORET;
+
++/* Under Xen we need to work with machine addresses. These macros give the
++ * machine address of a certain page to the generic kexec code instead of
++ * the pseudo physical address which would be given by the default macros.
++ */
++
++#ifdef CONFIG_XEN
++#define KEXEC_ARCH_HAS_PAGE_MACROS
++#define kexec_page_to_pfn(page) pfn_to_mfn(page_to_pfn(page))
++#define kexec_pfn_to_page(pfn) pfn_to_page(mfn_to_pfn(pfn))
++#define kexec_virt_to_phys(addr) virt_to_machine(addr)
++#define kexec_phys_to_virt(addr) phys_to_virt(machine_to_phys(addr))
++#endif
++
+ #endif /* __ASSEMBLY__ */
+
+ #endif /* _X86_64_KEXEC_H */
--- 0005/patches/linux-2.6.16.33/series
+++ work/patches/linux-2.6.16.33/series
@@ -4,6 +4,9 @@ git-2a8a3d5b65e86ec1dfef7d268c64a909eab9
git-3566561bfadffcb5dbc85d576be80c0dbf2cccc9.patch
linux-2.6.19-rc1-kexec-move_segment_code-i386.patch
linux-2.6.19-rc1-kexec-xen-i386.patch
+git-4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f.patch
+linux-2.6.19-rc1-kexec-move_segment_code-x86_64.patch
+linux-2.6.19-rc1-kexec-xen-x86_64.patch
blktap-aio-16_03_06.patch
device_bind.patch
fix-hz-suspend.patch
--- 0001/xen/arch/x86/x86_64/entry.S
+++ work/xen/arch/x86/x86_64/entry.S
@@ -559,6 +559,7 @@ ENTRY(hypercall_table)
.quad do_hvm_op
.quad do_sysctl /* 35 */
.quad do_domctl
+ .quad do_kexec_op
.rept NR_hypercalls-((.-hypercall_table)/8)
.quad do_ni_hypercall
.endr
@@ -601,6 +602,7 @@ ENTRY(hypercall_args_table)
.byte 2 /* do_hvm_op */
.byte 1 /* do_sysctl */ /* 35 */
.byte 1 /* do_domctl */
+ .byte 2 /* do_kexec */
.rept NR_hypercalls-(.-hypercall_args_table)
.byte 0 /* do_ni_hypercall */
.endr
--- 0004/xen/include/asm-x86/x86_64/elf.h
+++ work/xen/include/asm-x86/x86_64/elf.h
@@ -1,16 +1,82 @@
#ifndef __X86_64_ELF_H__
#define __X86_64_ELF_H__
-#include <xen/lib.h> /* for printk() used in stub */
+#include <asm/processor.h>
typedef struct {
- unsigned long dummy;
+ unsigned long r15;
+ unsigned long r14;
+ unsigned long r13;
+ unsigned long r12;
+ unsigned long rbp;
+ unsigned long rbx;
+ unsigned long r11;
+ unsigned long r10;
+ unsigned long r9;
+ unsigned long r8;
+ unsigned long rax;
+ unsigned long rcx;
+ unsigned long rdx;
+ unsigned long rsi;
+ unsigned long rdi;
+ unsigned long orig_rax;
+ unsigned long rip;
+ unsigned long cs;
+ unsigned long eflags;
+ unsigned long rsp;
+ unsigned long ss;
+ unsigned long thread_fs;
+ unsigned long thread_gs;
+ unsigned long ds;
+ unsigned long es;
+ unsigned long fs;
+ unsigned long gs;
} ELF_Gregset;
extern inline void elf_core_save_regs(ELF_Gregset *core_regs,
crash_xen_core_t *xen_core_regs)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ unsigned long tmp;
+
+ asm volatile("movq %%r15,%0" : "=m"(core_regs->r15));
+ asm volatile("movq %%r14,%0" : "=m"(core_regs->r14));
+ asm volatile("movq %%r13,%0" : "=m"(core_regs->r13));
+ asm volatile("movq %%r12,%0" : "=m"(core_regs->r12));
+ asm volatile("movq %%rbp,%0" : "=m"(core_regs->rbp));
+ asm volatile("movq %%rbx,%0" : "=m"(core_regs->rbx));
+ asm volatile("movq %%r11,%0" : "=m"(core_regs->r11));
+ asm volatile("movq %%r10,%0" : "=m"(core_regs->r10));
+ asm volatile("movq %%r9,%0" : "=m"(core_regs->r9));
+ asm volatile("movq %%r8,%0" : "=m"(core_regs->r8));
+ asm volatile("movq %%rax,%0" : "=m"(core_regs->rax));
+ asm volatile("movq %%rcx,%0" : "=m"(core_regs->rcx));
+ asm volatile("movq %%rdx,%0" : "=m"(core_regs->rdx));
+ asm volatile("movq %%rsi,%0" : "=m"(core_regs->rsi));
+ asm volatile("movq %%rdi,%0" : "=m"(core_regs->rdi));
+ /* orig_rax not filled in for now */
+ core_regs->rip = (unsigned long)current_text_addr();
+ asm volatile("movl %%cs, %%eax;" :"=a"(core_regs->cs));
+ asm volatile("pushfq; popq %0" :"=m"(core_regs->eflags));
+ asm volatile("movq %%rsp,%0" : "=m"(core_regs->rsp));
+ asm volatile("movl %%ss, %%eax;" :"=a"(core_regs->ss));
+ /* thread_fs not filled in for now */
+ /* thread_gs not filled in for now */
+ asm volatile("movl %%ds, %%eax;" :"=a"(core_regs->ds));
+ asm volatile("movl %%es, %%eax;" :"=a"(core_regs->es));
+ asm volatile("movl %%fs, %%eax;" :"=a"(core_regs->fs));
+ asm volatile("movl %%gs, %%eax;" :"=a"(core_regs->gs));
+
+ asm volatile("mov %%cr0, %0" : "=r" (tmp) : );
+ xen_core_regs->cr0 = tmp;
+
+ asm volatile("mov %%cr2, %0" : "=r" (tmp) : );
+ xen_core_regs->cr2 = tmp;
+
+ asm volatile("mov %%cr3, %0" : "=r" (tmp) : );
+ xen_core_regs->cr3 = tmp;
+
+ asm volatile("mov %%cr4, %0" : "=r" (tmp) : );
+ xen_core_regs->cr4 = tmp;
}
#endif /* __X86_64_ELF_H__ */
--- 0004/xen/include/asm-x86/x86_64/kexec.h
+++ work/xen/include/asm-x86/x86_64/kexec.h
@@ -1,14 +1,29 @@
+/******************************************************************************
+ * kexec.h
+ *
+ * Based heavily on machine_kexec.c and kexec.h from Linux 2.6.19-rc1
+ *
+ */
+
#ifndef __X86_64_KEXEC_H__
#define __X86_64_KEXEC_H__
-
-#include <xen/lib.h> /* for printk() used in stub */
+
#include <xen/types.h>
-#include <public/xen.h>
#include <xen/kexec.h>
+#include <asm/fixmap.h>
+
+typedef void (*relocate_new_kernel_t)(
+ unsigned long indirection_page,
+ unsigned long page_list,
+ unsigned long start_address);
static inline void machine_kexec(xen_kexec_image_t *image)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ relocate_new_kernel_t rnk;
+
+ rnk = (relocate_new_kernel_t) image->page_list[1];
+ (*rnk)(image->indirection_page, (unsigned long)image->page_list,
+ image->start_address);
}
#endif /* __X86_64_KEXEC_H__ */
^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH 04/04] Kexec / Kdump: x86_64 specific code
2006-11-08 13:48 [PATCH 00/04] Kexec / Kdump: Release 20061108 (xen-unstable-12281) Magnus Damm
@ 2006-11-08 13:49 ` Magnus Damm
0 siblings, 0 replies; 30+ messages in thread
From: Magnus Damm @ 2006-11-08 13:49 UTC (permalink / raw)
To: Keir Fraser
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, magnus.damm,
Isaku Yamahata, Magnus Damm, Horms
[PATCH 04/04] Kexec / Kdump: x86_64 specific code
This patch contains the x86_64 implementation of Kexec / Kdump for Xen.
Signed-Off-By: Magnus Damm <magnus@valinux.co.jp>
---
Applies on top of xen-unstable-12281.
buildconfigs/linux-defconfig_xen_x86_64 | 1
linux-2.6-xen-sparse/arch/x86_64/Kconfig | 2
linux-2.6-xen-sparse/arch/x86_64/kernel/e820-xen.c | 6
linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c | 13
linux-2.6-xen-sparse/include/asm-x86_64/kexec-xen.h | 30
linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h | 7
linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h | 2
patches/linux-2.6.16.29/git-4b...1f.patch | 375 +++
patches/linux-2.6.16.29/linux-2.6.16.29-crash-xen-x86_64.patch | 35
patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec...code-x86_64.patch | 161 +
patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-xen-x86_64.patch | 162 +
patches/linux-2.6.16.29/series | 4
xen/arch/x86/x86_64/entry.S | 2
xen/include/asm-x86/x86_64/elf.h | 58 +
xen/include/asm-x86/x86_64/kexec.h | 23
15 files changed, 872 insertions(+), 9 deletions(-)
--- 0002/buildconfigs/linux-defconfig_xen_x86_64
+++ work/buildconfigs/linux-defconfig_xen_x86_64
@@ -138,6 +138,7 @@ CONFIG_SWIOTLB=y
CONFIG_PHYSICAL_START=0x100000
CONFIG_SECCOMP=y
CONFIG_HZ_100=y
+CONFIG_KEXEC=y
# CONFIG_HZ_250 is not set
# CONFIG_HZ_1000 is not set
CONFIG_HZ=100
--- 0001/linux-2.6-xen-sparse/arch/x86_64/Kconfig
+++ work/linux-2.6-xen-sparse/arch/x86_64/Kconfig
@@ -435,7 +435,7 @@ config X86_MCE_AMD
config KEXEC
bool "kexec system call (EXPERIMENTAL)"
- depends on EXPERIMENTAL && !X86_64_XEN
+ depends on EXPERIMENTAL && !XEN_UNPRIVILEGED_GUEST
help
kexec is a system call that implements the ability to shutdown your
current kernel, and to start another kernel. It is like a reboot
--- 0001/linux-2.6-xen-sparse/arch/x86_64/kernel/e820-xen.c
+++ work/linux-2.6-xen-sparse/arch/x86_64/kernel/e820-xen.c
@@ -260,7 +260,11 @@ void __init e820_reserve_resources(struc
request_resource(res, &data_resource);
#endif
#ifdef CONFIG_KEXEC
- request_resource(res, &crashk_res);
+ if (crashk_res.start != crashk_res.end)
+ request_resource(res, &crashk_res);
+#ifdef CONFIG_XEN
+ xen_machine_kexec_register_resources(res);
+#endif
#endif
}
}
--- 0001/linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c
+++ work/linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c
@@ -80,6 +80,10 @@
#include <asm/mach-xen/setup_arch_post.h>
#include <xen/interface/memory.h>
+#ifdef CONFIG_XEN
+#include <xen/interface/kexec.h>
+#endif
+
extern unsigned long start_pfn;
extern struct edid_info edid_info;
@@ -450,6 +454,7 @@ static __init void parse_cmdline_early (
* after a kernel panic.
*/
else if (!memcmp(from, "crashkernel=", 12)) {
+#ifndef CONFIG_XEN
unsigned long size, base;
size = memparse(from+12, &from);
if (*from == '@') {
@@ -460,6 +465,10 @@ static __init void parse_cmdline_early (
crashk_res.start = base;
crashk_res.end = base + size - 1;
}
+#else
+ printk("Ignoring crashkernel command line, "
+ "parameter will be supplied by xen\n");
+#endif
}
#endif
@@ -812,11 +821,15 @@ void __init setup_arch(char **cmdline_p)
#endif
#endif /* !CONFIG_XEN */
#ifdef CONFIG_KEXEC
+#ifdef CONFIG_XEN
+ xen_machine_kexec_setup_resources();
+#else
if (crashk_res.start != crashk_res.end) {
reserve_bootmem(crashk_res.start,
crashk_res.end - crashk_res.start + 1);
}
#endif
+#endif
paging_init();
#ifdef CONFIG_X86_LOCAL_APIC
--- /dev/null
+++ work/linux-2.6-xen-sparse/include/asm-x86_64/kexec-xen.h
@@ -0,0 +1,30 @@
+#ifndef _X86_64_KEXEC_XEN_H
+#define _X86_64_KEXEC_XEN_H
+
+/* Kexec needs to know about the actual physical addresss.
+ * But in xen, on some architectures, a physical address is a
+ * pseudo-physical addresss. */
+
+#ifdef CONFIG_XEN
+#define kexec_page_to_pfn(page) pfn_to_mfn(page_to_pfn(page))
+#define kexec_pfn_to_page(pfn) pfn_to_page(mfn_to_pfn(pfn))
+#define kexec_virt_to_phys(addr) virt_to_machine(addr)
+#define kexec_phys_to_virt(addr) phys_to_virt(machine_to_phys(addr))
+#else
+#define kexec_page_to_pfn(page) page_to_pfn(page)
+#define kexec_pfn_to_page(pfn) pfn_to_page(pfn)
+#define kexec_virt_to_phys(addr) virt_to_phys(addr)
+#define kexec_phys_to_virt(addr) phys_to_virt(addr)
+#endif
+
+#endif /* _X86_64_KEXEC_XEN_H */
+
+/*
+ * Local variables:
+ * c-file-style: "linux"
+ * indent-tabs-mode: t
+ * c-indent-level: 8
+ * c-basic-offset: 8
+ * tab-width: 8
+ * End:
+ */
--- 0001/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h
+++ work/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h
@@ -396,4 +396,11 @@ HYPERVISOR_xenoprof_op(
return _hypercall2(int, xenoprof_op, op, arg);
}
+static inline int
+HYPERVISOR_kexec_op(
+ unsigned long op, void *args)
+{
+ return _hypercall2(int, kexec_op, op, args);
+}
+
#endif /* __HYPERCALL_H__ */
--- 0001/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h
+++ work/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h
@@ -90,6 +90,8 @@ extern unsigned long profile_pc(struct p
#define profile_pc(regs) instruction_pointer(regs)
#endif
+#include <linux/compiler.h>
+
void signal_fault(struct pt_regs *regs, void __user *frame, char *where);
struct task_struct;
--- /dev/null
+++ work/patches/linux-2.6.16.29/git-4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f.patch
@@ -0,0 +1,375 @@
+From: Magnus Damm <magnus@valinux.co.jp>
+Date: Tue, 26 Sep 2006 08:52:38 +0000 (+0200)
+Subject: [PATCH] Avoid overwriting the current pgd (V4, x86_64)
+X-Git-Tag: v2.6.19-rc1
+X-Git-Url: http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f
+
+[PATCH] Avoid overwriting the current pgd (V4, x86_64)
+
+kexec: Avoid overwriting the current pgd (V4, x86_64)
+
+This patch upgrades the x86_64-specific kexec code to avoid overwriting the
+current pgd. Overwriting the current pgd is bad when CONFIG_CRASH_DUMP is used
+to start a secondary kernel that dumps the memory of the previous kernel.
+
+The code introduces a new set of page tables. These tables are used to provide
+an executable identity mapping without overwriting the current pgd.
+
+Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
+Signed-off-by: Andi Kleen <ak@suse.de>
+---
+
+--- a/arch/x86_64/kernel/machine_kexec.c
++++ b/arch/x86_64/kernel/machine_kexec.c
+@@ -15,6 +15,15 @@
+ #include <asm/mmu_context.h>
+ #include <asm/io.h>
+
++#define PAGE_ALIGNED __attribute__ ((__aligned__(PAGE_SIZE)))
++static u64 kexec_pgd[512] PAGE_ALIGNED;
++static u64 kexec_pud0[512] PAGE_ALIGNED;
++static u64 kexec_pmd0[512] PAGE_ALIGNED;
++static u64 kexec_pte0[512] PAGE_ALIGNED;
++static u64 kexec_pud1[512] PAGE_ALIGNED;
++static u64 kexec_pmd1[512] PAGE_ALIGNED;
++static u64 kexec_pte1[512] PAGE_ALIGNED;
++
+ static void init_level2_page(pmd_t *level2p, unsigned long addr)
+ {
+ unsigned long end_addr;
+@@ -144,32 +153,19 @@ static void load_segments(void)
+ );
+ }
+
+-typedef NORET_TYPE void (*relocate_new_kernel_t)(unsigned long indirection_page,
+- unsigned long control_code_buffer,
+- unsigned long start_address,
+- unsigned long pgtable) ATTRIB_NORET;
+-
+-extern const unsigned char relocate_new_kernel[];
+-extern const unsigned long relocate_new_kernel_size;
+-
+ int machine_kexec_prepare(struct kimage *image)
+ {
+- unsigned long start_pgtable, control_code_buffer;
++ unsigned long start_pgtable;
+ int result;
+
+ /* Calculate the offsets */
+ start_pgtable = page_to_pfn(image->control_code_page) << PAGE_SHIFT;
+- control_code_buffer = start_pgtable + PAGE_SIZE;
+
+ /* Setup the identity mapped 64bit page table */
+ result = init_pgtable(image, start_pgtable);
+ if (result)
+ return result;
+
+- /* Place the code in the reboot code buffer */
+- memcpy(__va(control_code_buffer), relocate_new_kernel,
+- relocate_new_kernel_size);
+-
+ return 0;
+ }
+
+@@ -184,28 +180,34 @@ void machine_kexec_cleanup(struct kimage
+ */
+ NORET_TYPE void machine_kexec(struct kimage *image)
+ {
+- unsigned long page_list;
+- unsigned long control_code_buffer;
+- unsigned long start_pgtable;
+- relocate_new_kernel_t rnk;
++ unsigned long page_list[PAGES_NR];
++ void *control_page;
+
+ /* Interrupts aren't acceptable while we reboot */
+ local_irq_disable();
+
+- /* Calculate the offsets */
+- page_list = image->head;
+- start_pgtable = page_to_pfn(image->control_code_page) << PAGE_SHIFT;
+- control_code_buffer = start_pgtable + PAGE_SIZE;
++ control_page = page_address(image->control_code_page) + PAGE_SIZE;
++ memcpy(control_page, relocate_kernel, PAGE_SIZE);
+
+- /* Set the low half of the page table to my identity mapped
+- * page table for kexec. Leave the high half pointing at the
+- * kernel pages. Don't bother to flush the global pages
+- * as that will happen when I fully switch to my identity mapped
+- * page table anyway.
+- */
+- memcpy(__va(read_cr3()), __va(start_pgtable), PAGE_SIZE/2);
+- __flush_tlb();
++ page_list[PA_CONTROL_PAGE] = __pa(control_page);
++ page_list[VA_CONTROL_PAGE] = (unsigned long)relocate_kernel;
++ page_list[PA_PGD] = __pa(kexec_pgd);
++ page_list[VA_PGD] = (unsigned long)kexec_pgd;
++ page_list[PA_PUD_0] = __pa(kexec_pud0);
++ page_list[VA_PUD_0] = (unsigned long)kexec_pud0;
++ page_list[PA_PMD_0] = __pa(kexec_pmd0);
++ page_list[VA_PMD_0] = (unsigned long)kexec_pmd0;
++ page_list[PA_PTE_0] = __pa(kexec_pte0);
++ page_list[VA_PTE_0] = (unsigned long)kexec_pte0;
++ page_list[PA_PUD_1] = __pa(kexec_pud1);
++ page_list[VA_PUD_1] = (unsigned long)kexec_pud1;
++ page_list[PA_PMD_1] = __pa(kexec_pmd1);
++ page_list[VA_PMD_1] = (unsigned long)kexec_pmd1;
++ page_list[PA_PTE_1] = __pa(kexec_pte1);
++ page_list[VA_PTE_1] = (unsigned long)kexec_pte1;
+
++ page_list[PA_TABLE_PAGE] =
++ (unsigned long)__pa(page_address(image->control_code_page));
+
+ /* The segment registers are funny things, they have both a
+ * visible and an invisible part. Whenever the visible part is
+@@ -222,9 +224,10 @@ NORET_TYPE void machine_kexec(struct kim
+ */
+ set_gdt(phys_to_virt(0),0);
+ set_idt(phys_to_virt(0),0);
++
+ /* now call it */
+- rnk = (relocate_new_kernel_t) control_code_buffer;
+- (*rnk)(page_list, control_code_buffer, image->start, start_pgtable);
++ relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
++ image->start);
+ }
+
+ /* crashkernel=size@addr specifies the location to reserve for
+--- a/arch/x86_64/kernel/relocate_kernel.S
++++ b/arch/x86_64/kernel/relocate_kernel.S
+@@ -7,31 +7,169 @@
+ */
+
+ #include <linux/linkage.h>
++#include <asm/page.h>
++#include <asm/kexec.h>
+
+- /*
+- * Must be relocatable PIC code callable as a C function, that once
+- * it starts can not use the previous processes stack.
+- */
+- .globl relocate_new_kernel
++/*
++ * Must be relocatable PIC code callable as a C function
++ */
++
++#define PTR(x) (x << 3)
++#define PAGE_ALIGNED (1 << PAGE_SHIFT)
++#define PAGE_ATTR 0x63 /* _PAGE_PRESENT|_PAGE_RW|_PAGE_ACCESSED|_PAGE_DIRTY */
++
++ .text
++ .align PAGE_ALIGNED
+ .code64
++ .globl relocate_kernel
++relocate_kernel:
++ /* %rdi indirection_page
++ * %rsi page_list
++ * %rdx start address
++ */
++
++ /* map the control page at its virtual address */
++
++ movq $0x0000ff8000000000, %r10 /* mask */
++ mov $(39 - 3), %cl /* bits to shift */
++ movq PTR(VA_CONTROL_PAGE)(%rsi), %r11 /* address to map */
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PGD)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PUD_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PUD_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PMD_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PMD_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PTE_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PTE_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ /* identity map the control page at its physical address */
++
++ movq $0x0000ff8000000000, %r10 /* mask */
++ mov $(39 - 3), %cl /* bits to shift */
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r11 /* address to map */
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PGD)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PUD_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PUD_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PMD_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PMD_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PTE_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PTE_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
+ relocate_new_kernel:
+- /* %rdi page_list
+- * %rsi reboot_code_buffer
++ /* %rdi indirection_page
++ * %rsi page_list
+ * %rdx start address
+- * %rcx page_table
+- * %r8 arg5
+- * %r9 arg6
+ */
+
+ /* zero out flags, and disable interrupts */
+ pushq $0
+ popfq
+
+- /* set a new stack at the bottom of our page... */
+- lea 4096(%rsi), %rsp
++ /* get physical address of control page now */
++ /* this is impossible after page table switch */
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++
++ /* get physical address of page table now too */
++ movq PTR(PA_TABLE_PAGE)(%rsi), %rcx
++
++ /* switch to new set of page tables */
++ movq PTR(PA_PGD)(%rsi), %r9
++ movq %r9, %cr3
++
++ /* setup a new stack at the end of the physical control page */
++ lea 4096(%r8), %rsp
++
++ /* jump to identity mapped page */
++ addq $(identity_mapped - relocate_kernel), %r8
++ pushq %r8
++ ret
+
+- /* store the parameters back on the stack */
+- pushq %rdx /* store the start address */
++identity_mapped:
++ /* store the start address on the stack */
++ pushq %rdx
+
+ /* Set cr0 to a known state:
+ * 31 1 == Paging enabled
+@@ -136,8 +274,3 @@ relocate_new_kernel:
+ xorq %r15, %r15
+
+ ret
+-relocate_new_kernel_end:
+-
+- .globl relocate_new_kernel_size
+-relocate_new_kernel_size:
+- .quad relocate_new_kernel_end - relocate_new_kernel
+--- a/include/asm-x86_64/kexec.h
++++ b/include/asm-x86_64/kexec.h
+@@ -1,6 +1,27 @@
+ #ifndef _X86_64_KEXEC_H
+ #define _X86_64_KEXEC_H
+
++#define PA_CONTROL_PAGE 0
++#define VA_CONTROL_PAGE 1
++#define PA_PGD 2
++#define VA_PGD 3
++#define PA_PUD_0 4
++#define VA_PUD_0 5
++#define PA_PMD_0 6
++#define VA_PMD_0 7
++#define PA_PTE_0 8
++#define VA_PTE_0 9
++#define PA_PUD_1 10
++#define VA_PUD_1 11
++#define PA_PMD_1 12
++#define VA_PMD_1 13
++#define PA_PTE_1 14
++#define VA_PTE_1 15
++#define PA_TABLE_PAGE 16
++#define PAGES_NR 17
++
++#ifndef __ASSEMBLY__
++
+ #include <linux/string.h>
+
+ #include <asm/page.h>
+@@ -64,4 +85,12 @@ static inline void crash_setup_regs(stru
+ newregs->rip = (unsigned long)current_text_addr();
+ }
+ }
++
++NORET_TYPE void
++relocate_kernel(unsigned long indirection_page,
++ unsigned long page_list,
++ unsigned long start_address) ATTRIB_NORET;
++
++#endif /* __ASSEMBLY__ */
++
+ #endif /* _X86_64_KEXEC_H */
--- /dev/null
+++ work/patches/linux-2.6.16.29/linux-2.6.16.29-crash-xen-x86_64.patch
@@ -0,0 +1,35 @@
+--- 0001/arch/x86_64/kernel/crash.c
++++ work/arch/x86_64/kernel/crash.c 2006-10-26 15:44:27.000000000 +0900
+@@ -92,6 +92,7 @@ static void crash_save_self(struct pt_re
+ crash_save_this_cpu(regs, cpu);
+ }
+
++#ifndef CONFIG_XEN
+ #ifdef CONFIG_SMP
+ static atomic_t waiting_for_crash_ipi;
+
+@@ -156,6 +157,7 @@ static void nmi_shootdown_cpus(void)
+ /* There are no cpus to shootdown */
+ }
+ #endif
++#endif /* CONFIG_XEN */
+
+ void machine_crash_shutdown(struct pt_regs *regs)
+ {
+@@ -173,6 +175,8 @@ void machine_crash_shutdown(struct pt_re
+
+ /* Make a note of crashing cpu. Will be used in NMI callback.*/
+ crashing_cpu = smp_processor_id();
++
++#ifndef CONFIG_XEN
+ nmi_shootdown_cpus();
+
+ if(cpu_has_apic)
+@@ -181,6 +185,6 @@ void machine_crash_shutdown(struct pt_re
+ #if defined(CONFIG_X86_IO_APIC)
+ disable_IO_APIC();
+ #endif
+-
++#endif /* CONFIG_XEN */
+ crash_save_self(regs);
+ }
--- /dev/null
+++ work/patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-move_segment_code-x86_64.patch
@@ -0,0 +1,161 @@
+kexec: Move asm segment handling code to the assembly file (x86_64)
+
+This patch moves the idt, gdt, and segment handling code from machine_kexec.c
+to relocate_kernel.S. The main reason behind this move is to avoid code
+duplication in the Xen hypervisor. With this patch all code required to kexec
+is put on the control page.
+
+On top of that this patch also counts as a cleanup - I think it is much
+nicer to write assembly directly in assembly files than wrap inline assembly
+in C functions for no apparent reason.
+
+Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
+---
+
+ Applies to 2.6.19-rc1.
+
+ machine_kexec.c | 58 -----------------------------------------------------
+ relocate_kernel.S | 50 +++++++++++++++++++++++++++++++++++++++++----
+ 2 files changed, 45 insertions(+), 63 deletions(-)
+
+--- 0002/arch/x86_64/kernel/machine_kexec.c
++++ work/arch/x86_64/kernel/machine_kexec.c 2006-10-05 16:15:49.000000000 +0900
+@@ -112,47 +112,6 @@ static int init_pgtable(struct kimage *i
+ return init_level4_page(image, level4p, 0, end_pfn << PAGE_SHIFT);
+ }
+
+-static void set_idt(void *newidt, u16 limit)
+-{
+- struct desc_ptr curidt;
+-
+- /* x86-64 supports unaliged loads & stores */
+- curidt.size = limit;
+- curidt.address = (unsigned long)newidt;
+-
+- __asm__ __volatile__ (
+- "lidtq %0\n"
+- : : "m" (curidt)
+- );
+-};
+-
+-
+-static void set_gdt(void *newgdt, u16 limit)
+-{
+- struct desc_ptr curgdt;
+-
+- /* x86-64 supports unaligned loads & stores */
+- curgdt.size = limit;
+- curgdt.address = (unsigned long)newgdt;
+-
+- __asm__ __volatile__ (
+- "lgdtq %0\n"
+- : : "m" (curgdt)
+- );
+-};
+-
+-static void load_segments(void)
+-{
+- __asm__ __volatile__ (
+- "\tmovl %0,%%ds\n"
+- "\tmovl %0,%%es\n"
+- "\tmovl %0,%%ss\n"
+- "\tmovl %0,%%fs\n"
+- "\tmovl %0,%%gs\n"
+- : : "a" (__KERNEL_DS) : "memory"
+- );
+-}
+-
+ int machine_kexec_prepare(struct kimage *image)
+ {
+ unsigned long start_pgtable;
+@@ -209,23 +168,6 @@ NORET_TYPE void machine_kexec(struct kim
+ page_list[PA_TABLE_PAGE] =
+ (unsigned long)__pa(page_address(image->control_code_page));
+
+- /* The segment registers are funny things, they have both a
+- * visible and an invisible part. Whenever the visible part is
+- * set to a specific selector, the invisible part is loaded
+- * with from a table in memory. At no other time is the
+- * descriptor table in memory accessed.
+- *
+- * I take advantage of this here by force loading the
+- * segments, before I zap the gdt with an invalid value.
+- */
+- load_segments();
+- /* The gdt & idt are now invalid.
+- * If you want to load them you must set up your own idt & gdt.
+- */
+- set_gdt(phys_to_virt(0),0);
+- set_idt(phys_to_virt(0),0);
+-
+- /* now call it */
+ relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
+ image->start);
+ }
+--- 0002/arch/x86_64/kernel/relocate_kernel.S
++++ work/arch/x86_64/kernel/relocate_kernel.S 2006-10-05 16:18:07.000000000 +0900
+@@ -159,13 +159,39 @@ relocate_new_kernel:
+ movq PTR(PA_PGD)(%rsi), %r9
+ movq %r9, %cr3
+
++ /* setup idt */
++ movq %r8, %rax
++ addq $(idt_80 - relocate_kernel), %rax
++ lidtq (%rax)
++
++ /* setup gdt */
++ movq %r8, %rax
++ addq $(gdt - relocate_kernel), %rax
++ movq %r8, %r9
++ addq $((gdt_80 - relocate_kernel) + 2), %r9
++ movq %rax, (%r9)
++
++ movq %r8, %rax
++ addq $(gdt_80 - relocate_kernel), %rax
++ lgdtq (%rax)
++
++ /* setup data segment registers */
++ xorl %eax, %eax
++ movl %eax, %ds
++ movl %eax, %es
++ movl %eax, %fs
++ movl %eax, %gs
++ movl %eax, %ss
++
+ /* setup a new stack at the end of the physical control page */
+ lea 4096(%r8), %rsp
+
+- /* jump to identity mapped page */
+- addq $(identity_mapped - relocate_kernel), %r8
+- pushq %r8
+- ret
++ /* load new code segment and jump to identity mapped page */
++ movq %r8, %rax
++ addq $(identity_mapped - relocate_kernel), %rax
++ pushq $(gdt_cs - gdt)
++ pushq %rax
++ lretq
+
+ identity_mapped:
+ /* store the start address on the stack */
+@@ -272,5 +298,19 @@ identity_mapped:
+ xorq %r13, %r13
+ xorq %r14, %r14
+ xorq %r15, %r15
+-
+ ret
++
++ .align 16
++gdt:
++ .quad 0x0000000000000000 /* NULL descriptor */
++gdt_cs:
++ .quad 0x00af9a000000ffff
++gdt_end:
++
++gdt_80:
++ .word gdt_end - gdt - 1 /* limit */
++ .quad 0 /* base - filled in by code above */
++
++idt_80:
++ .word 0 /* limit */
++ .quad 0 /* base */
--- /dev/null
+++ work/patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-xen-x86_64.patch
@@ -0,0 +1,162 @@
+--- 0006/arch/x86_64/kernel/machine_kexec.c
++++ work/arch/x86_64/kernel/machine_kexec.c 2006-10-06 15:36:16.000000000 +0900
+@@ -24,6 +24,104 @@ static u64 kexec_pud1[512] PAGE_ALIGNED;
+ static u64 kexec_pmd1[512] PAGE_ALIGNED;
+ static u64 kexec_pte1[512] PAGE_ALIGNED;
+
++#ifdef CONFIG_XEN
++
++/* In the case of Xen, override hypervisor functions to be able to create
++ * a regular identity mapping page table...
++ */
++
++#include <xen/interface/kexec.h>
++#include <xen/interface/memory.h>
++
++#define x__pmd(x) ((pmd_t) { (x) } )
++#define x__pud(x) ((pud_t) { (x) } )
++#define x__pgd(x) ((pgd_t) { (x) } )
++
++#define x_pmd_val(x) ((x).pmd)
++#define x_pud_val(x) ((x).pud)
++#define x_pgd_val(x) ((x).pgd)
++
++static inline void x_set_pmd(pmd_t *dst, pmd_t val)
++{
++ x_pmd_val(*dst) = x_pmd_val(val);
++}
++
++static inline void x_set_pud(pud_t *dst, pud_t val)
++{
++ x_pud_val(*dst) = phys_to_machine(x_pud_val(val));
++}
++
++static inline void x_pud_clear (pud_t *pud)
++{
++ x_pud_val(*pud) = 0;
++}
++
++static inline void x_set_pgd(pgd_t *dst, pgd_t val)
++{
++ x_pgd_val(*dst) = phys_to_machine(x_pgd_val(val));
++}
++
++static inline void x_pgd_clear (pgd_t * pgd)
++{
++ x_pgd_val(*pgd) = 0;
++}
++
++#define X__PAGE_KERNEL_LARGE_EXEC \
++ _PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_PSE
++#define X_KERNPG_TABLE _PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY
++
++#define __ma(x) (pfn_to_mfn(__pa((x)) >> PAGE_SHIFT) << PAGE_SHIFT)
++
++#if PAGES_NR > KEXEC_XEN_NO_PAGES
++#error PAGES_NR is greater than KEXEC_XEN_NO_PAGES - Xen support will break
++#endif
++
++#if PA_CONTROL_PAGE != 0
++#error PA_CONTROL_PAGE is non zero - Xen support will break
++#endif
++
++void machine_kexec_setup_load_arg(xen_kexec_image_t *xki, struct kimage *image)
++{
++ void *control_page;
++ void *table_page;
++
++ memset(xki->page_list, 0, sizeof(xki->page_list));
++
++ control_page = page_address(image->control_code_page) + PAGE_SIZE;
++ memcpy(control_page, relocate_kernel, PAGE_SIZE);
++
++ table_page = page_address(image->control_code_page);
++
++ xki->page_list[PA_CONTROL_PAGE] = __ma(control_page);
++ xki->page_list[PA_TABLE_PAGE] = __ma(table_page);
++
++ xki->page_list[PA_PGD] = __ma(kexec_pgd);
++ xki->page_list[PA_PUD_0] = __ma(kexec_pud0);
++ xki->page_list[PA_PUD_1] = __ma(kexec_pud1);
++ xki->page_list[PA_PMD_0] = __ma(kexec_pmd0);
++ xki->page_list[PA_PMD_1] = __ma(kexec_pmd1);
++ xki->page_list[PA_PTE_0] = __ma(kexec_pte0);
++ xki->page_list[PA_PTE_1] = __ma(kexec_pte1);
++}
++
++#else /* CONFIG_XEN */
++
++#define x__pmd(x) __pmd(x)
++#define x__pud(x) __pud(x)
++#define x__pgd(x) __pgd(x)
++
++#define x_set_pmd(x, y) set_pmd(x, y)
++#define x_set_pud(x, y) set_pud(x, y)
++#define x_set_pgd(x, y) set_pgd(x, y)
++
++#define x_pud_clear(x) pud_clear(x)
++#define x_pgd_clear(x) pgd_clear(x)
++
++#define X__PAGE_KERNEL_LARGE_EXEC __PAGE_KERNEL_LARGE_EXEC
++#define X_KERNPG_TABLE _KERNPG_TABLE
++
++#endif /* CONFIG_XEN */
++
+ static void init_level2_page(pmd_t *level2p, unsigned long addr)
+ {
+ unsigned long end_addr;
+@@ -31,7 +129,7 @@ static void init_level2_page(pmd_t *leve
+ addr &= PAGE_MASK;
+ end_addr = addr + PUD_SIZE;
+ while (addr < end_addr) {
+- set_pmd(level2p++, __pmd(addr | __PAGE_KERNEL_LARGE_EXEC));
++ x_set_pmd(level2p++, x__pmd(addr | X__PAGE_KERNEL_LARGE_EXEC));
+ addr += PMD_SIZE;
+ }
+ }
+@@ -56,12 +154,12 @@ static int init_level3_page(struct kimag
+ }
+ level2p = (pmd_t *)page_address(page);
+ init_level2_page(level2p, addr);
+- set_pud(level3p++, __pud(__pa(level2p) | _KERNPG_TABLE));
++ x_set_pud(level3p++, x__pud(__pa(level2p) | X_KERNPG_TABLE));
+ addr += PUD_SIZE;
+ }
+ /* clear the unused entries */
+ while (addr < end_addr) {
+- pud_clear(level3p++);
++ x_pud_clear(level3p++);
+ addr += PUD_SIZE;
+ }
+ out:
+@@ -92,12 +190,12 @@ static int init_level4_page(struct kimag
+ if (result) {
+ goto out;
+ }
+- set_pgd(level4p++, __pgd(__pa(level3p) | _KERNPG_TABLE));
++ x_set_pgd(level4p++, x__pgd(__pa(level3p) | X_KERNPG_TABLE));
+ addr += PGDIR_SIZE;
+ }
+ /* clear the unused entries */
+ while (addr < end_addr) {
+- pgd_clear(level4p++);
++ x_pgd_clear(level4p++);
+ addr += PGDIR_SIZE;
+ }
+ out:
+@@ -108,8 +206,14 @@ out:
+ static int init_pgtable(struct kimage *image, unsigned long start_pgtable)
+ {
+ pgd_t *level4p;
++ unsigned long x_end_pfn = end_pfn;
++
++#ifdef CONFIG_XEN
++ x_end_pfn = HYPERVISOR_memory_op(XENMEM_maximum_ram_page, NULL);
++#endif
++
+ level4p = (pgd_t *)__va(start_pgtable);
+- return init_level4_page(image, level4p, 0, end_pfn << PAGE_SHIFT);
++ return init_level4_page(image, level4p, 0, x_end_pfn << PAGE_SHIFT);
+ }
+
+ int machine_kexec_prepare(struct kimage *image)
--- 0005/patches/linux-2.6.16.29/series
+++ work/patches/linux-2.6.16.29/series
@@ -5,6 +5,10 @@ git-3566561bfadffcb5dbc85d576be80c0dbf2c
linux-2.6.19-rc1-kexec-move_segment_code-i386.patch
linux-2.6.19-rc1-kexec-xen-i386.patch
linux-2.6.16.29-crash-xen-i386.patch
+git-4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f.patch
+linux-2.6.19-rc1-kexec-move_segment_code-x86_64.patch
+linux-2.6.19-rc1-kexec-xen-x86_64.patch
+linux-2.6.16.29-crash-xen-x86_64.patch
blktap-aio-16_03_06.patch
device_bind.patch
fix-hz-suspend.patch
--- 0001/xen/arch/x86/x86_64/entry.S
+++ work/xen/arch/x86/x86_64/entry.S
@@ -573,6 +573,7 @@ ENTRY(hypercall_table)
.quad do_hvm_op
.quad do_sysctl /* 35 */
.quad do_domctl
+ .quad do_kexec_op
.rept NR_hypercalls-((.-hypercall_table)/8)
.quad do_ni_hypercall
.endr
@@ -615,6 +616,7 @@ ENTRY(hypercall_args_table)
.byte 2 /* do_hvm_op */
.byte 1 /* do_sysctl */ /* 35 */
.byte 1 /* do_domctl */
+ .byte 2 /* do_kexec */
.rept NR_hypercalls-(.-hypercall_args_table)
.byte 0 /* do_ni_hypercall */
.endr
--- 0004/xen/include/asm-x86/x86_64/elf.h
+++ work/xen/include/asm-x86/x86_64/elf.h
@@ -1,15 +1,67 @@
#ifndef __X86_64_ELF_H__
#define __X86_64_ELF_H__
-#include <xen/lib.h> /* for printk() used in stub */
+#include <asm/processor.h>
typedef struct {
- unsigned long dummy;
+ unsigned long r15;
+ unsigned long r14;
+ unsigned long r13;
+ unsigned long r12;
+ unsigned long rbp;
+ unsigned long rbx;
+ unsigned long r11;
+ unsigned long r10;
+ unsigned long r9;
+ unsigned long r8;
+ unsigned long rax;
+ unsigned long rcx;
+ unsigned long rdx;
+ unsigned long rsi;
+ unsigned long rdi;
+ unsigned long orig_rax;
+ unsigned long rip;
+ unsigned long cs;
+ unsigned long eflags;
+ unsigned long rsp;
+ unsigned long ss;
+ unsigned long thread_fs;
+ unsigned long thread_gs;
+ unsigned long ds;
+ unsigned long es;
+ unsigned long fs;
+ unsigned long gs;
} ELF_Gregset;
extern inline void elf_core_save_regs(ELF_Gregset *dst)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ asm volatile("movq %%r15,%0" : "=m"(dst->r15));
+ asm volatile("movq %%r14,%0" : "=m"(dst->r14));
+ asm volatile("movq %%r13,%0" : "=m"(dst->r13));
+ asm volatile("movq %%r12,%0" : "=m"(dst->r12));
+ asm volatile("movq %%rbp,%0" : "=m"(dst->rbp));
+ asm volatile("movq %%rbx,%0" : "=m"(dst->rbx));
+ asm volatile("movq %%r11,%0" : "=m"(dst->r11));
+ asm volatile("movq %%r10,%0" : "=m"(dst->r10));
+ asm volatile("movq %%r9,%0" : "=m"(dst->r9));
+ asm volatile("movq %%r8,%0" : "=m"(dst->r8));
+ asm volatile("movq %%rax,%0" : "=m"(dst->rax));
+ asm volatile("movq %%rcx,%0" : "=m"(dst->rcx));
+ asm volatile("movq %%rdx,%0" : "=m"(dst->rdx));
+ asm volatile("movq %%rsi,%0" : "=m"(dst->rsi));
+ asm volatile("movq %%rdi,%0" : "=m"(dst->rdi));
+ /* orig_rax not filled in for now */
+ dst->rip = (unsigned long)current_text_addr();
+ asm volatile("movl %%cs, %%eax;" :"=a"(dst->cs));
+ asm volatile("pushfq; popq %0" :"=m"(dst->eflags));
+ asm volatile("movq %%rsp,%0" : "=m"(dst->rsp));
+ asm volatile("movl %%ss, %%eax;" :"=a"(dst->ss));
+ /* thread_fs not filled in for now */
+ /* thread_gs not filled in for now */
+ asm volatile("movl %%ds, %%eax;" :"=a"(dst->ds));
+ asm volatile("movl %%es, %%eax;" :"=a"(dst->es));
+ asm volatile("movl %%fs, %%eax;" :"=a"(dst->fs));
+ asm volatile("movl %%gs, %%eax;" :"=a"(dst->gs));
}
#endif /* __X86_64_ELF_H__ */
--- 0004/xen/include/asm-x86/x86_64/kexec.h
+++ work/xen/include/asm-x86/x86_64/kexec.h
@@ -1,14 +1,29 @@
+/******************************************************************************
+ * kexec.h
+ *
+ * Based heavily on machine_kexec.c and kexec.h from Linux 2.6.19-rc1
+ *
+ */
+
#ifndef __X86_64_KEXEC_H__
#define __X86_64_KEXEC_H__
-
-#include <xen/lib.h> /* for printk() used in stub */
+
#include <xen/types.h>
-#include <public/xen.h>
#include <xen/kexec.h>
+#include <asm/fixmap.h>
+
+typedef void (*relocate_new_kernel_t)(
+ unsigned long indirection_page,
+ unsigned long page_list,
+ unsigned long start_address);
static inline void machine_kexec(xen_kexec_image_t *image)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ relocate_new_kernel_t rnk;
+
+ rnk = (relocate_new_kernel_t) image->page_list[1];
+ (*rnk)(image->indirection_page, (unsigned long)image->page_list,
+ image->start_address);
}
#endif /* __X86_64_KEXEC_H__ */
^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH 04/04] Kexec / Kdump: x86_64 specific code
2006-10-30 10:03 [PATCH 00/04] Kexec / Kdump: Release 20061030 (xen-unstable-12025) Magnus Damm
@ 2006-10-30 10:04 ` Magnus Damm
0 siblings, 0 replies; 30+ messages in thread
From: Magnus Damm @ 2006-10-30 10:04 UTC (permalink / raw)
To: Keir Fraser
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, magnus.damm,
Isaku Yamahata, Magnus Damm, Horms
[PATCH 04/04] Kexec / Kdump: x86_64 specific code
This patch contains the x86_64 implementation of Kexec / Kdump for Xen.
Signed-Off-By: Magnus Damm <magnus@valinux.co.jp>
---
Applies on top of xen-unstable-12025.
buildconfigs/linux-defconfig_xen_x86_64 | 1
linux-2.6-xen-sparse/arch/x86_64/Kconfig | 2
linux-2.6-xen-sparse/arch/x86_64/kernel/e820-xen.c | 6
linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c | 13
linux-2.6-xen-sparse/include/asm-x86_64/kexec-xen.h | 30
linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h | 7
linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h | 2
patches/linux-2.6.16.29/git-4b..1f.patch | 375 +++++
patches/linux-2.6.16.29/linux-2.6.16.29-crash-xen-x86_64.patch | 35
patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec..code-x86_64.patch | 161 +++
patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-xen-x86_64.patch | 162 +++
patches/linux-2.6.16.29/series | 4
xen/arch/x86/x86_64/entry.S | 2
xen/include/asm-x86/x86_64/elf.h | 58 +
xen/include/asm-x86/x86_64/kexec.h | 23
15 files changed, 872 insertions(+), 9 deletions(-)
--- 0002/buildconfigs/linux-defconfig_xen_x86_64
+++ work/buildconfigs/linux-defconfig_xen_x86_64 2006-10-30 15:48:58.000000000 +0900
@@ -138,6 +138,7 @@ CONFIG_SWIOTLB=y
CONFIG_PHYSICAL_START=0x100000
CONFIG_SECCOMP=y
CONFIG_HZ_100=y
+CONFIG_KEXEC=y
# CONFIG_HZ_250 is not set
# CONFIG_HZ_1000 is not set
CONFIG_HZ=100
--- 0001/linux-2.6-xen-sparse/arch/x86_64/Kconfig
+++ work/linux-2.6-xen-sparse/arch/x86_64/Kconfig 2006-10-30 15:48:58.000000000 +0900
@@ -435,7 +435,7 @@ config X86_MCE_AMD
config KEXEC
bool "kexec system call (EXPERIMENTAL)"
- depends on EXPERIMENTAL && !X86_64_XEN
+ depends on EXPERIMENTAL && !XEN_UNPRIVILEGED_GUEST
help
kexec is a system call that implements the ability to shutdown your
current kernel, and to start another kernel. It is like a reboot
--- 0001/linux-2.6-xen-sparse/arch/x86_64/kernel/e820-xen.c
+++ work/linux-2.6-xen-sparse/arch/x86_64/kernel/e820-xen.c 2006-10-30 15:48:58.000000000 +0900
@@ -260,7 +260,11 @@ void __init e820_reserve_resources(struc
request_resource(res, &data_resource);
#endif
#ifdef CONFIG_KEXEC
- request_resource(res, &crashk_res);
+ if (crashk_res.start != crashk_res.end)
+ request_resource(res, &crashk_res);
+#ifdef CONFIG_XEN
+ xen_machine_kexec_register_resources(res);
+#endif
#endif
}
}
--- 0001/linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c
+++ work/linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c 2006-10-30 15:48:58.000000000 +0900
@@ -80,6 +80,10 @@
#include <asm/mach-xen/setup_arch_post.h>
#include <xen/interface/memory.h>
+#ifdef CONFIG_XEN
+#include <xen/interface/kexec.h>
+#endif
+
extern unsigned long start_pfn;
extern struct edid_info edid_info;
@@ -450,6 +454,7 @@ static __init void parse_cmdline_early (
* after a kernel panic.
*/
else if (!memcmp(from, "crashkernel=", 12)) {
+#ifndef CONFIG_XEN
unsigned long size, base;
size = memparse(from+12, &from);
if (*from == '@') {
@@ -460,6 +465,10 @@ static __init void parse_cmdline_early (
crashk_res.start = base;
crashk_res.end = base + size - 1;
}
+#else
+ printk("Ignoring crashkernel command line, "
+ "parameter will be supplied by xen\n");
+#endif
}
#endif
@@ -812,11 +821,15 @@ void __init setup_arch(char **cmdline_p)
#endif
#endif /* !CONFIG_XEN */
#ifdef CONFIG_KEXEC
+#ifdef CONFIG_XEN
+ xen_machine_kexec_setup_resources();
+#else
if (crashk_res.start != crashk_res.end) {
reserve_bootmem(crashk_res.start,
crashk_res.end - crashk_res.start + 1);
}
#endif
+#endif
paging_init();
#ifdef CONFIG_X86_LOCAL_APIC
--- /dev/null
+++ work/linux-2.6-xen-sparse/include/asm-x86_64/kexec-xen.h 2006-10-30 15:48:59.000000000 +0900
@@ -0,0 +1,30 @@
+#ifndef _X86_64_KEXEC_XEN_H
+#define _X86_64_KEXEC_XEN_H
+
+/* Kexec needs to know about the actual physical addresss.
+ * But in xen, on some architectures, a physical address is a
+ * pseudo-physical addresss. */
+
+#ifdef CONFIG_XEN
+#define kexec_page_to_pfn(page) pfn_to_mfn(page_to_pfn(page))
+#define kexec_pfn_to_page(pfn) pfn_to_page(mfn_to_pfn(pfn))
+#define kexec_virt_to_phys(addr) virt_to_machine(addr)
+#define kexec_phys_to_virt(addr) phys_to_virt(machine_to_phys(addr))
+#else
+#define kexec_page_to_pfn(page) page_to_pfn(page)
+#define kexec_pfn_to_page(pfn) pfn_to_page(pfn)
+#define kexec_virt_to_phys(addr) virt_to_phys(addr)
+#define kexec_phys_to_virt(addr) phys_to_virt(addr)
+#endif
+
+#endif /* _X86_64_KEXEC_XEN_H */
+
+/*
+ * Local variables:
+ * c-file-style: "linux"
+ * indent-tabs-mode: t
+ * c-indent-level: 8
+ * c-basic-offset: 8
+ * tab-width: 8
+ * End:
+ */
--- 0001/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h
+++ work/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h 2006-10-30 15:48:58.000000000 +0900
@@ -386,4 +386,11 @@ HYPERVISOR_xenoprof_op(
return _hypercall2(int, xenoprof_op, op, arg);
}
+static inline int
+HYPERVISOR_kexec_op(
+ unsigned long op, void *args)
+{
+ return _hypercall2(int, kexec_op, op, args);
+}
+
#endif /* __HYPERCALL_H__ */
--- 0001/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h
+++ work/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h 2006-10-30 15:48:58.000000000 +0900
@@ -90,6 +90,8 @@ extern unsigned long profile_pc(struct p
#define profile_pc(regs) instruction_pointer(regs)
#endif
+#include <linux/compiler.h>
+
void signal_fault(struct pt_regs *regs, void __user *frame, char *where);
struct task_struct;
--- /dev/null
+++ work/patches/linux-2.6.16.29/git-4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f.patch 2006-10-30 15:48:59.000000000 +0900
@@ -0,0 +1,375 @@
+From: Magnus Damm <magnus@valinux.co.jp>
+Date: Tue, 26 Sep 2006 08:52:38 +0000 (+0200)
+Subject: [PATCH] Avoid overwriting the current pgd (V4, x86_64)
+X-Git-Tag: v2.6.19-rc1
+X-Git-Url: http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f
+
+[PATCH] Avoid overwriting the current pgd (V4, x86_64)
+
+kexec: Avoid overwriting the current pgd (V4, x86_64)
+
+This patch upgrades the x86_64-specific kexec code to avoid overwriting the
+current pgd. Overwriting the current pgd is bad when CONFIG_CRASH_DUMP is used
+to start a secondary kernel that dumps the memory of the previous kernel.
+
+The code introduces a new set of page tables. These tables are used to provide
+an executable identity mapping without overwriting the current pgd.
+
+Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
+Signed-off-by: Andi Kleen <ak@suse.de>
+---
+
+--- a/arch/x86_64/kernel/machine_kexec.c
++++ b/arch/x86_64/kernel/machine_kexec.c
+@@ -15,6 +15,15 @@
+ #include <asm/mmu_context.h>
+ #include <asm/io.h>
+
++#define PAGE_ALIGNED __attribute__ ((__aligned__(PAGE_SIZE)))
++static u64 kexec_pgd[512] PAGE_ALIGNED;
++static u64 kexec_pud0[512] PAGE_ALIGNED;
++static u64 kexec_pmd0[512] PAGE_ALIGNED;
++static u64 kexec_pte0[512] PAGE_ALIGNED;
++static u64 kexec_pud1[512] PAGE_ALIGNED;
++static u64 kexec_pmd1[512] PAGE_ALIGNED;
++static u64 kexec_pte1[512] PAGE_ALIGNED;
++
+ static void init_level2_page(pmd_t *level2p, unsigned long addr)
+ {
+ unsigned long end_addr;
+@@ -144,32 +153,19 @@ static void load_segments(void)
+ );
+ }
+
+-typedef NORET_TYPE void (*relocate_new_kernel_t)(unsigned long indirection_page,
+- unsigned long control_code_buffer,
+- unsigned long start_address,
+- unsigned long pgtable) ATTRIB_NORET;
+-
+-extern const unsigned char relocate_new_kernel[];
+-extern const unsigned long relocate_new_kernel_size;
+-
+ int machine_kexec_prepare(struct kimage *image)
+ {
+- unsigned long start_pgtable, control_code_buffer;
++ unsigned long start_pgtable;
+ int result;
+
+ /* Calculate the offsets */
+ start_pgtable = page_to_pfn(image->control_code_page) << PAGE_SHIFT;
+- control_code_buffer = start_pgtable + PAGE_SIZE;
+
+ /* Setup the identity mapped 64bit page table */
+ result = init_pgtable(image, start_pgtable);
+ if (result)
+ return result;
+
+- /* Place the code in the reboot code buffer */
+- memcpy(__va(control_code_buffer), relocate_new_kernel,
+- relocate_new_kernel_size);
+-
+ return 0;
+ }
+
+@@ -184,28 +180,34 @@ void machine_kexec_cleanup(struct kimage
+ */
+ NORET_TYPE void machine_kexec(struct kimage *image)
+ {
+- unsigned long page_list;
+- unsigned long control_code_buffer;
+- unsigned long start_pgtable;
+- relocate_new_kernel_t rnk;
++ unsigned long page_list[PAGES_NR];
++ void *control_page;
+
+ /* Interrupts aren't acceptable while we reboot */
+ local_irq_disable();
+
+- /* Calculate the offsets */
+- page_list = image->head;
+- start_pgtable = page_to_pfn(image->control_code_page) << PAGE_SHIFT;
+- control_code_buffer = start_pgtable + PAGE_SIZE;
++ control_page = page_address(image->control_code_page) + PAGE_SIZE;
++ memcpy(control_page, relocate_kernel, PAGE_SIZE);
+
+- /* Set the low half of the page table to my identity mapped
+- * page table for kexec. Leave the high half pointing at the
+- * kernel pages. Don't bother to flush the global pages
+- * as that will happen when I fully switch to my identity mapped
+- * page table anyway.
+- */
+- memcpy(__va(read_cr3()), __va(start_pgtable), PAGE_SIZE/2);
+- __flush_tlb();
++ page_list[PA_CONTROL_PAGE] = __pa(control_page);
++ page_list[VA_CONTROL_PAGE] = (unsigned long)relocate_kernel;
++ page_list[PA_PGD] = __pa(kexec_pgd);
++ page_list[VA_PGD] = (unsigned long)kexec_pgd;
++ page_list[PA_PUD_0] = __pa(kexec_pud0);
++ page_list[VA_PUD_0] = (unsigned long)kexec_pud0;
++ page_list[PA_PMD_0] = __pa(kexec_pmd0);
++ page_list[VA_PMD_0] = (unsigned long)kexec_pmd0;
++ page_list[PA_PTE_0] = __pa(kexec_pte0);
++ page_list[VA_PTE_0] = (unsigned long)kexec_pte0;
++ page_list[PA_PUD_1] = __pa(kexec_pud1);
++ page_list[VA_PUD_1] = (unsigned long)kexec_pud1;
++ page_list[PA_PMD_1] = __pa(kexec_pmd1);
++ page_list[VA_PMD_1] = (unsigned long)kexec_pmd1;
++ page_list[PA_PTE_1] = __pa(kexec_pte1);
++ page_list[VA_PTE_1] = (unsigned long)kexec_pte1;
+
++ page_list[PA_TABLE_PAGE] =
++ (unsigned long)__pa(page_address(image->control_code_page));
+
+ /* The segment registers are funny things, they have both a
+ * visible and an invisible part. Whenever the visible part is
+@@ -222,9 +224,10 @@ NORET_TYPE void machine_kexec(struct kim
+ */
+ set_gdt(phys_to_virt(0),0);
+ set_idt(phys_to_virt(0),0);
++
+ /* now call it */
+- rnk = (relocate_new_kernel_t) control_code_buffer;
+- (*rnk)(page_list, control_code_buffer, image->start, start_pgtable);
++ relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
++ image->start);
+ }
+
+ /* crashkernel=size@addr specifies the location to reserve for
+--- a/arch/x86_64/kernel/relocate_kernel.S
++++ b/arch/x86_64/kernel/relocate_kernel.S
+@@ -7,31 +7,169 @@
+ */
+
+ #include <linux/linkage.h>
++#include <asm/page.h>
++#include <asm/kexec.h>
+
+- /*
+- * Must be relocatable PIC code callable as a C function, that once
+- * it starts can not use the previous processes stack.
+- */
+- .globl relocate_new_kernel
++/*
++ * Must be relocatable PIC code callable as a C function
++ */
++
++#define PTR(x) (x << 3)
++#define PAGE_ALIGNED (1 << PAGE_SHIFT)
++#define PAGE_ATTR 0x63 /* _PAGE_PRESENT|_PAGE_RW|_PAGE_ACCESSED|_PAGE_DIRTY */
++
++ .text
++ .align PAGE_ALIGNED
+ .code64
++ .globl relocate_kernel
++relocate_kernel:
++ /* %rdi indirection_page
++ * %rsi page_list
++ * %rdx start address
++ */
++
++ /* map the control page at its virtual address */
++
++ movq $0x0000ff8000000000, %r10 /* mask */
++ mov $(39 - 3), %cl /* bits to shift */
++ movq PTR(VA_CONTROL_PAGE)(%rsi), %r11 /* address to map */
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PGD)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PUD_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PUD_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PMD_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PMD_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PTE_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PTE_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ /* identity map the control page at its physical address */
++
++ movq $0x0000ff8000000000, %r10 /* mask */
++ mov $(39 - 3), %cl /* bits to shift */
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r11 /* address to map */
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PGD)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PUD_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PUD_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PMD_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PMD_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PTE_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PTE_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
+ relocate_new_kernel:
+- /* %rdi page_list
+- * %rsi reboot_code_buffer
++ /* %rdi indirection_page
++ * %rsi page_list
+ * %rdx start address
+- * %rcx page_table
+- * %r8 arg5
+- * %r9 arg6
+ */
+
+ /* zero out flags, and disable interrupts */
+ pushq $0
+ popfq
+
+- /* set a new stack at the bottom of our page... */
+- lea 4096(%rsi), %rsp
++ /* get physical address of control page now */
++ /* this is impossible after page table switch */
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++
++ /* get physical address of page table now too */
++ movq PTR(PA_TABLE_PAGE)(%rsi), %rcx
++
++ /* switch to new set of page tables */
++ movq PTR(PA_PGD)(%rsi), %r9
++ movq %r9, %cr3
++
++ /* setup a new stack at the end of the physical control page */
++ lea 4096(%r8), %rsp
++
++ /* jump to identity mapped page */
++ addq $(identity_mapped - relocate_kernel), %r8
++ pushq %r8
++ ret
+
+- /* store the parameters back on the stack */
+- pushq %rdx /* store the start address */
++identity_mapped:
++ /* store the start address on the stack */
++ pushq %rdx
+
+ /* Set cr0 to a known state:
+ * 31 1 == Paging enabled
+@@ -136,8 +274,3 @@ relocate_new_kernel:
+ xorq %r15, %r15
+
+ ret
+-relocate_new_kernel_end:
+-
+- .globl relocate_new_kernel_size
+-relocate_new_kernel_size:
+- .quad relocate_new_kernel_end - relocate_new_kernel
+--- a/include/asm-x86_64/kexec.h
++++ b/include/asm-x86_64/kexec.h
+@@ -1,6 +1,27 @@
+ #ifndef _X86_64_KEXEC_H
+ #define _X86_64_KEXEC_H
+
++#define PA_CONTROL_PAGE 0
++#define VA_CONTROL_PAGE 1
++#define PA_PGD 2
++#define VA_PGD 3
++#define PA_PUD_0 4
++#define VA_PUD_0 5
++#define PA_PMD_0 6
++#define VA_PMD_0 7
++#define PA_PTE_0 8
++#define VA_PTE_0 9
++#define PA_PUD_1 10
++#define VA_PUD_1 11
++#define PA_PMD_1 12
++#define VA_PMD_1 13
++#define PA_PTE_1 14
++#define VA_PTE_1 15
++#define PA_TABLE_PAGE 16
++#define PAGES_NR 17
++
++#ifndef __ASSEMBLY__
++
+ #include <linux/string.h>
+
+ #include <asm/page.h>
+@@ -64,4 +85,12 @@ static inline void crash_setup_regs(stru
+ newregs->rip = (unsigned long)current_text_addr();
+ }
+ }
++
++NORET_TYPE void
++relocate_kernel(unsigned long indirection_page,
++ unsigned long page_list,
++ unsigned long start_address) ATTRIB_NORET;
++
++#endif /* __ASSEMBLY__ */
++
+ #endif /* _X86_64_KEXEC_H */
--- /dev/null
+++ work/patches/linux-2.6.16.29/linux-2.6.16.29-crash-xen-x86_64.patch 2006-10-30 15:48:59.000000000 +0900
@@ -0,0 +1,35 @@
+--- 0001/arch/x86_64/kernel/crash.c
++++ work/arch/x86_64/kernel/crash.c 2006-10-26 15:44:27.000000000 +0900
+@@ -92,6 +92,7 @@ static void crash_save_self(struct pt_re
+ crash_save_this_cpu(regs, cpu);
+ }
+
++#ifndef CONFIG_XEN
+ #ifdef CONFIG_SMP
+ static atomic_t waiting_for_crash_ipi;
+
+@@ -156,6 +157,7 @@ static void nmi_shootdown_cpus(void)
+ /* There are no cpus to shootdown */
+ }
+ #endif
++#endif /* CONFIG_XEN */
+
+ void machine_crash_shutdown(struct pt_regs *regs)
+ {
+@@ -173,6 +175,8 @@ void machine_crash_shutdown(struct pt_re
+
+ /* Make a note of crashing cpu. Will be used in NMI callback.*/
+ crashing_cpu = smp_processor_id();
++
++#ifndef CONFIG_XEN
+ nmi_shootdown_cpus();
+
+ if(cpu_has_apic)
+@@ -181,6 +185,6 @@ void machine_crash_shutdown(struct pt_re
+ #if defined(CONFIG_X86_IO_APIC)
+ disable_IO_APIC();
+ #endif
+-
++#endif /* CONFIG_XEN */
+ crash_save_self(regs);
+ }
--- /dev/null
+++ work/patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-move_segment_code-x86_64.patch 2006-10-30 15:48:59.000000000 +0900
@@ -0,0 +1,161 @@
+kexec: Move asm segment handling code to the assembly file (x86_64)
+
+This patch moves the idt, gdt, and segment handling code from machine_kexec.c
+to relocate_kernel.S. The main reason behind this move is to avoid code
+duplication in the Xen hypervisor. With this patch all code required to kexec
+is put on the control page.
+
+On top of that this patch also counts as a cleanup - I think it is much
+nicer to write assembly directly in assembly files than wrap inline assembly
+in C functions for no apparent reason.
+
+Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
+---
+
+ Applies to 2.6.19-rc1.
+
+ machine_kexec.c | 58 -----------------------------------------------------
+ relocate_kernel.S | 50 +++++++++++++++++++++++++++++++++++++++++----
+ 2 files changed, 45 insertions(+), 63 deletions(-)
+
+--- 0002/arch/x86_64/kernel/machine_kexec.c
++++ work/arch/x86_64/kernel/machine_kexec.c 2006-10-05 16:15:49.000000000 +0900
+@@ -112,47 +112,6 @@ static int init_pgtable(struct kimage *i
+ return init_level4_page(image, level4p, 0, end_pfn << PAGE_SHIFT);
+ }
+
+-static void set_idt(void *newidt, u16 limit)
+-{
+- struct desc_ptr curidt;
+-
+- /* x86-64 supports unaliged loads & stores */
+- curidt.size = limit;
+- curidt.address = (unsigned long)newidt;
+-
+- __asm__ __volatile__ (
+- "lidtq %0\n"
+- : : "m" (curidt)
+- );
+-};
+-
+-
+-static void set_gdt(void *newgdt, u16 limit)
+-{
+- struct desc_ptr curgdt;
+-
+- /* x86-64 supports unaligned loads & stores */
+- curgdt.size = limit;
+- curgdt.address = (unsigned long)newgdt;
+-
+- __asm__ __volatile__ (
+- "lgdtq %0\n"
+- : : "m" (curgdt)
+- );
+-};
+-
+-static void load_segments(void)
+-{
+- __asm__ __volatile__ (
+- "\tmovl %0,%%ds\n"
+- "\tmovl %0,%%es\n"
+- "\tmovl %0,%%ss\n"
+- "\tmovl %0,%%fs\n"
+- "\tmovl %0,%%gs\n"
+- : : "a" (__KERNEL_DS) : "memory"
+- );
+-}
+-
+ int machine_kexec_prepare(struct kimage *image)
+ {
+ unsigned long start_pgtable;
+@@ -209,23 +168,6 @@ NORET_TYPE void machine_kexec(struct kim
+ page_list[PA_TABLE_PAGE] =
+ (unsigned long)__pa(page_address(image->control_code_page));
+
+- /* The segment registers are funny things, they have both a
+- * visible and an invisible part. Whenever the visible part is
+- * set to a specific selector, the invisible part is loaded
+- * with from a table in memory. At no other time is the
+- * descriptor table in memory accessed.
+- *
+- * I take advantage of this here by force loading the
+- * segments, before I zap the gdt with an invalid value.
+- */
+- load_segments();
+- /* The gdt & idt are now invalid.
+- * If you want to load them you must set up your own idt & gdt.
+- */
+- set_gdt(phys_to_virt(0),0);
+- set_idt(phys_to_virt(0),0);
+-
+- /* now call it */
+ relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
+ image->start);
+ }
+--- 0002/arch/x86_64/kernel/relocate_kernel.S
++++ work/arch/x86_64/kernel/relocate_kernel.S 2006-10-05 16:18:07.000000000 +0900
+@@ -159,13 +159,39 @@ relocate_new_kernel:
+ movq PTR(PA_PGD)(%rsi), %r9
+ movq %r9, %cr3
+
++ /* setup idt */
++ movq %r8, %rax
++ addq $(idt_80 - relocate_kernel), %rax
++ lidtq (%rax)
++
++ /* setup gdt */
++ movq %r8, %rax
++ addq $(gdt - relocate_kernel), %rax
++ movq %r8, %r9
++ addq $((gdt_80 - relocate_kernel) + 2), %r9
++ movq %rax, (%r9)
++
++ movq %r8, %rax
++ addq $(gdt_80 - relocate_kernel), %rax
++ lgdtq (%rax)
++
++ /* setup data segment registers */
++ xorl %eax, %eax
++ movl %eax, %ds
++ movl %eax, %es
++ movl %eax, %fs
++ movl %eax, %gs
++ movl %eax, %ss
++
+ /* setup a new stack at the end of the physical control page */
+ lea 4096(%r8), %rsp
+
+- /* jump to identity mapped page */
+- addq $(identity_mapped - relocate_kernel), %r8
+- pushq %r8
+- ret
++ /* load new code segment and jump to identity mapped page */
++ movq %r8, %rax
++ addq $(identity_mapped - relocate_kernel), %rax
++ pushq $(gdt_cs - gdt)
++ pushq %rax
++ lretq
+
+ identity_mapped:
+ /* store the start address on the stack */
+@@ -272,5 +298,19 @@ identity_mapped:
+ xorq %r13, %r13
+ xorq %r14, %r14
+ xorq %r15, %r15
+-
+ ret
++
++ .align 16
++gdt:
++ .quad 0x0000000000000000 /* NULL descriptor */
++gdt_cs:
++ .quad 0x00af9a000000ffff
++gdt_end:
++
++gdt_80:
++ .word gdt_end - gdt - 1 /* limit */
++ .quad 0 /* base - filled in by code above */
++
++idt_80:
++ .word 0 /* limit */
++ .quad 0 /* base */
--- /dev/null
+++ work/patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-xen-x86_64.patch 2006-10-30 15:48:59.000000000 +0900
@@ -0,0 +1,162 @@
+--- 0006/arch/x86_64/kernel/machine_kexec.c
++++ work/arch/x86_64/kernel/machine_kexec.c 2006-10-06 15:36:16.000000000 +0900
+@@ -24,6 +24,104 @@ static u64 kexec_pud1[512] PAGE_ALIGNED;
+ static u64 kexec_pmd1[512] PAGE_ALIGNED;
+ static u64 kexec_pte1[512] PAGE_ALIGNED;
+
++#ifdef CONFIG_XEN
++
++/* In the case of Xen, override hypervisor functions to be able to create
++ * a regular identity mapping page table...
++ */
++
++#include <xen/interface/kexec.h>
++#include <xen/interface/memory.h>
++
++#define x__pmd(x) ((pmd_t) { (x) } )
++#define x__pud(x) ((pud_t) { (x) } )
++#define x__pgd(x) ((pgd_t) { (x) } )
++
++#define x_pmd_val(x) ((x).pmd)
++#define x_pud_val(x) ((x).pud)
++#define x_pgd_val(x) ((x).pgd)
++
++static inline void x_set_pmd(pmd_t *dst, pmd_t val)
++{
++ x_pmd_val(*dst) = x_pmd_val(val);
++}
++
++static inline void x_set_pud(pud_t *dst, pud_t val)
++{
++ x_pud_val(*dst) = phys_to_machine(x_pud_val(val));
++}
++
++static inline void x_pud_clear (pud_t *pud)
++{
++ x_pud_val(*pud) = 0;
++}
++
++static inline void x_set_pgd(pgd_t *dst, pgd_t val)
++{
++ x_pgd_val(*dst) = phys_to_machine(x_pgd_val(val));
++}
++
++static inline void x_pgd_clear (pgd_t * pgd)
++{
++ x_pgd_val(*pgd) = 0;
++}
++
++#define X__PAGE_KERNEL_LARGE_EXEC \
++ _PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_PSE
++#define X_KERNPG_TABLE _PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY
++
++#define __ma(x) (pfn_to_mfn(__pa((x)) >> PAGE_SHIFT) << PAGE_SHIFT)
++
++#if PAGES_NR > KEXEC_XEN_NO_PAGES
++#error PAGES_NR is greater than KEXEC_XEN_NO_PAGES - Xen support will break
++#endif
++
++#if PA_CONTROL_PAGE != 0
++#error PA_CONTROL_PAGE is non zero - Xen support will break
++#endif
++
++void machine_kexec_setup_load_arg(xen_kexec_image_t *xki, struct kimage *image)
++{
++ void *control_page;
++ void *table_page;
++
++ memset(xki->page_list, 0, sizeof(xki->page_list));
++
++ control_page = page_address(image->control_code_page) + PAGE_SIZE;
++ memcpy(control_page, relocate_kernel, PAGE_SIZE);
++
++ table_page = page_address(image->control_code_page);
++
++ xki->page_list[PA_CONTROL_PAGE] = __ma(control_page);
++ xki->page_list[PA_TABLE_PAGE] = __ma(table_page);
++
++ xki->page_list[PA_PGD] = __ma(kexec_pgd);
++ xki->page_list[PA_PUD_0] = __ma(kexec_pud0);
++ xki->page_list[PA_PUD_1] = __ma(kexec_pud1);
++ xki->page_list[PA_PMD_0] = __ma(kexec_pmd0);
++ xki->page_list[PA_PMD_1] = __ma(kexec_pmd1);
++ xki->page_list[PA_PTE_0] = __ma(kexec_pte0);
++ xki->page_list[PA_PTE_1] = __ma(kexec_pte1);
++}
++
++#else /* CONFIG_XEN */
++
++#define x__pmd(x) __pmd(x)
++#define x__pud(x) __pud(x)
++#define x__pgd(x) __pgd(x)
++
++#define x_set_pmd(x, y) set_pmd(x, y)
++#define x_set_pud(x, y) set_pud(x, y)
++#define x_set_pgd(x, y) set_pgd(x, y)
++
++#define x_pud_clear(x) pud_clear(x)
++#define x_pgd_clear(x) pgd_clear(x)
++
++#define X__PAGE_KERNEL_LARGE_EXEC __PAGE_KERNEL_LARGE_EXEC
++#define X_KERNPG_TABLE _KERNPG_TABLE
++
++#endif /* CONFIG_XEN */
++
+ static void init_level2_page(pmd_t *level2p, unsigned long addr)
+ {
+ unsigned long end_addr;
+@@ -31,7 +129,7 @@ static void init_level2_page(pmd_t *leve
+ addr &= PAGE_MASK;
+ end_addr = addr + PUD_SIZE;
+ while (addr < end_addr) {
+- set_pmd(level2p++, __pmd(addr | __PAGE_KERNEL_LARGE_EXEC));
++ x_set_pmd(level2p++, x__pmd(addr | X__PAGE_KERNEL_LARGE_EXEC));
+ addr += PMD_SIZE;
+ }
+ }
+@@ -56,12 +154,12 @@ static int init_level3_page(struct kimag
+ }
+ level2p = (pmd_t *)page_address(page);
+ init_level2_page(level2p, addr);
+- set_pud(level3p++, __pud(__pa(level2p) | _KERNPG_TABLE));
++ x_set_pud(level3p++, x__pud(__pa(level2p) | X_KERNPG_TABLE));
+ addr += PUD_SIZE;
+ }
+ /* clear the unused entries */
+ while (addr < end_addr) {
+- pud_clear(level3p++);
++ x_pud_clear(level3p++);
+ addr += PUD_SIZE;
+ }
+ out:
+@@ -92,12 +190,12 @@ static int init_level4_page(struct kimag
+ if (result) {
+ goto out;
+ }
+- set_pgd(level4p++, __pgd(__pa(level3p) | _KERNPG_TABLE));
++ x_set_pgd(level4p++, x__pgd(__pa(level3p) | X_KERNPG_TABLE));
+ addr += PGDIR_SIZE;
+ }
+ /* clear the unused entries */
+ while (addr < end_addr) {
+- pgd_clear(level4p++);
++ x_pgd_clear(level4p++);
+ addr += PGDIR_SIZE;
+ }
+ out:
+@@ -108,8 +206,14 @@ out:
+ static int init_pgtable(struct kimage *image, unsigned long start_pgtable)
+ {
+ pgd_t *level4p;
++ unsigned long x_end_pfn = end_pfn;
++
++#ifdef CONFIG_XEN
++ x_end_pfn = HYPERVISOR_memory_op(XENMEM_maximum_ram_page, NULL);
++#endif
++
+ level4p = (pgd_t *)__va(start_pgtable);
+- return init_level4_page(image, level4p, 0, end_pfn << PAGE_SHIFT);
++ return init_level4_page(image, level4p, 0, x_end_pfn << PAGE_SHIFT);
+ }
+
+ int machine_kexec_prepare(struct kimage *image)
--- 0005/patches/linux-2.6.16.29/series
+++ work/patches/linux-2.6.16.29/series 2006-10-30 15:48:58.000000000 +0900
@@ -5,6 +5,10 @@ git-3566561bfadffcb5dbc85d576be80c0dbf2c
linux-2.6.19-rc1-kexec-move_segment_code-i386.patch
linux-2.6.19-rc1-kexec-xen-i386.patch
linux-2.6.16.29-crash-xen-i386.patch
+git-4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f.patch
+linux-2.6.19-rc1-kexec-move_segment_code-x86_64.patch
+linux-2.6.19-rc1-kexec-xen-x86_64.patch
+linux-2.6.16.29-crash-xen-x86_64.patch
blktap-aio-16_03_06.patch
device_bind.patch
fix-hz-suspend.patch
--- 0001/xen/arch/x86/x86_64/entry.S
+++ work/xen/arch/x86/x86_64/entry.S 2006-10-30 15:48:58.000000000 +0900
@@ -573,6 +573,7 @@ ENTRY(hypercall_table)
.quad do_hvm_op
.quad do_sysctl /* 35 */
.quad do_domctl
+ .quad do_kexec_op
.rept NR_hypercalls-((.-hypercall_table)/8)
.quad do_ni_hypercall
.endr
@@ -615,6 +616,7 @@ ENTRY(hypercall_args_table)
.byte 2 /* do_hvm_op */
.byte 1 /* do_sysctl */ /* 35 */
.byte 1 /* do_domctl */
+ .byte 2 /* do_kexec */
.rept NR_hypercalls-(.-hypercall_args_table)
.byte 0 /* do_ni_hypercall */
.endr
--- 0004/xen/include/asm-x86/x86_64/elf.h
+++ work/xen/include/asm-x86/x86_64/elf.h 2006-10-30 15:48:58.000000000 +0900
@@ -1,15 +1,67 @@
#ifndef __X86_64_ELF_H__
#define __X86_64_ELF_H__
-#include <xen/lib.h> /* for printk() used in stub */
+#include <asm/processor.h>
typedef struct {
- unsigned long dummy;
+ unsigned long r15;
+ unsigned long r14;
+ unsigned long r13;
+ unsigned long r12;
+ unsigned long rbp;
+ unsigned long rbx;
+ unsigned long r11;
+ unsigned long r10;
+ unsigned long r9;
+ unsigned long r8;
+ unsigned long rax;
+ unsigned long rcx;
+ unsigned long rdx;
+ unsigned long rsi;
+ unsigned long rdi;
+ unsigned long orig_rax;
+ unsigned long rip;
+ unsigned long cs;
+ unsigned long eflags;
+ unsigned long rsp;
+ unsigned long ss;
+ unsigned long thread_fs;
+ unsigned long thread_gs;
+ unsigned long ds;
+ unsigned long es;
+ unsigned long fs;
+ unsigned long gs;
} ELF_Gregset;
extern inline void elf_core_save_regs(ELF_Gregset *dst)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ asm volatile("movq %%r15,%0" : "=m"(dst->r15));
+ asm volatile("movq %%r14,%0" : "=m"(dst->r14));
+ asm volatile("movq %%r13,%0" : "=m"(dst->r13));
+ asm volatile("movq %%r12,%0" : "=m"(dst->r12));
+ asm volatile("movq %%rbp,%0" : "=m"(dst->rbp));
+ asm volatile("movq %%rbx,%0" : "=m"(dst->rbx));
+ asm volatile("movq %%r11,%0" : "=m"(dst->r11));
+ asm volatile("movq %%r10,%0" : "=m"(dst->r10));
+ asm volatile("movq %%r9,%0" : "=m"(dst->r9));
+ asm volatile("movq %%r8,%0" : "=m"(dst->r8));
+ asm volatile("movq %%rax,%0" : "=m"(dst->rax));
+ asm volatile("movq %%rcx,%0" : "=m"(dst->rcx));
+ asm volatile("movq %%rdx,%0" : "=m"(dst->rdx));
+ asm volatile("movq %%rsi,%0" : "=m"(dst->rsi));
+ asm volatile("movq %%rdi,%0" : "=m"(dst->rdi));
+ /* orig_rax not filled in for now */
+ dst->rip = (unsigned long)current_text_addr();
+ asm volatile("movl %%cs, %%eax;" :"=a"(dst->cs));
+ asm volatile("pushfq; popq %0" :"=m"(dst->eflags));
+ asm volatile("movq %%rsp,%0" : "=m"(dst->rsp));
+ asm volatile("movl %%ss, %%eax;" :"=a"(dst->ss));
+ /* thread_fs not filled in for now */
+ /* thread_gs not filled in for now */
+ asm volatile("movl %%ds, %%eax;" :"=a"(dst->ds));
+ asm volatile("movl %%es, %%eax;" :"=a"(dst->es));
+ asm volatile("movl %%fs, %%eax;" :"=a"(dst->fs));
+ asm volatile("movl %%gs, %%eax;" :"=a"(dst->gs));
}
#endif /* __X86_64_ELF_H__ */
--- 0004/xen/include/asm-x86/x86_64/kexec.h
+++ work/xen/include/asm-x86/x86_64/kexec.h 2006-10-30 15:48:58.000000000 +0900
@@ -1,14 +1,29 @@
+/******************************************************************************
+ * kexec.h
+ *
+ * Based heavily on machine_kexec.c and kexec.h from Linux 2.6.19-rc1
+ *
+ */
+
#ifndef __X86_64_KEXEC_H__
#define __X86_64_KEXEC_H__
-
-#include <xen/lib.h> /* for printk() used in stub */
+
#include <xen/types.h>
-#include <public/xen.h>
#include <xen/kexec.h>
+#include <asm/fixmap.h>
+
+typedef void (*relocate_new_kernel_t)(
+ unsigned long indirection_page,
+ unsigned long page_list,
+ unsigned long start_address);
static inline void machine_kexec(xen_kexec_image_t *image)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ relocate_new_kernel_t rnk;
+
+ rnk = (relocate_new_kernel_t) image->page_list[1];
+ (*rnk)(image->indirection_page, (unsigned long)image->page_list,
+ image->start_address);
}
#endif /* __X86_64_KEXEC_H__ */
^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH 04/04] Kexec / Kdump: x86_64 specific code
2006-10-23 9:05 [PATCH 00/04] Kexec / Kdump: Release 20061023 (xen-unstable-11856) Magnus Damm
@ 2006-10-23 9:05 ` Magnus Damm
0 siblings, 0 replies; 30+ messages in thread
From: Magnus Damm @ 2006-10-23 9:05 UTC (permalink / raw)
To: Keir Fraser
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, magnus.damm,
Isaku Yamahata, Magnus Damm, Horms
[PATCH 04/04] Kexec / Kdump: x86_64 specific code
This patch contains the x86_64 implementation of Kexec / Kdump for Xen.
Signed-Off-By: Magnus Damm <magnus@valinux.co.jp>
---
Applies on top of xen-unstable-11856.
buildconfigs/linux-defconfig_xen_x86_64 | 1
linux-2.6-xen-sparse/arch/x86_64/Kconfig | 2
linux-2.6-xen-sparse/arch/x86_64/kernel/Makefile | 2
linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c | 27
linux-2.6-xen-sparse/include/asm-x86_64/kexec-xen.h | 64 +
linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h | 7
linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h | 2
patches/linux-2.6.16.29/git-4b...1f.patch | 375 ++++
patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec..code-x86_64.patch | 161 ++
patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-xen-x86_64.patch | 162 ++
patches/linux-2.6.16.29/series | 3
xen/arch/x86/x86_64/entry.S | 2
xen/include/asm-x86/x86_64/elf.h | 49 +
xen/include/asm-x86/x86_64/kexec.h | 60 +
14 files changed, 903 insertions(+), 14 deletions(-)
--- 0002/buildconfigs/linux-defconfig_xen_x86_64
+++ work/buildconfigs/linux-defconfig_xen_x86_64 2006-10-23 11:36:17.000000000 +0900
@@ -138,6 +138,7 @@ CONFIG_SWIOTLB=y
CONFIG_PHYSICAL_START=0x100000
CONFIG_SECCOMP=y
CONFIG_HZ_100=y
+CONFIG_KEXEC=y
# CONFIG_HZ_250 is not set
# CONFIG_HZ_1000 is not set
CONFIG_HZ=100
--- 0001/linux-2.6-xen-sparse/arch/x86_64/Kconfig
+++ work/linux-2.6-xen-sparse/arch/x86_64/Kconfig 2006-10-23 11:36:17.000000000 +0900
@@ -435,7 +435,7 @@ config X86_MCE_AMD
config KEXEC
bool "kexec system call (EXPERIMENTAL)"
- depends on EXPERIMENTAL && !X86_64_XEN
+ depends on EXPERIMENTAL && !XEN_UNPRIVILEGED_GUEST
help
kexec is a system call that implements the ability to shutdown your
current kernel, and to start another kernel. It is like a reboot
--- 0001/linux-2.6-xen-sparse/arch/x86_64/kernel/Makefile
+++ work/linux-2.6-xen-sparse/arch/x86_64/kernel/Makefile 2006-10-23 11:36:17.000000000 +0900
@@ -59,7 +59,7 @@ pci-dma-y += ../../i386/kernel/pci-dma
microcode-$(subst m,y,$(CONFIG_MICROCODE)) := ../../i386/kernel/microcode-xen.o
quirks-y := ../../i386/kernel/quirks-xen.o
-n-obj-xen := i8259.o reboot.o i8237.o smpboot.o trampoline.o
+n-obj-xen := i8259.o reboot.o i8237.o smpboot.o trampoline.o crash.o
include $(srctree)/scripts/Makefile.xen
--- 0001/linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c
+++ work/linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c 2006-10-23 11:36:17.000000000 +0900
@@ -80,6 +80,10 @@
#include <asm/mach-xen/setup_arch_post.h>
#include <xen/interface/memory.h>
+#ifdef CONFIG_XEN
+#include <xen/interface/kexec.h>
+#endif
+
extern unsigned long start_pfn;
extern struct edid_info edid_info;
@@ -450,6 +454,7 @@ static __init void parse_cmdline_early (
* after a kernel panic.
*/
else if (!memcmp(from, "crashkernel=", 12)) {
+#ifndef CONFIG_XEN
unsigned long size, base;
size = memparse(from+12, &from);
if (*from == '@') {
@@ -460,6 +465,10 @@ static __init void parse_cmdline_early (
crashk_res.start = base;
crashk_res.end = base + size - 1;
}
+#else
+ printk("Ignoring crashkernel command line, "
+ "parameter will be supplied by xen\n");
+#endif
}
#endif
@@ -812,10 +821,23 @@ void __init setup_arch(char **cmdline_p)
#endif
#endif /* !CONFIG_XEN */
#ifdef CONFIG_KEXEC
+#ifndef CONFIG_XEN
if (crashk_res.start != crashk_res.end) {
reserve_bootmem(crashk_res.start,
crashk_res.end - crashk_res.start + 1);
}
+#else
+ {
+ xen_kexec_reserve_t reservation;
+ BUG_ON(HYPERVISOR_kexec_op(KEXEC_CMD_kexec_reserve,
+ &reservation));
+ if (reservation.size) {
+ crashk_res.start = reservation.start;
+ crashk_res.end = reservation.start +
+ reservation.size - 1;
+ }
+ }
+#endif
#endif
paging_init();
@@ -954,6 +976,11 @@ void __init setup_arch(char **cmdline_p)
iommu_hole_init();
#endif
+#ifdef CONFIG_KEXEC
+ if (crashk_res.start != crashk_res.end)
+ request_resource(&ioport_resource, &crashk_res);
+#endif
+
#ifdef CONFIG_XEN
{
struct physdev_set_iopl set_iopl;
--- /dev/null
+++ work/linux-2.6-xen-sparse/include/asm-x86_64/kexec-xen.h 2006-10-23 11:36:18.000000000 +0900
@@ -0,0 +1,64 @@
+/*
+ * include/asm-x86_64/kexec-xen.h
+ *
+ * Created By: Horms <horms@verge.net.au>
+ */
+
+#ifndef _X86_64_KEXEC_XEN_H
+#define _X86_64_KEXEC_XEN_H
+
+#include <asm/ptrace.h>
+#include <asm/types.h>
+#include <xen/interface/arch-x86_64.h>
+
+static inline void crash_translate_regs(struct pt_regs *linux_regs,
+ struct cpu_user_regs *xen_regs)
+{
+ xen_regs->r15 = linux_regs->r15;
+ xen_regs->r14 = linux_regs->r14;
+ xen_regs->r13 = linux_regs->r13;
+ xen_regs->r12 = linux_regs->r12;
+ xen_regs->rbp = linux_regs->rbp;
+ xen_regs->rbx = linux_regs->rbx;
+ xen_regs->r11 = linux_regs->r11;
+ xen_regs->r10 = linux_regs->r10;
+ xen_regs->r9 = linux_regs->r9;
+ xen_regs->r8 = linux_regs->r8;
+ xen_regs->rax = linux_regs->rax;
+ xen_regs->rcx = linux_regs->rcx;
+ xen_regs->rdx = linux_regs->rdx;
+ xen_regs->rsi = linux_regs->rsi;
+ xen_regs->rdi = linux_regs->rdi;
+ xen_regs->rip = linux_regs->rip;
+ xen_regs->cs = linux_regs->cs;
+ xen_regs->rflags = linux_regs->eflags;
+ xen_regs->rsp = linux_regs->rsp;
+ xen_regs->ss = linux_regs->ss;
+}
+
+/* Kexec needs to know about the actual physical addresss.
+ * But in xen, on some architectures, a physical address is a
+ * pseudo-physical addresss. */
+#ifdef CONFIG_XEN
+#define kexec_page_to_pfn(page) pfn_to_mfn(page_to_pfn(page))
+#define kexec_pfn_to_page(pfn) pfn_to_page(mfn_to_pfn(pfn))
+#define kexec_virt_to_phys(addr) virt_to_machine(addr)
+#define kexec_phys_to_virt(addr) phys_to_virt(machine_to_phys(addr))
+#else
+#define kexec_page_to_pfn(page) page_to_pfn(page)
+#define kexec_pfn_to_page(pfn) pfn_to_page(pfn)
+#define kexec_virt_to_phys(addr) virt_to_phys(addr)
+#define kexec_phys_to_virt(addr) phys_to_virt(addr)
+#endif
+
+#endif /* _X86_64_KEXEC_XEN_H */
+
+/*
+ * Local variables:
+ * c-file-style: "linux"
+ * indent-tabs-mode: t
+ * c-indent-level: 8
+ * c-basic-offset: 8
+ * tab-width: 8
+ * End:
+ */
--- 0001/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h
+++ work/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h 2006-10-23 11:36:17.000000000 +0900
@@ -386,4 +386,11 @@ HYPERVISOR_xenoprof_op(
return _hypercall2(int, xenoprof_op, op, arg);
}
+static inline int
+HYPERVISOR_kexec_op(
+ unsigned long op, void *args)
+{
+ return _hypercall2(int, kexec_op, op, args);
+}
+
#endif /* __HYPERCALL_H__ */
--- 0001/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h
+++ work/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h 2006-10-23 11:36:17.000000000 +0900
@@ -90,6 +90,8 @@ extern unsigned long profile_pc(struct p
#define profile_pc(regs) instruction_pointer(regs)
#endif
+#include <linux/compiler.h>
+
void signal_fault(struct pt_regs *regs, void __user *frame, char *where);
struct task_struct;
--- /dev/null
+++ work/patches/linux-2.6.16.29/git-4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f.patch 2006-10-23 11:36:18.000000000 +0900
@@ -0,0 +1,375 @@
+From: Magnus Damm <magnus@valinux.co.jp>
+Date: Tue, 26 Sep 2006 08:52:38 +0000 (+0200)
+Subject: [PATCH] Avoid overwriting the current pgd (V4, x86_64)
+X-Git-Tag: v2.6.19-rc1
+X-Git-Url: http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f
+
+[PATCH] Avoid overwriting the current pgd (V4, x86_64)
+
+kexec: Avoid overwriting the current pgd (V4, x86_64)
+
+This patch upgrades the x86_64-specific kexec code to avoid overwriting the
+current pgd. Overwriting the current pgd is bad when CONFIG_CRASH_DUMP is used
+to start a secondary kernel that dumps the memory of the previous kernel.
+
+The code introduces a new set of page tables. These tables are used to provide
+an executable identity mapping without overwriting the current pgd.
+
+Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
+Signed-off-by: Andi Kleen <ak@suse.de>
+---
+
+--- a/arch/x86_64/kernel/machine_kexec.c
++++ b/arch/x86_64/kernel/machine_kexec.c
+@@ -15,6 +15,15 @@
+ #include <asm/mmu_context.h>
+ #include <asm/io.h>
+
++#define PAGE_ALIGNED __attribute__ ((__aligned__(PAGE_SIZE)))
++static u64 kexec_pgd[512] PAGE_ALIGNED;
++static u64 kexec_pud0[512] PAGE_ALIGNED;
++static u64 kexec_pmd0[512] PAGE_ALIGNED;
++static u64 kexec_pte0[512] PAGE_ALIGNED;
++static u64 kexec_pud1[512] PAGE_ALIGNED;
++static u64 kexec_pmd1[512] PAGE_ALIGNED;
++static u64 kexec_pte1[512] PAGE_ALIGNED;
++
+ static void init_level2_page(pmd_t *level2p, unsigned long addr)
+ {
+ unsigned long end_addr;
+@@ -144,32 +153,19 @@ static void load_segments(void)
+ );
+ }
+
+-typedef NORET_TYPE void (*relocate_new_kernel_t)(unsigned long indirection_page,
+- unsigned long control_code_buffer,
+- unsigned long start_address,
+- unsigned long pgtable) ATTRIB_NORET;
+-
+-extern const unsigned char relocate_new_kernel[];
+-extern const unsigned long relocate_new_kernel_size;
+-
+ int machine_kexec_prepare(struct kimage *image)
+ {
+- unsigned long start_pgtable, control_code_buffer;
++ unsigned long start_pgtable;
+ int result;
+
+ /* Calculate the offsets */
+ start_pgtable = page_to_pfn(image->control_code_page) << PAGE_SHIFT;
+- control_code_buffer = start_pgtable + PAGE_SIZE;
+
+ /* Setup the identity mapped 64bit page table */
+ result = init_pgtable(image, start_pgtable);
+ if (result)
+ return result;
+
+- /* Place the code in the reboot code buffer */
+- memcpy(__va(control_code_buffer), relocate_new_kernel,
+- relocate_new_kernel_size);
+-
+ return 0;
+ }
+
+@@ -184,28 +180,34 @@ void machine_kexec_cleanup(struct kimage
+ */
+ NORET_TYPE void machine_kexec(struct kimage *image)
+ {
+- unsigned long page_list;
+- unsigned long control_code_buffer;
+- unsigned long start_pgtable;
+- relocate_new_kernel_t rnk;
++ unsigned long page_list[PAGES_NR];
++ void *control_page;
+
+ /* Interrupts aren't acceptable while we reboot */
+ local_irq_disable();
+
+- /* Calculate the offsets */
+- page_list = image->head;
+- start_pgtable = page_to_pfn(image->control_code_page) << PAGE_SHIFT;
+- control_code_buffer = start_pgtable + PAGE_SIZE;
++ control_page = page_address(image->control_code_page) + PAGE_SIZE;
++ memcpy(control_page, relocate_kernel, PAGE_SIZE);
+
+- /* Set the low half of the page table to my identity mapped
+- * page table for kexec. Leave the high half pointing at the
+- * kernel pages. Don't bother to flush the global pages
+- * as that will happen when I fully switch to my identity mapped
+- * page table anyway.
+- */
+- memcpy(__va(read_cr3()), __va(start_pgtable), PAGE_SIZE/2);
+- __flush_tlb();
++ page_list[PA_CONTROL_PAGE] = __pa(control_page);
++ page_list[VA_CONTROL_PAGE] = (unsigned long)relocate_kernel;
++ page_list[PA_PGD] = __pa(kexec_pgd);
++ page_list[VA_PGD] = (unsigned long)kexec_pgd;
++ page_list[PA_PUD_0] = __pa(kexec_pud0);
++ page_list[VA_PUD_0] = (unsigned long)kexec_pud0;
++ page_list[PA_PMD_0] = __pa(kexec_pmd0);
++ page_list[VA_PMD_0] = (unsigned long)kexec_pmd0;
++ page_list[PA_PTE_0] = __pa(kexec_pte0);
++ page_list[VA_PTE_0] = (unsigned long)kexec_pte0;
++ page_list[PA_PUD_1] = __pa(kexec_pud1);
++ page_list[VA_PUD_1] = (unsigned long)kexec_pud1;
++ page_list[PA_PMD_1] = __pa(kexec_pmd1);
++ page_list[VA_PMD_1] = (unsigned long)kexec_pmd1;
++ page_list[PA_PTE_1] = __pa(kexec_pte1);
++ page_list[VA_PTE_1] = (unsigned long)kexec_pte1;
+
++ page_list[PA_TABLE_PAGE] =
++ (unsigned long)__pa(page_address(image->control_code_page));
+
+ /* The segment registers are funny things, they have both a
+ * visible and an invisible part. Whenever the visible part is
+@@ -222,9 +224,10 @@ NORET_TYPE void machine_kexec(struct kim
+ */
+ set_gdt(phys_to_virt(0),0);
+ set_idt(phys_to_virt(0),0);
++
+ /* now call it */
+- rnk = (relocate_new_kernel_t) control_code_buffer;
+- (*rnk)(page_list, control_code_buffer, image->start, start_pgtable);
++ relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
++ image->start);
+ }
+
+ /* crashkernel=size@addr specifies the location to reserve for
+--- a/arch/x86_64/kernel/relocate_kernel.S
++++ b/arch/x86_64/kernel/relocate_kernel.S
+@@ -7,31 +7,169 @@
+ */
+
+ #include <linux/linkage.h>
++#include <asm/page.h>
++#include <asm/kexec.h>
+
+- /*
+- * Must be relocatable PIC code callable as a C function, that once
+- * it starts can not use the previous processes stack.
+- */
+- .globl relocate_new_kernel
++/*
++ * Must be relocatable PIC code callable as a C function
++ */
++
++#define PTR(x) (x << 3)
++#define PAGE_ALIGNED (1 << PAGE_SHIFT)
++#define PAGE_ATTR 0x63 /* _PAGE_PRESENT|_PAGE_RW|_PAGE_ACCESSED|_PAGE_DIRTY */
++
++ .text
++ .align PAGE_ALIGNED
+ .code64
++ .globl relocate_kernel
++relocate_kernel:
++ /* %rdi indirection_page
++ * %rsi page_list
++ * %rdx start address
++ */
++
++ /* map the control page at its virtual address */
++
++ movq $0x0000ff8000000000, %r10 /* mask */
++ mov $(39 - 3), %cl /* bits to shift */
++ movq PTR(VA_CONTROL_PAGE)(%rsi), %r11 /* address to map */
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PGD)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PUD_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PUD_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PMD_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PMD_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PTE_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PTE_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ /* identity map the control page at its physical address */
++
++ movq $0x0000ff8000000000, %r10 /* mask */
++ mov $(39 - 3), %cl /* bits to shift */
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r11 /* address to map */
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PGD)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PUD_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PUD_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PMD_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PMD_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PTE_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PTE_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
+ relocate_new_kernel:
+- /* %rdi page_list
+- * %rsi reboot_code_buffer
++ /* %rdi indirection_page
++ * %rsi page_list
+ * %rdx start address
+- * %rcx page_table
+- * %r8 arg5
+- * %r9 arg6
+ */
+
+ /* zero out flags, and disable interrupts */
+ pushq $0
+ popfq
+
+- /* set a new stack at the bottom of our page... */
+- lea 4096(%rsi), %rsp
++ /* get physical address of control page now */
++ /* this is impossible after page table switch */
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++
++ /* get physical address of page table now too */
++ movq PTR(PA_TABLE_PAGE)(%rsi), %rcx
++
++ /* switch to new set of page tables */
++ movq PTR(PA_PGD)(%rsi), %r9
++ movq %r9, %cr3
++
++ /* setup a new stack at the end of the physical control page */
++ lea 4096(%r8), %rsp
++
++ /* jump to identity mapped page */
++ addq $(identity_mapped - relocate_kernel), %r8
++ pushq %r8
++ ret
+
+- /* store the parameters back on the stack */
+- pushq %rdx /* store the start address */
++identity_mapped:
++ /* store the start address on the stack */
++ pushq %rdx
+
+ /* Set cr0 to a known state:
+ * 31 1 == Paging enabled
+@@ -136,8 +274,3 @@ relocate_new_kernel:
+ xorq %r15, %r15
+
+ ret
+-relocate_new_kernel_end:
+-
+- .globl relocate_new_kernel_size
+-relocate_new_kernel_size:
+- .quad relocate_new_kernel_end - relocate_new_kernel
+--- a/include/asm-x86_64/kexec.h
++++ b/include/asm-x86_64/kexec.h
+@@ -1,6 +1,27 @@
+ #ifndef _X86_64_KEXEC_H
+ #define _X86_64_KEXEC_H
+
++#define PA_CONTROL_PAGE 0
++#define VA_CONTROL_PAGE 1
++#define PA_PGD 2
++#define VA_PGD 3
++#define PA_PUD_0 4
++#define VA_PUD_0 5
++#define PA_PMD_0 6
++#define VA_PMD_0 7
++#define PA_PTE_0 8
++#define VA_PTE_0 9
++#define PA_PUD_1 10
++#define VA_PUD_1 11
++#define PA_PMD_1 12
++#define VA_PMD_1 13
++#define PA_PTE_1 14
++#define VA_PTE_1 15
++#define PA_TABLE_PAGE 16
++#define PAGES_NR 17
++
++#ifndef __ASSEMBLY__
++
+ #include <linux/string.h>
+
+ #include <asm/page.h>
+@@ -64,4 +85,12 @@ static inline void crash_setup_regs(stru
+ newregs->rip = (unsigned long)current_text_addr();
+ }
+ }
++
++NORET_TYPE void
++relocate_kernel(unsigned long indirection_page,
++ unsigned long page_list,
++ unsigned long start_address) ATTRIB_NORET;
++
++#endif /* __ASSEMBLY__ */
++
+ #endif /* _X86_64_KEXEC_H */
--- /dev/null
+++ work/patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-move_segment_code-x86_64.patch 2006-10-23 11:36:18.000000000 +0900
@@ -0,0 +1,161 @@
+kexec: Move asm segment handling code to the assembly file (x86_64)
+
+This patch moves the idt, gdt, and segment handling code from machine_kexec.c
+to relocate_kernel.S. The main reason behind this move is to avoid code
+duplication in the Xen hypervisor. With this patch all code required to kexec
+is put on the control page.
+
+On top of that this patch also counts as a cleanup - I think it is much
+nicer to write assembly directly in assembly files than wrap inline assembly
+in C functions for no apparent reason.
+
+Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
+---
+
+ Applies to 2.6.19-rc1.
+
+ machine_kexec.c | 58 -----------------------------------------------------
+ relocate_kernel.S | 50 +++++++++++++++++++++++++++++++++++++++++----
+ 2 files changed, 45 insertions(+), 63 deletions(-)
+
+--- 0002/arch/x86_64/kernel/machine_kexec.c
++++ work/arch/x86_64/kernel/machine_kexec.c 2006-10-05 16:15:49.000000000 +0900
+@@ -112,47 +112,6 @@ static int init_pgtable(struct kimage *i
+ return init_level4_page(image, level4p, 0, end_pfn << PAGE_SHIFT);
+ }
+
+-static void set_idt(void *newidt, u16 limit)
+-{
+- struct desc_ptr curidt;
+-
+- /* x86-64 supports unaliged loads & stores */
+- curidt.size = limit;
+- curidt.address = (unsigned long)newidt;
+-
+- __asm__ __volatile__ (
+- "lidtq %0\n"
+- : : "m" (curidt)
+- );
+-};
+-
+-
+-static void set_gdt(void *newgdt, u16 limit)
+-{
+- struct desc_ptr curgdt;
+-
+- /* x86-64 supports unaligned loads & stores */
+- curgdt.size = limit;
+- curgdt.address = (unsigned long)newgdt;
+-
+- __asm__ __volatile__ (
+- "lgdtq %0\n"
+- : : "m" (curgdt)
+- );
+-};
+-
+-static void load_segments(void)
+-{
+- __asm__ __volatile__ (
+- "\tmovl %0,%%ds\n"
+- "\tmovl %0,%%es\n"
+- "\tmovl %0,%%ss\n"
+- "\tmovl %0,%%fs\n"
+- "\tmovl %0,%%gs\n"
+- : : "a" (__KERNEL_DS) : "memory"
+- );
+-}
+-
+ int machine_kexec_prepare(struct kimage *image)
+ {
+ unsigned long start_pgtable;
+@@ -209,23 +168,6 @@ NORET_TYPE void machine_kexec(struct kim
+ page_list[PA_TABLE_PAGE] =
+ (unsigned long)__pa(page_address(image->control_code_page));
+
+- /* The segment registers are funny things, they have both a
+- * visible and an invisible part. Whenever the visible part is
+- * set to a specific selector, the invisible part is loaded
+- * with from a table in memory. At no other time is the
+- * descriptor table in memory accessed.
+- *
+- * I take advantage of this here by force loading the
+- * segments, before I zap the gdt with an invalid value.
+- */
+- load_segments();
+- /* The gdt & idt are now invalid.
+- * If you want to load them you must set up your own idt & gdt.
+- */
+- set_gdt(phys_to_virt(0),0);
+- set_idt(phys_to_virt(0),0);
+-
+- /* now call it */
+ relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
+ image->start);
+ }
+--- 0002/arch/x86_64/kernel/relocate_kernel.S
++++ work/arch/x86_64/kernel/relocate_kernel.S 2006-10-05 16:18:07.000000000 +0900
+@@ -159,13 +159,39 @@ relocate_new_kernel:
+ movq PTR(PA_PGD)(%rsi), %r9
+ movq %r9, %cr3
+
++ /* setup idt */
++ movq %r8, %rax
++ addq $(idt_80 - relocate_kernel), %rax
++ lidtq (%rax)
++
++ /* setup gdt */
++ movq %r8, %rax
++ addq $(gdt - relocate_kernel), %rax
++ movq %r8, %r9
++ addq $((gdt_80 - relocate_kernel) + 2), %r9
++ movq %rax, (%r9)
++
++ movq %r8, %rax
++ addq $(gdt_80 - relocate_kernel), %rax
++ lgdtq (%rax)
++
++ /* setup data segment registers */
++ xorl %eax, %eax
++ movl %eax, %ds
++ movl %eax, %es
++ movl %eax, %fs
++ movl %eax, %gs
++ movl %eax, %ss
++
+ /* setup a new stack at the end of the physical control page */
+ lea 4096(%r8), %rsp
+
+- /* jump to identity mapped page */
+- addq $(identity_mapped - relocate_kernel), %r8
+- pushq %r8
+- ret
++ /* load new code segment and jump to identity mapped page */
++ movq %r8, %rax
++ addq $(identity_mapped - relocate_kernel), %rax
++ pushq $(gdt_cs - gdt)
++ pushq %rax
++ lretq
+
+ identity_mapped:
+ /* store the start address on the stack */
+@@ -272,5 +298,19 @@ identity_mapped:
+ xorq %r13, %r13
+ xorq %r14, %r14
+ xorq %r15, %r15
+-
+ ret
++
++ .align 16
++gdt:
++ .quad 0x0000000000000000 /* NULL descriptor */
++gdt_cs:
++ .quad 0x00af9a000000ffff
++gdt_end:
++
++gdt_80:
++ .word gdt_end - gdt - 1 /* limit */
++ .quad 0 /* base - filled in by code above */
++
++idt_80:
++ .word 0 /* limit */
++ .quad 0 /* base */
--- /dev/null
+++ work/patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-xen-x86_64.patch 2006-10-23 11:36:18.000000000 +0900
@@ -0,0 +1,162 @@
+--- 0006/arch/x86_64/kernel/machine_kexec.c
++++ work/arch/x86_64/kernel/machine_kexec.c 2006-10-06 15:36:16.000000000 +0900
+@@ -24,6 +24,104 @@ static u64 kexec_pud1[512] PAGE_ALIGNED;
+ static u64 kexec_pmd1[512] PAGE_ALIGNED;
+ static u64 kexec_pte1[512] PAGE_ALIGNED;
+
++#ifdef CONFIG_XEN
++
++/* In the case of Xen, override hypervisor functions to be able to create
++ * a regular identity mapping page table...
++ */
++
++#include <xen/interface/kexec.h>
++#include <xen/interface/memory.h>
++
++#define x__pmd(x) ((pmd_t) { (x) } )
++#define x__pud(x) ((pud_t) { (x) } )
++#define x__pgd(x) ((pgd_t) { (x) } )
++
++#define x_pmd_val(x) ((x).pmd)
++#define x_pud_val(x) ((x).pud)
++#define x_pgd_val(x) ((x).pgd)
++
++static inline void x_set_pmd(pmd_t *dst, pmd_t val)
++{
++ x_pmd_val(*dst) = x_pmd_val(val);
++}
++
++static inline void x_set_pud(pud_t *dst, pud_t val)
++{
++ x_pud_val(*dst) = phys_to_machine(x_pud_val(val));
++}
++
++static inline void x_pud_clear (pud_t *pud)
++{
++ x_pud_val(*pud) = 0;
++}
++
++static inline void x_set_pgd(pgd_t *dst, pgd_t val)
++{
++ x_pgd_val(*dst) = phys_to_machine(x_pgd_val(val));
++}
++
++static inline void x_pgd_clear (pgd_t * pgd)
++{
++ x_pgd_val(*pgd) = 0;
++}
++
++#define X__PAGE_KERNEL_LARGE_EXEC \
++ _PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_PSE
++#define X_KERNPG_TABLE _PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY
++
++#define __ma(x) (pfn_to_mfn(__pa((x)) >> PAGE_SHIFT) << PAGE_SHIFT)
++
++#if PAGES_NR > KEXEC_XEN_NO_PAGES
++#error PAGES_NR is greater than KEXEC_XEN_NO_PAGES - Xen support will break
++#endif
++
++#if PA_CONTROL_PAGE != 0
++#error PA_CONTROL_PAGE is non zero - Xen support will break
++#endif
++
++void machine_kexec_setup_load_arg(xen_kexec_image_t *xki, struct kimage *image)
++{
++ void *control_page;
++ void *table_page;
++
++ memset(xki->page_list, 0, sizeof(xki->page_list));
++
++ control_page = page_address(image->control_code_page) + PAGE_SIZE;
++ memcpy(control_page, relocate_kernel, PAGE_SIZE);
++
++ table_page = page_address(image->control_code_page);
++
++ xki->page_list[PA_CONTROL_PAGE] = __ma(control_page);
++ xki->page_list[PA_TABLE_PAGE] = __ma(table_page);
++
++ xki->page_list[PA_PGD] = __ma(kexec_pgd);
++ xki->page_list[PA_PUD_0] = __ma(kexec_pud0);
++ xki->page_list[PA_PUD_1] = __ma(kexec_pud1);
++ xki->page_list[PA_PMD_0] = __ma(kexec_pmd0);
++ xki->page_list[PA_PMD_1] = __ma(kexec_pmd1);
++ xki->page_list[PA_PTE_0] = __ma(kexec_pte0);
++ xki->page_list[PA_PTE_1] = __ma(kexec_pte1);
++}
++
++#else /* CONFIG_XEN */
++
++#define x__pmd(x) __pmd(x)
++#define x__pud(x) __pud(x)
++#define x__pgd(x) __pgd(x)
++
++#define x_set_pmd(x, y) set_pmd(x, y)
++#define x_set_pud(x, y) set_pud(x, y)
++#define x_set_pgd(x, y) set_pgd(x, y)
++
++#define x_pud_clear(x) pud_clear(x)
++#define x_pgd_clear(x) pgd_clear(x)
++
++#define X__PAGE_KERNEL_LARGE_EXEC __PAGE_KERNEL_LARGE_EXEC
++#define X_KERNPG_TABLE _KERNPG_TABLE
++
++#endif /* CONFIG_XEN */
++
+ static void init_level2_page(pmd_t *level2p, unsigned long addr)
+ {
+ unsigned long end_addr;
+@@ -31,7 +129,7 @@ static void init_level2_page(pmd_t *leve
+ addr &= PAGE_MASK;
+ end_addr = addr + PUD_SIZE;
+ while (addr < end_addr) {
+- set_pmd(level2p++, __pmd(addr | __PAGE_KERNEL_LARGE_EXEC));
++ x_set_pmd(level2p++, x__pmd(addr | X__PAGE_KERNEL_LARGE_EXEC));
+ addr += PMD_SIZE;
+ }
+ }
+@@ -56,12 +154,12 @@ static int init_level3_page(struct kimag
+ }
+ level2p = (pmd_t *)page_address(page);
+ init_level2_page(level2p, addr);
+- set_pud(level3p++, __pud(__pa(level2p) | _KERNPG_TABLE));
++ x_set_pud(level3p++, x__pud(__pa(level2p) | X_KERNPG_TABLE));
+ addr += PUD_SIZE;
+ }
+ /* clear the unused entries */
+ while (addr < end_addr) {
+- pud_clear(level3p++);
++ x_pud_clear(level3p++);
+ addr += PUD_SIZE;
+ }
+ out:
+@@ -92,12 +190,12 @@ static int init_level4_page(struct kimag
+ if (result) {
+ goto out;
+ }
+- set_pgd(level4p++, __pgd(__pa(level3p) | _KERNPG_TABLE));
++ x_set_pgd(level4p++, x__pgd(__pa(level3p) | X_KERNPG_TABLE));
+ addr += PGDIR_SIZE;
+ }
+ /* clear the unused entries */
+ while (addr < end_addr) {
+- pgd_clear(level4p++);
++ x_pgd_clear(level4p++);
+ addr += PGDIR_SIZE;
+ }
+ out:
+@@ -108,8 +206,14 @@ out:
+ static int init_pgtable(struct kimage *image, unsigned long start_pgtable)
+ {
+ pgd_t *level4p;
++ unsigned long x_end_pfn = end_pfn;
++
++#ifdef CONFIG_XEN
++ x_end_pfn = HYPERVISOR_memory_op(XENMEM_maximum_ram_page, NULL);
++#endif
++
+ level4p = (pgd_t *)__va(start_pgtable);
+- return init_level4_page(image, level4p, 0, end_pfn << PAGE_SHIFT);
++ return init_level4_page(image, level4p, 0, x_end_pfn << PAGE_SHIFT);
+ }
+
+ int machine_kexec_prepare(struct kimage *image)
--- 0005/patches/linux-2.6.16.29/series
+++ work/patches/linux-2.6.16.29/series 2006-10-23 11:36:17.000000000 +0900
@@ -4,6 +4,9 @@ git-2a8a3d5b65e86ec1dfef7d268c64a909eab9
git-3566561bfadffcb5dbc85d576be80c0dbf2cccc9.patch
linux-2.6.19-rc1-kexec-move_segment_code-i386.patch
linux-2.6.19-rc1-kexec-xen-i386.patch
+git-4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f.patch
+linux-2.6.19-rc1-kexec-move_segment_code-x86_64.patch
+linux-2.6.19-rc1-kexec-xen-x86_64.patch
blktap-aio-16_03_06.patch
device_bind.patch
fix-hz-suspend.patch
--- 0001/xen/arch/x86/x86_64/entry.S
+++ work/xen/arch/x86/x86_64/entry.S 2006-10-23 11:36:17.000000000 +0900
@@ -573,6 +573,7 @@ ENTRY(hypercall_table)
.quad do_hvm_op
.quad do_sysctl /* 35 */
.quad do_domctl
+ .quad do_kexec_op
.rept NR_hypercalls-((.-hypercall_table)/8)
.quad do_ni_hypercall
.endr
@@ -615,6 +616,7 @@ ENTRY(hypercall_args_table)
.byte 2 /* do_hvm_op */
.byte 1 /* do_sysctl */ /* 35 */
.byte 1 /* do_domctl */
+ .byte 2 /* do_kexec */
.rept NR_hypercalls-(.-hypercall_args_table)
.byte 0 /* do_ni_hypercall */
.endr
--- 0004/xen/include/asm-x86/x86_64/elf.h
+++ work/xen/include/asm-x86/x86_64/elf.h 2006-10-23 11:36:17.000000000 +0900
@@ -1,14 +1,51 @@
+/*
+ * Based on include/asm-x86_64/elf.h:ELF_CORE_COPY_REGS from Linux 2.6.16
+ */
+
#ifndef __X86_64_ELF_H__
#define __X86_64_ELF_H__
-#include <xen/lib.h> /* for printk() used in stub */
+#define ELF_NGREG 27
-#define ELF_NGREG 1 /* XXX: Define to be at least as large as
- however many register slots are needed when
- crash notes are written during crash dump */
+/* XXX: Xen doesn't have orig_rax, so it is omitted.
+ * Xen dosn't have threads, so fs and gs are read from the CPU and
+ * thus values 21 and 22 are just duplicates of 25 and 26
+ * respectively. All these values could be passed from dom0 in the
+ * case of it crashing, but does that help?
+ *
+ * Lastly, I'm not sure why ds, es, fs and gs are read from
+ * the CPU rather than regs, but linux does this
+ */
-#define ELF_CORE_COPY_REGS(pr_reg, regs) \
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+#define ELF_CORE_COPY_REGS(pr_reg, regs) do { \
+ unsigned v; \
+ (pr_reg)[0] = (regs)->r15; \
+ (pr_reg)[1] = (regs)->r14; \
+ (pr_reg)[2] = (regs)->r13; \
+ (pr_reg)[3] = (regs)->r12; \
+ (pr_reg)[4] = (regs)->rbp; \
+ (pr_reg)[5] = (regs)->rbx; \
+ (pr_reg)[6] = (regs)->r11; \
+ (pr_reg)[7] = (regs)->r10; \
+ (pr_reg)[8] = (regs)->r9; \
+ (pr_reg)[9] = (regs)->r8; \
+ (pr_reg)[10] = (regs)->rax; \
+ (pr_reg)[11] = (regs)->rcx; \
+ (pr_reg)[12] = (regs)->rdx; \
+ (pr_reg)[13] = (regs)->rsi; \
+ (pr_reg)[14] = (regs)->rdi; \
+ (pr_reg)[16] = (regs)->rip; \
+ (pr_reg)[17] = (regs)->cs; \
+ (pr_reg)[18] = (regs)->eflags; \
+ (pr_reg)[19] = (regs)->rsp; \
+ (pr_reg)[20] = (regs)->ss; \
+ asm("movl %%fs,%0" : "=r" (v)); (pr_reg)[21] = v; \
+ asm("movl %%gs,%0" : "=r" (v)); (pr_reg)[22] = v; \
+ asm("movl %%ds,%0" : "=r" (v)); (pr_reg)[23] = v; \
+ asm("movl %%es,%0" : "=r" (v)); (pr_reg)[24] = v; \
+ asm("movl %%fs,%0" : "=r" (v)); (pr_reg)[25] = v; \
+ asm("movl %%gs,%0" : "=r" (v)); (pr_reg)[26] = v; \
+} while(0);
#endif /* __X86_64_ELF_H__ */
--- 0004/xen/include/asm-x86/x86_64/kexec.h
+++ work/xen/include/asm-x86/x86_64/kexec.h 2006-10-23 11:36:17.000000000 +0900
@@ -1,20 +1,68 @@
+/******************************************************************************
+ * kexec.h
+ *
+ * Based heavily on machine_kexec.c and kexec.h from Linux 2.6.19-rc1
+ *
+ */
+
#ifndef __X86_64_KEXEC_H__
#define __X86_64_KEXEC_H__
-
-#include <xen/lib.h> /* for printk() used in stub */
+
+#include <xen/lib.h>
#include <xen/types.h>
#include <public/xen.h>
#include <xen/kexec.h>
-
+#include <asm/processor.h>
+#include <xen/string.h>
+#include <asm/fixmap.h>
+
+/*
+ * Saving the registers of the cpu on which panic occured in
+ * crash_kexec to save a valid sp. The registers of other cpus
+ * will be saved in machine_crash_shutdown while shooting down them.
+ */
static inline void crash_setup_regs(struct cpu_user_regs *newregs,
- struct cpu_user_regs *oldregs)
+ struct cpu_user_regs *oldregs)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ if (oldregs)
+ memcpy(newregs, oldregs, sizeof(*newregs));
+ else {
+ __asm__ __volatile__("movq %%rbx,%0" : "=m"(newregs->rbx));
+ __asm__ __volatile__("movq %%rcx,%0" : "=m"(newregs->rcx));
+ __asm__ __volatile__("movq %%rdx,%0" : "=m"(newregs->rdx));
+ __asm__ __volatile__("movq %%rsi,%0" : "=m"(newregs->rsi));
+ __asm__ __volatile__("movq %%rdi,%0" : "=m"(newregs->rdi));
+ __asm__ __volatile__("movq %%rbp,%0" : "=m"(newregs->rbp));
+ __asm__ __volatile__("movq %%rax,%0" : "=m"(newregs->rax));
+ __asm__ __volatile__("movq %%rsp,%0" : "=m"(newregs->rsp));
+ __asm__ __volatile__("movq %%r8,%0" : "=m"(newregs->r8));
+ __asm__ __volatile__("movq %%r9,%0" : "=m"(newregs->r9));
+ __asm__ __volatile__("movq %%r10,%0" : "=m"(newregs->r10));
+ __asm__ __volatile__("movq %%r11,%0" : "=m"(newregs->r11));
+ __asm__ __volatile__("movq %%r12,%0" : "=m"(newregs->r12));
+ __asm__ __volatile__("movq %%r13,%0" : "=m"(newregs->r13));
+ __asm__ __volatile__("movq %%r14,%0" : "=m"(newregs->r14));
+ __asm__ __volatile__("movq %%r15,%0" : "=m"(newregs->r15));
+ __asm__ __volatile__("movl %%ss, %%eax;" :"=a"(newregs->ss));
+ __asm__ __volatile__("movl %%cs, %%eax;" :"=a"(newregs->cs));
+ __asm__ __volatile__("pushfq; popq %0" :"=m"(newregs->eflags));
+
+ newregs->rip = (unsigned long)current_text_addr();
+ }
}
+typedef void (*relocate_new_kernel_t)(
+ unsigned long indirection_page,
+ unsigned long page_list,
+ unsigned long start_address);
+
static inline void machine_kexec(xen_kexec_image_t *image)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ relocate_new_kernel_t rnk;
+
+ rnk = (relocate_new_kernel_t) image->page_list[1];
+ (*rnk)(image->indirection_page, (unsigned long)image->page_list,
+ image->start_address);
}
#endif /* __X86_64_KEXEC_H__ */
^ permalink raw reply [flat|nested] 30+ messages in thread
* [PATCH 04/04] Kexec / Kdump: x86_64 specific code
2006-10-16 8:33 [PATCH 00/04] Kexec / Kdump: Release 20061016 (xen-unstable-11760) Magnus Damm
@ 2006-10-16 8:33 ` Magnus Damm
0 siblings, 0 replies; 30+ messages in thread
From: Magnus Damm @ 2006-10-16 8:33 UTC (permalink / raw)
To: Keir Fraser
Cc: Ian Pratt, Kazuo Moriwaka, xen-devel, Akio Takebe, magnus.damm,
Isaku Yamahata, Magnus Damm, Horms
[PATCH 04/04] Kexec / Kdump: x86_64 specific code
This patch contains the x86_64 implementation of Kexec / Kdump for Xen.
Signed-Off-By: Magnus Damm <magnus@valinux.co.jp>
---
Applies on top of xen-unstable-11760.
buildconfigs/linux-defconfig_xen_x86_64 | 1
linux-2.6-xen-sparse/arch/x86_64/Kconfig | 2
linux-2.6-xen-sparse/arch/x86_64/kernel/Makefile | 2
linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c | 27
linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h | 7
linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h | 2
patches/linux-2.6.16.29/series | 3
linux-2.6-xen-sparse/include/asm-x86_64/kexec-xen.h | 64 +
patches/linux-2.6.16.29/git-4b..1f.patch | 375 ++++++
patches/linux-2.6.16.29/linux-2.6.19-rc1-kexe..code-x86_64.patch | 161 ++++
patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-xen-x86_64.patch | 162 ++++
xen/arch/x86/x86_64/entry.S | 2
xen/arch/x86/x86_64/machine_kexec.c | 19
xen/include/asm-x86/x86_64/elf.h | 48 +
xen/include/asm-x86/x86_64/kexec.h | 33
xen/include/public/kexec.h | 3
16 files changed, 897 insertions(+), 14 deletions(-)
--- 0002/buildconfigs/linux-defconfig_xen_x86_64
+++ work/buildconfigs/linux-defconfig_xen_x86_64 2006-10-16 12:41:27.000000000 +0900
@@ -138,6 +138,7 @@ CONFIG_SWIOTLB=y
CONFIG_PHYSICAL_START=0x100000
CONFIG_SECCOMP=y
CONFIG_HZ_100=y
+CONFIG_KEXEC=y
# CONFIG_HZ_250 is not set
# CONFIG_HZ_1000 is not set
CONFIG_HZ=100
--- 0001/linux-2.6-xen-sparse/arch/x86_64/Kconfig
+++ work/linux-2.6-xen-sparse/arch/x86_64/Kconfig 2006-10-16 12:41:27.000000000 +0900
@@ -435,7 +435,7 @@ config X86_MCE_AMD
config KEXEC
bool "kexec system call (EXPERIMENTAL)"
- depends on EXPERIMENTAL && !X86_64_XEN
+ depends on EXPERIMENTAL && !XEN_UNPRIVILEGED_GUEST
help
kexec is a system call that implements the ability to shutdown your
current kernel, and to start another kernel. It is like a reboot
--- 0001/linux-2.6-xen-sparse/arch/x86_64/kernel/Makefile
+++ work/linux-2.6-xen-sparse/arch/x86_64/kernel/Makefile 2006-10-16 12:41:27.000000000 +0900
@@ -59,7 +59,7 @@ pci-dma-y += ../../i386/kernel/pci-dma
microcode-$(subst m,y,$(CONFIG_MICROCODE)) := ../../i386/kernel/microcode-xen.o
quirks-y := ../../i386/kernel/quirks-xen.o
-n-obj-xen := i8259.o reboot.o i8237.o smpboot.o trampoline.o
+n-obj-xen := i8259.o reboot.o i8237.o smpboot.o trampoline.o crash.o
include $(srctree)/scripts/Makefile.xen
--- 0001/linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c
+++ work/linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c 2006-10-16 12:41:56.000000000 +0900
@@ -80,6 +80,10 @@
#include <asm/mach-xen/setup_arch_post.h>
#include <xen/interface/memory.h>
+#ifdef CONFIG_XEN
+#include <xen/interface/kexec.h>
+#endif
+
extern unsigned long start_pfn;
extern struct edid_info edid_info;
@@ -450,6 +454,7 @@ static __init void parse_cmdline_early (
* after a kernel panic.
*/
else if (!memcmp(from, "crashkernel=", 12)) {
+#ifndef CONFIG_XEN
unsigned long size, base;
size = memparse(from+12, &from);
if (*from == '@') {
@@ -460,6 +465,10 @@ static __init void parse_cmdline_early (
crashk_res.start = base;
crashk_res.end = base + size - 1;
}
+#else
+ printk("Ignoring crashkernel command line, "
+ "parameter will be supplied by xen\n");
+#endif
}
#endif
@@ -812,10 +821,23 @@ void __init setup_arch(char **cmdline_p)
#endif
#endif /* !CONFIG_XEN */
#ifdef CONFIG_KEXEC
+#ifndef CONFIG_XEN
if (crashk_res.start != crashk_res.end) {
reserve_bootmem(crashk_res.start,
crashk_res.end - crashk_res.start + 1);
}
+#else
+ {
+ xen_kexec_reserve_t reservation;
+ BUG_ON(HYPERVISOR_kexec(KEXEC_CMD_kexec_reserve, 0,
+ &reservation));
+ if (reservation.size) {
+ crashk_res.start = reservation.start;
+ crashk_res.end = reservation.start +
+ reservation.size - 1;
+ }
+ }
+#endif
#endif
paging_init();
@@ -954,6 +976,11 @@ void __init setup_arch(char **cmdline_p)
iommu_hole_init();
#endif
+#ifdef CONFIG_KEXEC
+ if (crashk_res.start != crashk_res.end)
+ request_resource(&ioport_resource, &crashk_res);
+#endif
+
#ifdef CONFIG_XEN
{
struct physdev_set_iopl set_iopl;
--- /dev/null
+++ work/linux-2.6-xen-sparse/include/asm-x86_64/kexec-xen.h 2006-10-16 12:41:28.000000000 +0900
@@ -0,0 +1,64 @@
+/*
+ * include/asm-x86_64/kexec-xen.h
+ *
+ * Created By: Horms <horms@verge.net.au>
+ */
+
+#ifndef _X86_64_KEXEC_XEN_H
+#define _X86_64_KEXEC_XEN_H
+
+#include <asm/ptrace.h>
+#include <asm/types.h>
+#include <xen/interface/arch-x86_64.h>
+
+static inline void crash_translate_regs(struct pt_regs *linux_regs,
+ struct cpu_user_regs *xen_regs)
+{
+ xen_regs->r15 = linux_regs->r15;
+ xen_regs->r14 = linux_regs->r14;
+ xen_regs->r13 = linux_regs->r13;
+ xen_regs->r12 = linux_regs->r12;
+ xen_regs->rbp = linux_regs->rbp;
+ xen_regs->rbx = linux_regs->rbx;
+ xen_regs->r11 = linux_regs->r11;
+ xen_regs->r10 = linux_regs->r10;
+ xen_regs->r9 = linux_regs->r9;
+ xen_regs->r8 = linux_regs->r8;
+ xen_regs->rax = linux_regs->rax;
+ xen_regs->rcx = linux_regs->rcx;
+ xen_regs->rdx = linux_regs->rdx;
+ xen_regs->rsi = linux_regs->rsi;
+ xen_regs->rdi = linux_regs->rdi;
+ xen_regs->rip = linux_regs->rip;
+ xen_regs->cs = linux_regs->cs;
+ xen_regs->rflags = linux_regs->eflags;
+ xen_regs->rsp = linux_regs->rsp;
+ xen_regs->ss = linux_regs->ss;
+}
+
+/* Kexec needs to know about the actual physical addresss.
+ * But in xen, on some architectures, a physical address is a
+ * pseudo-physical addresss. */
+#ifdef CONFIG_XEN
+#define kexec_page_to_pfn(page) pfn_to_mfn(page_to_pfn(page))
+#define kexec_pfn_to_page(pfn) pfn_to_page(mfn_to_pfn(pfn))
+#define kexec_virt_to_phys(addr) virt_to_machine(addr)
+#define kexec_phys_to_virt(addr) phys_to_virt(machine_to_phys(addr))
+#else
+#define kexec_page_to_pfn(page) page_to_pfn(page)
+#define kexec_pfn_to_page(pfn) pfn_to_page(pfn)
+#define kexec_virt_to_phys(addr) virt_to_phys(addr)
+#define kexec_phys_to_virt(addr) phys_to_virt(addr)
+#endif
+
+#endif /* _X86_64_KEXEC_XEN_H */
+
+/*
+ * Local variables:
+ * c-file-style: "linux"
+ * indent-tabs-mode: t
+ * c-indent-level: 8
+ * c-basic-offset: 8
+ * tab-width: 8
+ * End:
+ */
--- 0001/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h
+++ work/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h 2006-10-16 12:41:27.000000000 +0900
@@ -386,4 +386,11 @@ HYPERVISOR_xenoprof_op(
return _hypercall2(int, xenoprof_op, op, arg);
}
+static inline int
+HYPERVISOR_kexec(
+ unsigned long op, unsigned int arg1, void * extra_args)
+{
+ return _hypercall3(int, kexec_op, op, arg1, extra_args);
+}
+
#endif /* __HYPERCALL_H__ */
--- 0001/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h
+++ work/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h 2006-10-16 12:41:27.000000000 +0900
@@ -90,6 +90,8 @@ extern unsigned long profile_pc(struct p
#define profile_pc(regs) instruction_pointer(regs)
#endif
+#include <linux/compiler.h>
+
void signal_fault(struct pt_regs *regs, void __user *frame, char *where);
struct task_struct;
--- /dev/null
+++ work/patches/linux-2.6.16.29/git-4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f.patch 2006-10-16 12:41:28.000000000 +0900
@@ -0,0 +1,375 @@
+From: Magnus Damm <magnus@valinux.co.jp>
+Date: Tue, 26 Sep 2006 08:52:38 +0000 (+0200)
+Subject: [PATCH] Avoid overwriting the current pgd (V4, x86_64)
+X-Git-Tag: v2.6.19-rc1
+X-Git-Url: http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f
+
+[PATCH] Avoid overwriting the current pgd (V4, x86_64)
+
+kexec: Avoid overwriting the current pgd (V4, x86_64)
+
+This patch upgrades the x86_64-specific kexec code to avoid overwriting the
+current pgd. Overwriting the current pgd is bad when CONFIG_CRASH_DUMP is used
+to start a secondary kernel that dumps the memory of the previous kernel.
+
+The code introduces a new set of page tables. These tables are used to provide
+an executable identity mapping without overwriting the current pgd.
+
+Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
+Signed-off-by: Andi Kleen <ak@suse.de>
+---
+
+--- a/arch/x86_64/kernel/machine_kexec.c
++++ b/arch/x86_64/kernel/machine_kexec.c
+@@ -15,6 +15,15 @@
+ #include <asm/mmu_context.h>
+ #include <asm/io.h>
+
++#define PAGE_ALIGNED __attribute__ ((__aligned__(PAGE_SIZE)))
++static u64 kexec_pgd[512] PAGE_ALIGNED;
++static u64 kexec_pud0[512] PAGE_ALIGNED;
++static u64 kexec_pmd0[512] PAGE_ALIGNED;
++static u64 kexec_pte0[512] PAGE_ALIGNED;
++static u64 kexec_pud1[512] PAGE_ALIGNED;
++static u64 kexec_pmd1[512] PAGE_ALIGNED;
++static u64 kexec_pte1[512] PAGE_ALIGNED;
++
+ static void init_level2_page(pmd_t *level2p, unsigned long addr)
+ {
+ unsigned long end_addr;
+@@ -144,32 +153,19 @@ static void load_segments(void)
+ );
+ }
+
+-typedef NORET_TYPE void (*relocate_new_kernel_t)(unsigned long indirection_page,
+- unsigned long control_code_buffer,
+- unsigned long start_address,
+- unsigned long pgtable) ATTRIB_NORET;
+-
+-extern const unsigned char relocate_new_kernel[];
+-extern const unsigned long relocate_new_kernel_size;
+-
+ int machine_kexec_prepare(struct kimage *image)
+ {
+- unsigned long start_pgtable, control_code_buffer;
++ unsigned long start_pgtable;
+ int result;
+
+ /* Calculate the offsets */
+ start_pgtable = page_to_pfn(image->control_code_page) << PAGE_SHIFT;
+- control_code_buffer = start_pgtable + PAGE_SIZE;
+
+ /* Setup the identity mapped 64bit page table */
+ result = init_pgtable(image, start_pgtable);
+ if (result)
+ return result;
+
+- /* Place the code in the reboot code buffer */
+- memcpy(__va(control_code_buffer), relocate_new_kernel,
+- relocate_new_kernel_size);
+-
+ return 0;
+ }
+
+@@ -184,28 +180,34 @@ void machine_kexec_cleanup(struct kimage
+ */
+ NORET_TYPE void machine_kexec(struct kimage *image)
+ {
+- unsigned long page_list;
+- unsigned long control_code_buffer;
+- unsigned long start_pgtable;
+- relocate_new_kernel_t rnk;
++ unsigned long page_list[PAGES_NR];
++ void *control_page;
+
+ /* Interrupts aren't acceptable while we reboot */
+ local_irq_disable();
+
+- /* Calculate the offsets */
+- page_list = image->head;
+- start_pgtable = page_to_pfn(image->control_code_page) << PAGE_SHIFT;
+- control_code_buffer = start_pgtable + PAGE_SIZE;
++ control_page = page_address(image->control_code_page) + PAGE_SIZE;
++ memcpy(control_page, relocate_kernel, PAGE_SIZE);
+
+- /* Set the low half of the page table to my identity mapped
+- * page table for kexec. Leave the high half pointing at the
+- * kernel pages. Don't bother to flush the global pages
+- * as that will happen when I fully switch to my identity mapped
+- * page table anyway.
+- */
+- memcpy(__va(read_cr3()), __va(start_pgtable), PAGE_SIZE/2);
+- __flush_tlb();
++ page_list[PA_CONTROL_PAGE] = __pa(control_page);
++ page_list[VA_CONTROL_PAGE] = (unsigned long)relocate_kernel;
++ page_list[PA_PGD] = __pa(kexec_pgd);
++ page_list[VA_PGD] = (unsigned long)kexec_pgd;
++ page_list[PA_PUD_0] = __pa(kexec_pud0);
++ page_list[VA_PUD_0] = (unsigned long)kexec_pud0;
++ page_list[PA_PMD_0] = __pa(kexec_pmd0);
++ page_list[VA_PMD_0] = (unsigned long)kexec_pmd0;
++ page_list[PA_PTE_0] = __pa(kexec_pte0);
++ page_list[VA_PTE_0] = (unsigned long)kexec_pte0;
++ page_list[PA_PUD_1] = __pa(kexec_pud1);
++ page_list[VA_PUD_1] = (unsigned long)kexec_pud1;
++ page_list[PA_PMD_1] = __pa(kexec_pmd1);
++ page_list[VA_PMD_1] = (unsigned long)kexec_pmd1;
++ page_list[PA_PTE_1] = __pa(kexec_pte1);
++ page_list[VA_PTE_1] = (unsigned long)kexec_pte1;
+
++ page_list[PA_TABLE_PAGE] =
++ (unsigned long)__pa(page_address(image->control_code_page));
+
+ /* The segment registers are funny things, they have both a
+ * visible and an invisible part. Whenever the visible part is
+@@ -222,9 +224,10 @@ NORET_TYPE void machine_kexec(struct kim
+ */
+ set_gdt(phys_to_virt(0),0);
+ set_idt(phys_to_virt(0),0);
++
+ /* now call it */
+- rnk = (relocate_new_kernel_t) control_code_buffer;
+- (*rnk)(page_list, control_code_buffer, image->start, start_pgtable);
++ relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
++ image->start);
+ }
+
+ /* crashkernel=size@addr specifies the location to reserve for
+--- a/arch/x86_64/kernel/relocate_kernel.S
++++ b/arch/x86_64/kernel/relocate_kernel.S
+@@ -7,31 +7,169 @@
+ */
+
+ #include <linux/linkage.h>
++#include <asm/page.h>
++#include <asm/kexec.h>
+
+- /*
+- * Must be relocatable PIC code callable as a C function, that once
+- * it starts can not use the previous processes stack.
+- */
+- .globl relocate_new_kernel
++/*
++ * Must be relocatable PIC code callable as a C function
++ */
++
++#define PTR(x) (x << 3)
++#define PAGE_ALIGNED (1 << PAGE_SHIFT)
++#define PAGE_ATTR 0x63 /* _PAGE_PRESENT|_PAGE_RW|_PAGE_ACCESSED|_PAGE_DIRTY */
++
++ .text
++ .align PAGE_ALIGNED
+ .code64
++ .globl relocate_kernel
++relocate_kernel:
++ /* %rdi indirection_page
++ * %rsi page_list
++ * %rdx start address
++ */
++
++ /* map the control page at its virtual address */
++
++ movq $0x0000ff8000000000, %r10 /* mask */
++ mov $(39 - 3), %cl /* bits to shift */
++ movq PTR(VA_CONTROL_PAGE)(%rsi), %r11 /* address to map */
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PGD)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PUD_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PUD_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PMD_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PMD_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PTE_0)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PTE_0)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ /* identity map the control page at its physical address */
++
++ movq $0x0000ff8000000000, %r10 /* mask */
++ mov $(39 - 3), %cl /* bits to shift */
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r11 /* address to map */
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PGD)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PUD_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PUD_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PMD_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PMD_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_PTE_1)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
++ shrq $9, %r10
++ sub $9, %cl
++
++ movq %r11, %r9
++ andq %r10, %r9
++ shrq %cl, %r9
++
++ movq PTR(VA_PTE_1)(%rsi), %r8
++ addq %r8, %r9
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++ orq $PAGE_ATTR, %r8
++ movq %r8, (%r9)
++
+ relocate_new_kernel:
+- /* %rdi page_list
+- * %rsi reboot_code_buffer
++ /* %rdi indirection_page
++ * %rsi page_list
+ * %rdx start address
+- * %rcx page_table
+- * %r8 arg5
+- * %r9 arg6
+ */
+
+ /* zero out flags, and disable interrupts */
+ pushq $0
+ popfq
+
+- /* set a new stack at the bottom of our page... */
+- lea 4096(%rsi), %rsp
++ /* get physical address of control page now */
++ /* this is impossible after page table switch */
++ movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
++
++ /* get physical address of page table now too */
++ movq PTR(PA_TABLE_PAGE)(%rsi), %rcx
++
++ /* switch to new set of page tables */
++ movq PTR(PA_PGD)(%rsi), %r9
++ movq %r9, %cr3
++
++ /* setup a new stack at the end of the physical control page */
++ lea 4096(%r8), %rsp
++
++ /* jump to identity mapped page */
++ addq $(identity_mapped - relocate_kernel), %r8
++ pushq %r8
++ ret
+
+- /* store the parameters back on the stack */
+- pushq %rdx /* store the start address */
++identity_mapped:
++ /* store the start address on the stack */
++ pushq %rdx
+
+ /* Set cr0 to a known state:
+ * 31 1 == Paging enabled
+@@ -136,8 +274,3 @@ relocate_new_kernel:
+ xorq %r15, %r15
+
+ ret
+-relocate_new_kernel_end:
+-
+- .globl relocate_new_kernel_size
+-relocate_new_kernel_size:
+- .quad relocate_new_kernel_end - relocate_new_kernel
+--- a/include/asm-x86_64/kexec.h
++++ b/include/asm-x86_64/kexec.h
+@@ -1,6 +1,27 @@
+ #ifndef _X86_64_KEXEC_H
+ #define _X86_64_KEXEC_H
+
++#define PA_CONTROL_PAGE 0
++#define VA_CONTROL_PAGE 1
++#define PA_PGD 2
++#define VA_PGD 3
++#define PA_PUD_0 4
++#define VA_PUD_0 5
++#define PA_PMD_0 6
++#define VA_PMD_0 7
++#define PA_PTE_0 8
++#define VA_PTE_0 9
++#define PA_PUD_1 10
++#define VA_PUD_1 11
++#define PA_PMD_1 12
++#define VA_PMD_1 13
++#define PA_PTE_1 14
++#define VA_PTE_1 15
++#define PA_TABLE_PAGE 16
++#define PAGES_NR 17
++
++#ifndef __ASSEMBLY__
++
+ #include <linux/string.h>
+
+ #include <asm/page.h>
+@@ -64,4 +85,12 @@ static inline void crash_setup_regs(stru
+ newregs->rip = (unsigned long)current_text_addr();
+ }
+ }
++
++NORET_TYPE void
++relocate_kernel(unsigned long indirection_page,
++ unsigned long page_list,
++ unsigned long start_address) ATTRIB_NORET;
++
++#endif /* __ASSEMBLY__ */
++
+ #endif /* _X86_64_KEXEC_H */
--- /dev/null
+++ work/patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-move_segment_code-x86_64.patch 2006-10-16 12:41:28.000000000 +0900
@@ -0,0 +1,161 @@
+kexec: Move asm segment handling code to the assembly file (x86_64)
+
+This patch moves the idt, gdt, and segment handling code from machine_kexec.c
+to relocate_kernel.S. The main reason behind this move is to avoid code
+duplication in the Xen hypervisor. With this patch all code required to kexec
+is put on the control page.
+
+On top of that this patch also counts as a cleanup - I think it is much
+nicer to write assembly directly in assembly files than wrap inline assembly
+in C functions for no apparent reason.
+
+Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
+---
+
+ Applies to 2.6.19-rc1.
+
+ machine_kexec.c | 58 -----------------------------------------------------
+ relocate_kernel.S | 50 +++++++++++++++++++++++++++++++++++++++++----
+ 2 files changed, 45 insertions(+), 63 deletions(-)
+
+--- 0002/arch/x86_64/kernel/machine_kexec.c
++++ work/arch/x86_64/kernel/machine_kexec.c 2006-10-05 16:15:49.000000000 +0900
+@@ -112,47 +112,6 @@ static int init_pgtable(struct kimage *i
+ return init_level4_page(image, level4p, 0, end_pfn << PAGE_SHIFT);
+ }
+
+-static void set_idt(void *newidt, u16 limit)
+-{
+- struct desc_ptr curidt;
+-
+- /* x86-64 supports unaliged loads & stores */
+- curidt.size = limit;
+- curidt.address = (unsigned long)newidt;
+-
+- __asm__ __volatile__ (
+- "lidtq %0\n"
+- : : "m" (curidt)
+- );
+-};
+-
+-
+-static void set_gdt(void *newgdt, u16 limit)
+-{
+- struct desc_ptr curgdt;
+-
+- /* x86-64 supports unaligned loads & stores */
+- curgdt.size = limit;
+- curgdt.address = (unsigned long)newgdt;
+-
+- __asm__ __volatile__ (
+- "lgdtq %0\n"
+- : : "m" (curgdt)
+- );
+-};
+-
+-static void load_segments(void)
+-{
+- __asm__ __volatile__ (
+- "\tmovl %0,%%ds\n"
+- "\tmovl %0,%%es\n"
+- "\tmovl %0,%%ss\n"
+- "\tmovl %0,%%fs\n"
+- "\tmovl %0,%%gs\n"
+- : : "a" (__KERNEL_DS) : "memory"
+- );
+-}
+-
+ int machine_kexec_prepare(struct kimage *image)
+ {
+ unsigned long start_pgtable;
+@@ -209,23 +168,6 @@ NORET_TYPE void machine_kexec(struct kim
+ page_list[PA_TABLE_PAGE] =
+ (unsigned long)__pa(page_address(image->control_code_page));
+
+- /* The segment registers are funny things, they have both a
+- * visible and an invisible part. Whenever the visible part is
+- * set to a specific selector, the invisible part is loaded
+- * with from a table in memory. At no other time is the
+- * descriptor table in memory accessed.
+- *
+- * I take advantage of this here by force loading the
+- * segments, before I zap the gdt with an invalid value.
+- */
+- load_segments();
+- /* The gdt & idt are now invalid.
+- * If you want to load them you must set up your own idt & gdt.
+- */
+- set_gdt(phys_to_virt(0),0);
+- set_idt(phys_to_virt(0),0);
+-
+- /* now call it */
+ relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
+ image->start);
+ }
+--- 0002/arch/x86_64/kernel/relocate_kernel.S
++++ work/arch/x86_64/kernel/relocate_kernel.S 2006-10-05 16:18:07.000000000 +0900
+@@ -159,13 +159,39 @@ relocate_new_kernel:
+ movq PTR(PA_PGD)(%rsi), %r9
+ movq %r9, %cr3
+
++ /* setup idt */
++ movq %r8, %rax
++ addq $(idt_80 - relocate_kernel), %rax
++ lidtq (%rax)
++
++ /* setup gdt */
++ movq %r8, %rax
++ addq $(gdt - relocate_kernel), %rax
++ movq %r8, %r9
++ addq $((gdt_80 - relocate_kernel) + 2), %r9
++ movq %rax, (%r9)
++
++ movq %r8, %rax
++ addq $(gdt_80 - relocate_kernel), %rax
++ lgdtq (%rax)
++
++ /* setup data segment registers */
++ xorl %eax, %eax
++ movl %eax, %ds
++ movl %eax, %es
++ movl %eax, %fs
++ movl %eax, %gs
++ movl %eax, %ss
++
+ /* setup a new stack at the end of the physical control page */
+ lea 4096(%r8), %rsp
+
+- /* jump to identity mapped page */
+- addq $(identity_mapped - relocate_kernel), %r8
+- pushq %r8
+- ret
++ /* load new code segment and jump to identity mapped page */
++ movq %r8, %rax
++ addq $(identity_mapped - relocate_kernel), %rax
++ pushq $(gdt_cs - gdt)
++ pushq %rax
++ lretq
+
+ identity_mapped:
+ /* store the start address on the stack */
+@@ -272,5 +298,19 @@ identity_mapped:
+ xorq %r13, %r13
+ xorq %r14, %r14
+ xorq %r15, %r15
+-
+ ret
++
++ .align 16
++gdt:
++ .quad 0x0000000000000000 /* NULL descriptor */
++gdt_cs:
++ .quad 0x00af9a000000ffff
++gdt_end:
++
++gdt_80:
++ .word gdt_end - gdt - 1 /* limit */
++ .quad 0 /* base - filled in by code above */
++
++idt_80:
++ .word 0 /* limit */
++ .quad 0 /* base */
--- /dev/null
+++ work/patches/linux-2.6.16.29/linux-2.6.19-rc1-kexec-xen-x86_64.patch 2006-10-16 12:41:28.000000000 +0900
@@ -0,0 +1,162 @@
+--- 0006/arch/x86_64/kernel/machine_kexec.c
++++ work/arch/x86_64/kernel/machine_kexec.c 2006-10-06 15:36:16.000000000 +0900
+@@ -24,6 +24,104 @@ static u64 kexec_pud1[512] PAGE_ALIGNED;
+ static u64 kexec_pmd1[512] PAGE_ALIGNED;
+ static u64 kexec_pte1[512] PAGE_ALIGNED;
+
++#ifdef CONFIG_XEN
++
++/* In the case of Xen, override hypervisor functions to be able to create
++ * a regular identity mapping page table...
++ */
++
++#include <xen/interface/kexec.h>
++#include <xen/interface/memory.h>
++
++#define x__pmd(x) ((pmd_t) { (x) } )
++#define x__pud(x) ((pud_t) { (x) } )
++#define x__pgd(x) ((pgd_t) { (x) } )
++
++#define x_pmd_val(x) ((x).pmd)
++#define x_pud_val(x) ((x).pud)
++#define x_pgd_val(x) ((x).pgd)
++
++static inline void x_set_pmd(pmd_t *dst, pmd_t val)
++{
++ x_pmd_val(*dst) = x_pmd_val(val);
++}
++
++static inline void x_set_pud(pud_t *dst, pud_t val)
++{
++ x_pud_val(*dst) = phys_to_machine(x_pud_val(val));
++}
++
++static inline void x_pud_clear (pud_t *pud)
++{
++ x_pud_val(*pud) = 0;
++}
++
++static inline void x_set_pgd(pgd_t *dst, pgd_t val)
++{
++ x_pgd_val(*dst) = phys_to_machine(x_pgd_val(val));
++}
++
++static inline void x_pgd_clear (pgd_t * pgd)
++{
++ x_pgd_val(*pgd) = 0;
++}
++
++#define X__PAGE_KERNEL_LARGE_EXEC \
++ _PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_PSE
++#define X_KERNPG_TABLE _PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY
++
++#define __ma(x) (pfn_to_mfn(__pa((x)) >> PAGE_SHIFT) << PAGE_SHIFT)
++
++#if PAGES_NR > KEXEC_XEN_NO_PAGES
++#error PAGES_NR is greater than KEXEC_XEN_NO_PAGES - Xen support will break
++#endif
++
++#if PA_CONTROL_PAGE != 0
++#error PA_CONTROL_PAGE is non zero - Xen support will break
++#endif
++
++void machine_kexec_setup_load_arg(xen_kexec_image_t *xki, struct kimage *image)
++{
++ void *control_page;
++ void *table_page;
++
++ memset(xki->page_list, 0, sizeof(xki->page_list));
++
++ control_page = page_address(image->control_code_page) + PAGE_SIZE;
++ memcpy(control_page, relocate_kernel, PAGE_SIZE);
++
++ table_page = page_address(image->control_code_page);
++
++ xki->page_list[PA_CONTROL_PAGE] = __ma(control_page);
++ xki->page_list[PA_TABLE_PAGE] = __ma(table_page);
++
++ xki->page_list[PA_PGD] = __ma(kexec_pgd);
++ xki->page_list[PA_PUD_0] = __ma(kexec_pud0);
++ xki->page_list[PA_PUD_1] = __ma(kexec_pud1);
++ xki->page_list[PA_PMD_0] = __ma(kexec_pmd0);
++ xki->page_list[PA_PMD_1] = __ma(kexec_pmd1);
++ xki->page_list[PA_PTE_0] = __ma(kexec_pte0);
++ xki->page_list[PA_PTE_1] = __ma(kexec_pte1);
++}
++
++#else /* CONFIG_XEN */
++
++#define x__pmd(x) __pmd(x)
++#define x__pud(x) __pud(x)
++#define x__pgd(x) __pgd(x)
++
++#define x_set_pmd(x, y) set_pmd(x, y)
++#define x_set_pud(x, y) set_pud(x, y)
++#define x_set_pgd(x, y) set_pgd(x, y)
++
++#define x_pud_clear(x) pud_clear(x)
++#define x_pgd_clear(x) pgd_clear(x)
++
++#define X__PAGE_KERNEL_LARGE_EXEC __PAGE_KERNEL_LARGE_EXEC
++#define X_KERNPG_TABLE _KERNPG_TABLE
++
++#endif /* CONFIG_XEN */
++
+ static void init_level2_page(pmd_t *level2p, unsigned long addr)
+ {
+ unsigned long end_addr;
+@@ -31,7 +129,7 @@ static void init_level2_page(pmd_t *leve
+ addr &= PAGE_MASK;
+ end_addr = addr + PUD_SIZE;
+ while (addr < end_addr) {
+- set_pmd(level2p++, __pmd(addr | __PAGE_KERNEL_LARGE_EXEC));
++ x_set_pmd(level2p++, x__pmd(addr | X__PAGE_KERNEL_LARGE_EXEC));
+ addr += PMD_SIZE;
+ }
+ }
+@@ -56,12 +154,12 @@ static int init_level3_page(struct kimag
+ }
+ level2p = (pmd_t *)page_address(page);
+ init_level2_page(level2p, addr);
+- set_pud(level3p++, __pud(__pa(level2p) | _KERNPG_TABLE));
++ x_set_pud(level3p++, x__pud(__pa(level2p) | X_KERNPG_TABLE));
+ addr += PUD_SIZE;
+ }
+ /* clear the unused entries */
+ while (addr < end_addr) {
+- pud_clear(level3p++);
++ x_pud_clear(level3p++);
+ addr += PUD_SIZE;
+ }
+ out:
+@@ -92,12 +190,12 @@ static int init_level4_page(struct kimag
+ if (result) {
+ goto out;
+ }
+- set_pgd(level4p++, __pgd(__pa(level3p) | _KERNPG_TABLE));
++ x_set_pgd(level4p++, x__pgd(__pa(level3p) | X_KERNPG_TABLE));
+ addr += PGDIR_SIZE;
+ }
+ /* clear the unused entries */
+ while (addr < end_addr) {
+- pgd_clear(level4p++);
++ x_pgd_clear(level4p++);
+ addr += PGDIR_SIZE;
+ }
+ out:
+@@ -108,8 +206,14 @@ out:
+ static int init_pgtable(struct kimage *image, unsigned long start_pgtable)
+ {
+ pgd_t *level4p;
++ unsigned long x_end_pfn = end_pfn;
++
++#ifdef CONFIG_XEN
++ x_end_pfn = HYPERVISOR_memory_op(XENMEM_maximum_ram_page, NULL);
++#endif
++
+ level4p = (pgd_t *)__va(start_pgtable);
+- return init_level4_page(image, level4p, 0, end_pfn << PAGE_SHIFT);
++ return init_level4_page(image, level4p, 0, x_end_pfn << PAGE_SHIFT);
+ }
+
+ int machine_kexec_prepare(struct kimage *image)
--- 0005/patches/linux-2.6.16.29/series
+++ work/patches/linux-2.6.16.29/series 2006-10-16 12:41:27.000000000 +0900
@@ -4,6 +4,9 @@ git-2a8a3d5b65e86ec1dfef7d268c64a909eab9
git-3566561bfadffcb5dbc85d576be80c0dbf2cccc9.patch
linux-2.6.19-rc1-kexec-move_segment_code-i386.patch
linux-2.6.19-rc1-kexec-xen-i386.patch
+git-4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f.patch
+linux-2.6.19-rc1-kexec-move_segment_code-x86_64.patch
+linux-2.6.19-rc1-kexec-xen-x86_64.patch
blktap-aio-16_03_06.patch
device_bind.patch
fix-hz-suspend.patch
--- 0001/xen/arch/x86/x86_64/entry.S
+++ work/xen/arch/x86/x86_64/entry.S 2006-10-16 12:41:27.000000000 +0900
@@ -573,6 +573,7 @@ ENTRY(hypercall_table)
.quad do_hvm_op
.quad do_sysctl /* 35 */
.quad do_domctl
+ .quad do_kexec_op
.rept NR_hypercalls-((.-hypercall_table)/8)
.quad do_ni_hypercall
.endr
@@ -615,6 +616,7 @@ ENTRY(hypercall_args_table)
.byte 2 /* do_hvm_op */
.byte 1 /* do_sysctl */ /* 35 */
.byte 1 /* do_domctl */
+ .byte 3 /* do_kexec */
.rept NR_hypercalls-(.-hypercall_args_table)
.byte 0 /* do_ni_hypercall */
.endr
--- 0004/xen/arch/x86/x86_64/machine_kexec.c
+++ work/xen/arch/x86/x86_64/machine_kexec.c 2006-10-16 12:41:27.000000000 +0900
@@ -4,18 +4,27 @@
*
* Created By: Horms <horms@verge.net.au>
*
- * Should be losely based on arch/x86_64/kernel/machine_kexec.c
+ * Losely based on arch/x86_64/kernel/machine_kexec.c
*/
-
-#include <xen/lib.h> /* for printk() used in stub */
+
#include <xen/types.h>
#include <public/kexec.h>
+#include <asm/fixmap.h>
+
+typedef void (*relocate_new_kernel_t)(
+ unsigned long indirection_page,
+ unsigned long page_list,
+ unsigned long start_address);
void machine_kexec(xen_kexec_image_t *image)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
-}
+ relocate_new_kernel_t rnk;
+ rnk = (relocate_new_kernel_t) fix_to_virt(FIX_KEXEC_BASE_0);
+ (*rnk)(image->indirection_page, (unsigned long)image->page_list,
+ image->start_address);
+ }
+
/*
* Local variables:
* mode: C
--- 0004/xen/include/asm-x86/x86_64/elf.h
+++ work/xen/include/asm-x86/x86_64/elf.h 2006-10-16 12:41:27.000000000 +0900
@@ -3,17 +3,55 @@
*
* Created By: Horms
*
- * Should pull be based on include/asm-x86_64/elf.h:ELF_CORE_COPY_REGS
- * from Linux 2.6.16
+ * Based on include/asm-x86_64/elf.h:ELF_CORE_COPY_REGS from Linux 2.6.16
*/
#ifndef __X86_ELF_X86_64_H__
#define __X86_ELF_X86_64_H__
-#include <xen/lib.h> /* for printk() used in stub */
+#include <xen/lib.h>
-#define ELF_CORE_COPY_REGS(pr_reg, regs) \
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+#include <xen/lib.h>
+
+/* XXX: Xen doesn't have orig_rax, so it is omitted.
+ * Xen dosn't have threads, so fs and gs are read from the CPU and
+ * thus values 21 and 22 are just duplicates of 25 and 26
+ * respectively. All these values could be passed from dom0 in the
+ * case of it crashing, but does that help?
+ *
+ * Lastly, I'm not sure why ds, es, fs and gs are read from
+ * the CPU rather than regs, but linux does this
+ */
+
+#define ELF_CORE_COPY_REGS(pr_reg, regs) do { \
+ unsigned v; \
+ (pr_reg)[0] = (regs)->r15; \
+ (pr_reg)[1] = (regs)->r14; \
+ (pr_reg)[2] = (regs)->r13; \
+ (pr_reg)[3] = (regs)->r12; \
+ (pr_reg)[4] = (regs)->rbp; \
+ (pr_reg)[5] = (regs)->rbx; \
+ (pr_reg)[6] = (regs)->r11; \
+ (pr_reg)[7] = (regs)->r10; \
+ (pr_reg)[8] = (regs)->r9; \
+ (pr_reg)[9] = (regs)->r8; \
+ (pr_reg)[10] = (regs)->rax; \
+ (pr_reg)[11] = (regs)->rcx; \
+ (pr_reg)[12] = (regs)->rdx; \
+ (pr_reg)[13] = (regs)->rsi; \
+ (pr_reg)[14] = (regs)->rdi; \
+ (pr_reg)[16] = (regs)->rip; \
+ (pr_reg)[17] = (regs)->cs; \
+ (pr_reg)[18] = (regs)->eflags; \
+ (pr_reg)[19] = (regs)->rsp; \
+ (pr_reg)[20] = (regs)->ss; \
+ asm("movl %%fs,%0" : "=r" (v)); (pr_reg)[21] = v; \
+ asm("movl %%gs,%0" : "=r" (v)); (pr_reg)[22] = v; \
+ asm("movl %%ds,%0" : "=r" (v)); (pr_reg)[23] = v; \
+ asm("movl %%es,%0" : "=r" (v)); (pr_reg)[24] = v; \
+ asm("movl %%fs,%0" : "=r" (v)); (pr_reg)[25] = v; \
+ asm("movl %%gs,%0" : "=r" (v)); (pr_reg)[26] = v; \
+} while(0);
#endif /* __X86_ELF_X86_64_H__ */
--- 0004/xen/include/asm-x86/x86_64/kexec.h
+++ work/xen/include/asm-x86/x86_64/kexec.h 2006-10-16 12:41:27.000000000 +0900
@@ -10,14 +10,43 @@
#ifndef __X86_64_KEXEC_H__
#define __X86_64_KEXEC_H__
-#include <xen/lib.h> /* for printk() used in stub */
+#include <xen/lib.h>
#include <xen/types.h>
#include <public/xen.h>
+/*
+ * Saving the registers of the cpu on which panic occured in
+ * crash_kexec to save a valid sp. The registers of other cpus
+ * will be saved in machine_crash_shutdown while shooting down them.
+ */
static void crash_setup_regs(struct cpu_user_regs *newregs,
struct cpu_user_regs *oldregs)
{
- printk("STUB: " __FILE__ ": %s: not implemented\n", __FUNCTION__);
+ if (oldregs)
+ memcpy(newregs, oldregs, sizeof(*newregs));
+ else {
+ __asm__ __volatile__("movq %%rbx,%0" : "=m"(newregs->rbx));
+ __asm__ __volatile__("movq %%rcx,%0" : "=m"(newregs->rcx));
+ __asm__ __volatile__("movq %%rdx,%0" : "=m"(newregs->rdx));
+ __asm__ __volatile__("movq %%rsi,%0" : "=m"(newregs->rsi));
+ __asm__ __volatile__("movq %%rdi,%0" : "=m"(newregs->rdi));
+ __asm__ __volatile__("movq %%rbp,%0" : "=m"(newregs->rbp));
+ __asm__ __volatile__("movq %%rax,%0" : "=m"(newregs->rax));
+ __asm__ __volatile__("movq %%rsp,%0" : "=m"(newregs->rsp));
+ __asm__ __volatile__("movq %%r8,%0" : "=m"(newregs->r8));
+ __asm__ __volatile__("movq %%r9,%0" : "=m"(newregs->r9));
+ __asm__ __volatile__("movq %%r10,%0" : "=m"(newregs->r10));
+ __asm__ __volatile__("movq %%r11,%0" : "=m"(newregs->r11));
+ __asm__ __volatile__("movq %%r12,%0" : "=m"(newregs->r12));
+ __asm__ __volatile__("movq %%r13,%0" : "=m"(newregs->r13));
+ __asm__ __volatile__("movq %%r14,%0" : "=m"(newregs->r14));
+ __asm__ __volatile__("movq %%r15,%0" : "=m"(newregs->r15));
+ __asm__ __volatile__("movl %%ss, %%eax;" :"=a"(newregs->ss));
+ __asm__ __volatile__("movl %%cs, %%eax;" :"=a"(newregs->cs));
+ __asm__ __volatile__("pushfq; popq %0" :"=m"(newregs->eflags));
+
+ newregs->rip = (unsigned long)current_text_addr();
+ }
}
#endif /* __X86_64_KEXEC_H__ */
--- 0004/xen/include/public/kexec.h
+++ work/xen/include/public/kexec.h 2006-10-16 12:41:27.000000000 +0900
@@ -50,6 +50,9 @@ typedef struct xen_kexec_image {
#if defined(__i386__) || defined(__x86_64__)
unsigned long page_list[KEXEC_XEN_NO_PAGES];
#endif
+#if defined(__x86_64__)
+ unsigned long page_table_b;
+#endif
unsigned long indirection_page;
unsigned long start_address;
} xen_kexec_image_t;
^ permalink raw reply [flat|nested] 30+ messages in thread