[PATCH][ia64][kdump] Fix problems on INIT

[PATCH][ia64][kdump] Fix problems on INIT

[Kenji Kaneshige]

Hi,

Here are patches for kdump on ia64 to fix the follwing problems:

  - [PATCH 1/2] Fix kernel panic in kdump on INIT

    I encountered the kernel panic when I sent an INIT without
    configuring kdump kernel image. The cause of this problem seems
    that machine_kdump_on_init() uses printk() in INIT
    context. Replacing it with ia64_mca_printk() fixes this problem.
    The panic message is attached at the end of this mail.

  - [PATCH 2/2] Fix kernel hangup in kdump on INIT

    Even after fixing the kernel panic problem mentioned above, I
    encontered the kernel hangup problem when I sent an INIT without
    configuring kdump kernel image. The kdump_init_notifier() on
    monarch CPU stops its operation if the kdump kernel image is not
    configured. On the other hand, kdump_init_notifier() on
    non-monarch CPUs get into spin because they don't know the fact
    the monarch stops its operation. This is the cause of this
    problem.

Thanks,
Kenji Kaneshige


machine_kdump_on_init(): kdump not configured
Unable to handle kernel NULL pointer dereference (address 0000000000000020)
INIT swapper 1[0]: Oops 11012296146944 [1]
Modules linked in: sunrpc binfmt_misc dm_mirror dm_mod fan sg eepro100 container e100 mii button thermal processor usb_storage ehci_hcd ohci_hcd

Pid: 0, CPU 1, comm:       INIT swapper 1
psr : 0000121808022030 ifs : 8000000000000813 ip  : [<a00000010006a3a1>]    Not tainted
ip is at try_to_wake_up+0x5c1/0x740
unat: 0000000000000000 pfs : 0000000000000813 rsc : 0000000000000003
rnat: 0000000000000000 bsps: 0000000000000000 pr  : 000000000000a685
ldrs: 0000000000000000 ccv : 0000000000000000 fpsr: 0009804c0270033f
csd : 0000000000000000 ssd : 0000000000000000
b0  : a00000010006a350 b6  : a000000100071540 b7  : e00000007fd08010
f6  : 1003e0000005b41d7540b f7  : 1003e00000000000005dc
f8  : 1003e000000002aaaaaab f9  : 1003e0000005b41d77388
f10 : 1003e35a3e8b515f27bd8 f11 : 1003e000000000000000f
r1  : a000000100bf9e20 r2  : 0000000000000000 r3  : a0000001009faa70
r8  : 0000003ba3d925e4 r9  : 0000003b9e207e0a r10 : e0000000010368d0
r11 : 0000003b9e207e0a r12 : e00000000115fa80 r13 : e000000001150000
r14 : 0000000000000400 r15 : 0000000000000400 r16 : e0000000010368c0
r17 : e000000001036ff8 r18 : e0000000010368c0 r19 : e00000007da10000
r20 : 0000000000000000 r21 : e000000001150000 r22 : e000000001036910
r23 : 0000000000000000 r24 : 0000000000000020 r25 : 0000000000000000
r26 : e000000001150040 r27 : e000000001150000 r28 : e000000001036fd8
r29 : a000000100a1a350 r30 : 0000000000000000 r31 : e000000001036880

Call Trace:
 [<a000000100013c20>] show_stack+0x40/0xa0
                                spà0000000115f650 bspà00000001151148
 [<a0000001000148a0>] show_regs+0x840/0x880
                                spà0000000115f820 bspà000000011510f0
 [<a000000100036e80>] die+0x1a0/0x2a0
                                spà0000000115f820 bspà000000011510a8
 [<a000000100061610>] ia64_do_page_fault+0x8d0/0xa00
                                spà0000000115f820 bspà00000001151058
 [<a00000010000b240>] ia64_leave_kernel+0x0/0x270
                                spà0000000115f8b0 bspà00000001151058
 [<a00000010006a3a0>] try_to_wake_up+0x5c0/0x740
                                spà0000000115fa80 bspà00000001150fb8
 [<a00000010006a550>] default_wake_function+0x30/0x60
                                spà0000000115fa90 bspà00000001150f88
 [<a0000001000a9670>] autoremove_wake_function+0x30/0xa0
                                spà0000000115fa90 bspà00000001150f50
 [<a0000001000641a0>] __wake_up_common+0x60/0x120
                                spà0000000115fa90 bspà00000001150f00
 [<a000000100067fd0>] __wake_up+0x50/0x80
                                spà0000000115fa90 bspà00000001150ec0
 [<a0000001000784b0>] wake_up_klogd+0x70/0xa0
                                spà0000000115fa90 bspà00000001150ea8
 [<a000000100079940>] release_console_sem+0x4a0/0x4c0
                                spà0000000115fa90 bspà00000001150e60
 [<a000000100078e40>] vprintk+0x7c0/0x940
                                spà0000000115fa90 bspà00000001150dd8
 [<a000000100079050>] printk+0x90/0xc0
                                spà0000000115fb00 bspà00000001150d78
 [<a00000010005f380>] machine_kdump_on_init+0x40/0xc0
                                spà0000000115fb50 bspà00000001150d60
 [<a00000010005f650>] kdump_init_notifier+0x250/0x280
                                spà0000000115fb50 bspà00000001150d30
 [<a000000100097c10>] notifier_call_chain+0x50/0xe0
                                spà0000000115fb50 bspà00000001150ce8
 [<a000000100097ce0>] __atomic_notifier_call_chain+0x40/0x60
                                spà0000000115fb50 bspà00000001150ca8
 [<a000000100097d40>] atomic_notifier_call_chain+0x40/0x60
                                spà0000000115fb50 bspà00000001150c78
 [<a0000001000b2160>] notify_die+0x80/0xa0
                                spà0000000115fb50 bspà00000001150c30
 [<a000000100049220>] ia64_init_handler+0x800/0x900
                                spà0000000115fb70 bspà00000001150bd8
 [<a000000100049b40>] ia64_os_init_virtual_begin+0x40/0x140
                                spà0000000115fb80 bspà00000001150bd8

[PATCH 1/2] Fix kernel panic in kdump on INIT

[Kenji Kaneshige]

Fix the problem that kdump on INIT causes a kernel panic if kdump
kernel image is not configured. The cause of this problem is
machine_kexec_on_init() is using printk in INIT context. It should
use ia64_mca_printk() instead.

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>

---
 arch/ia64/kernel/crash.c   |    2 +-
 arch/ia64/kernel/mca_drv.h |    2 --
 include/asm-ia64/mca.h     |    2 ++
 3 files changed, 3 insertions(+), 3 deletions(-)

Index: linux-2.6.23-rc4/arch/ia64/kernel/crash.c
=================================--- linux-2.6.23-rc4.orig/arch/ia64/kernel/crash.c
+++ linux-2.6.23-rc4/arch/ia64/kernel/crash.c
@@ -119,7 +119,7 @@ static void
 machine_kdump_on_init(void)
 {
 	if (!ia64_kimage) {
-		printk(KERN_NOTICE "machine_kdump_on_init(): "
+		ia64_mca_printk(KERN_NOTICE "machine_kdump_on_init(): "
 				"kdump not configured\n");
 		return;
 	}
Index: linux-2.6.23-rc4/arch/ia64/kernel/mca_drv.h
=================================--- linux-2.6.23-rc4.orig/arch/ia64/kernel/mca_drv.h
+++ linux-2.6.23-rc4/arch/ia64/kernel/mca_drv.h
@@ -118,7 +118,5 @@ struct mca_table_entry {
 
 extern const struct mca_table_entry *search_mca_tables (unsigned long addr);
 extern int mca_recover_range(unsigned long);
-extern void ia64_mca_printk(const char * fmt, ...)
-	 __attribute__ ((format (printf, 1, 2)));
 extern void ia64_mlogbuf_dump(void);
 
Index: linux-2.6.23-rc4/include/asm-ia64/mca.h
=================================--- linux-2.6.23-rc4.orig/include/asm-ia64/mca.h
+++ linux-2.6.23-rc4/include/asm-ia64/mca.h
@@ -151,6 +151,8 @@ extern void ia64_mca_cmc_vector_setup(vo
 extern int  ia64_reg_MCA_extension(int (*fn)(void *, struct ia64_sal_os_state *));
 extern void ia64_unreg_MCA_extension(void);
 extern u64 ia64_get_rnat(u64 *);
+extern void ia64_mca_printk(const char * fmt, ...)
+	 __attribute__ ((format (printf, 1, 2)));
 
 struct ia64_mca_notify_die {
 	struct ia64_sal_os_state *sos;

[PATCH 2/2] Fix kernel hangup in kdump on INIT

[Kenji Kaneshige]

Fix the problem that kdump on INIT hung up if kdump kernel image is
not configured.

The kdump_init_notifier() on monarch CPU stops its operation at
DIE_INIT_MONARCH_LEAVE time if the kdump kernel image is not
configured. On the other hand, kdump_init_notifier() on non-monarch
CPUs get into spin because they don't know the fact the monarch stops
its operation. This is the cause of this problem. To fix this problem,
we need to check the kdump kernel image at the top of the
kdump_init_notifier() function.

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>

---
 arch/ia64/kernel/crash.c |   13 ++++++++-----
 1 files changed, 8 insertions(+), 5 deletions(-)

Index: linux-2.6.23-rc4/arch/ia64/kernel/crash.c
=================================--- linux-2.6.23-rc4.orig/arch/ia64/kernel/crash.c
+++ linux-2.6.23-rc4/arch/ia64/kernel/crash.c
@@ -118,11 +118,6 @@ machine_crash_shutdown(struct pt_regs *p
 static void
 machine_kdump_on_init(void)
 {
-	if (!ia64_kimage) {
-		ia64_mca_printk(KERN_NOTICE "machine_kdump_on_init(): "
-				"kdump not configured\n");
-		return;
-	}
 	local_irq_disable();
 	kexec_disable_iosapic();
 	machine_kexec(ia64_kimage);
@@ -156,6 +151,14 @@ kdump_init_notifier(struct notifier_bloc
 	if (!kdump_on_init)
 		return NOTIFY_DONE;
 
+	if (!ia64_kimage) {
+		if (val = DIE_INIT_MONARCH_LEAVE)
+			ia64_mca_printk(KERN_NOTICE
+					"%s: kdump not configured\n",
+					__FUNCTION__);
+		return NOTIFY_DONE;
+	}
+
 	if (val != DIE_INIT_MONARCH_LEAVE &&
 	    val != DIE_INIT_SLAVE_LEAVE &&
 	    val != DIE_INIT_MONARCH_PROCESS &&