From: Brian Gerst <brgerst@gmail.com>
To: Tejun Heo <tj@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>,
linux-kernel@vger.kernel.org, Brian Gerst <brgerst@gmail.com>
Subject: [PATCH 4/5] x86-64: Remove the PDA
Date: Sun, 18 Jan 2009 19:52:24 -0500 [thread overview]
Message-ID: <1232326345-3534-4-git-send-email-brgerst@gmail.com> (raw)
In-Reply-To: <1232326345-3534-3-git-send-email-brgerst@gmail.com>
Now that the PDA is empty except for the stack canary, it can be removed.
The irqstack is moved to the start of the per-cpu section. If the stack
protector is enabled, the canary overlaps the bottom 48 bytes of the irqstack.
Signed-off-by: Brian Gerst <brgerst@gmail.com>
---
arch/x86/include/asm/pda.h | 5 -----
arch/x86/include/asm/percpu.h | 6 ------
arch/x86/include/asm/processor.h | 25 ++++++++++++++++++++++++-
arch/x86/kernel/asm-offsets_64.c | 4 ----
arch/x86/kernel/cpu/common.c | 10 +++-------
arch/x86/kernel/head_64.S | 13 +++++--------
arch/x86/kernel/process_64.c | 6 +++---
arch/x86/kernel/setup_percpu.c | 34 ++++------------------------------
arch/x86/kernel/vmlinux_64.lds.S | 8 ++++++--
9 files changed, 45 insertions(+), 66 deletions(-)
diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h
index 6ca7bc0..ba46416 100644
--- a/arch/x86/include/asm/pda.h
+++ b/arch/x86/include/asm/pda.h
@@ -17,11 +17,6 @@ struct x8664_pda {
unsigned long unused4;
int unused5;
unsigned int unused6; /* 36 was cpunumber */
-#ifdef CONFIG_CC_STACKPROTECTOR
- unsigned long stack_canary; /* 40 stack canary value */
- /* gcc-ABI: this canary MUST be at
- offset 40!!! */
-#endif
short in_bootmem; /* pda lives in bootmem */
} ____cacheline_aligned_in_smp;
diff --git a/arch/x86/include/asm/percpu.h b/arch/x86/include/asm/percpu.h
index 165d527..ce980db 100644
--- a/arch/x86/include/asm/percpu.h
+++ b/arch/x86/include/asm/percpu.h
@@ -133,12 +133,6 @@ do { \
/* We can use this directly for local CPU (faster). */
DECLARE_PER_CPU(unsigned long, this_cpu_off);
-#ifdef CONFIG_X86_64
-extern void load_pda_offset(int cpu);
-#else
-static inline void load_pda_offset(int cpu) { }
-#endif
-
#endif /* !__ASSEMBLY__ */
#ifdef CONFIG_SMP
diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index f511246..cfe3237 100644
--- a/arch/x86/include/asm/processor.h
+++ b/arch/x86/include/asm/processor.h
@@ -379,8 +379,31 @@ union thread_xstate {
#ifdef CONFIG_X86_64
DECLARE_PER_CPU(struct orig_ist, orig_ist);
-DECLARE_PER_CPU(char[IRQ_STACK_SIZE], irq_stack);
+union irq_stack_union {
+ char irq_stack[IRQ_STACK_SIZE];
+ /*
+ * GCC hardcodes the stack canary as %gs:40. Since the
+ * irq_stack is the object at %gs:0, we reserve the bottom
+ * 48 bytes of the irq stack for the canary.
+ */
+ struct {
+ char gs_base[40];
+#ifdef CONFIG_CC_STACKPROTECTOR
+ unsigned long stack_canary;
+#endif
+ };
+};
+
+DECLARE_PER_CPU(union irq_stack_union, irq_stack_union);
DECLARE_PER_CPU(char *, irq_stack_ptr);
+
+static inline void load_gs_base(int cpu)
+{
+ /* Memory clobbers used to order pda/percpu accesses */
+ mb();
+ wrmsrl(MSR_GS_BASE, (unsigned long)per_cpu(irq_stack_union.gs_base, cpu));
+ mb();
+}
#endif
extern void print_cpu_info(struct cpuinfo_x86 *);
diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c
index 64c834a..94f9c8b 100644
--- a/arch/x86/kernel/asm-offsets_64.c
+++ b/arch/x86/kernel/asm-offsets_64.c
@@ -48,10 +48,6 @@ int main(void)
#endif
BLANK();
#undef ENTRY
-#define ENTRY(entry) DEFINE(pda_ ## entry, offsetof(struct x8664_pda, entry))
- DEFINE(pda_size, sizeof(struct x8664_pda));
- BLANK();
-#undef ENTRY
#ifdef CONFIG_PARAVIRT
BLANK();
OFFSET(PARAVIRT_enabled, pv_info, paravirt_enabled);
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index f83a4d6..94ea5cb 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -881,13 +881,9 @@ __setup("clearcpuid=", setup_disablecpuid);
#ifdef CONFIG_X86_64
struct desc_ptr idt_descr = { 256 * 16 - 1, (unsigned long) idt_table };
-DEFINE_PER_CPU_PAGE_ALIGNED(char[IRQ_STACK_SIZE], irq_stack);
-#ifdef CONFIG_SMP
-DEFINE_PER_CPU(char *, irq_stack_ptr); /* will be set during per cpu init */
-#else
+DEFINE_PER_CPU_FIRST(union irq_stack_union, irq_stack_union) __aligned(PAGE_SIZE);
DEFINE_PER_CPU(char *, irq_stack_ptr) =
- per_cpu_var(irq_stack) + IRQ_STACK_SIZE - 64;
-#endif
+ per_cpu_var(irq_stack_union.irq_stack) + IRQ_STACK_SIZE - 64;
DEFINE_PER_CPU(unsigned long, kernel_stack) =
(unsigned long)&init_thread_union - KERNEL_STACK_OFFSET + THREAD_SIZE;
@@ -960,7 +956,7 @@ void __cpuinit cpu_init(void)
loadsegment(fs, 0);
loadsegment(gs, 0);
- load_pda_offset(cpu);
+ load_gs_base(cpu);
#ifdef CONFIG_NUMA
if (cpu != 0 && percpu_read(node_number) == 0 &&
diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S
index 98ea26a..8c83de6 100644
--- a/arch/x86/kernel/head_64.S
+++ b/arch/x86/kernel/head_64.S
@@ -216,6 +216,7 @@ ENTRY(secondary_startup_64)
cmpl $0, per_cpu__cpu_number(%rax)
jne 1f
addq %rax, early_gdt_descr_base(%rip)
+ addq %rax, per_cpu__irq_stack_ptr(%rax)
1:
#endif
/*
@@ -242,13 +243,9 @@ ENTRY(secondary_startup_64)
/* Set up %gs.
*
- * On SMP, %gs should point to the per-cpu area. For initial
- * boot, make %gs point to the init data section. For a
- * secondary CPU,initial_gs should be set to its pda address
- * before the CPU runs this code.
- *
- * On UP, initial_gs points to PER_CPU_VAR(__pda) and doesn't
- * change.
+ * The base of %gs always points to the bottom of the irqstack
+ * union. If the stack protector canary is enabled, it is located
+ * at %gs:40.
*/
movl $MSR_GS_BASE,%ecx
movq initial_gs(%rip),%rax
@@ -281,7 +278,7 @@ ENTRY(secondary_startup_64)
#ifdef CONFIG_SMP
.quad __per_cpu_load
#else
- .quad PER_CPU_VAR(__pda)
+ .quad per_cpu__irq_stack_union
#endif
__FINITDATA
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 4523ff8..604b69e 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -627,12 +627,12 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
(unsigned long)task_stack_page(next_p) +
THREAD_SIZE - KERNEL_STACK_OFFSET);
#ifdef CONFIG_CC_STACKPROTECTOR
- write_pda(stack_canary, next_p->stack_canary);
+ percpu_write(irq_stack_union.stack_canary, canary);
/*
* Build time only check to make sure the stack_canary is at
- * offset 40 in the pda; this is a gcc ABI requirement
+ * %gs:40; this is a gcc ABI requirement
*/
- BUILD_BUG_ON(offsetof(struct x8664_pda, stack_canary) != 40);
+ BUILD_BUG_ON(offsetof(union irq_stack_union, stack_canary) != 40);
#endif
/*
diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c
index efbafbb..65c10c4 100644
--- a/arch/x86/kernel/setup_percpu.c
+++ b/arch/x86/kernel/setup_percpu.c
@@ -77,30 +77,6 @@ static void __init setup_node_to_cpumask_map(void);
static inline void setup_node_to_cpumask_map(void) { }
#endif
-/*
- * Define load_pda_offset() and per-cpu __pda for x86_64.
- * load_pda_offset() is responsible for loading the offset of pda into
- * %gs.
- *
- * On SMP, pda offset also duals as percpu base address and thus it
- * should be at the start of per-cpu area. To achieve this, it's
- * preallocated in vmlinux_64.lds.S directly instead of using
- * DEFINE_PER_CPU().
- */
-#ifdef CONFIG_X86_64
-void __cpuinit load_pda_offset(int cpu)
-{
- /* Memory clobbers used to order pda/percpu accesses */
- mb();
- wrmsrl(MSR_GS_BASE, cpu_pda(cpu));
- mb();
-}
-#ifndef CONFIG_SMP
-DEFINE_PER_CPU(struct x8664_pda, __pda);
-#endif
-EXPORT_PER_CPU_SYMBOL(__pda);
-#endif /* CONFIG_SMP && CONFIG_X86_64 */
-
#ifdef CONFIG_X86_64
/* correctly size the local cpu masks */
@@ -207,15 +183,13 @@ void __init setup_per_cpu_areas(void)
per_cpu(cpu_number, cpu) = cpu;
#ifdef CONFIG_X86_64
per_cpu(irq_stack_ptr, cpu) =
- (char *)per_cpu(irq_stack, cpu) + IRQ_STACK_SIZE - 64;
+ per_cpu(irq_stack_union.irq_stack, cpu) + IRQ_STACK_SIZE - 64;
/*
- * CPU0 modified pda in the init data area, reload pda
- * offset for CPU0 and clear the area for others.
+ * CPU0 modified data in the init per-cpu area, reload %gs
+ * offset for CPU0.
*/
if (cpu == 0)
- load_pda_offset(0);
- else
- memset(cpu_pda(cpu), 0, sizeof(*cpu_pda(cpu)));
+ load_gs_base(cpu);
#endif
DBG("PERCPU: cpu %4d %p\n", cpu, ptr);
diff --git a/arch/x86/kernel/vmlinux_64.lds.S b/arch/x86/kernel/vmlinux_64.lds.S
index a09abb8..c974099 100644
--- a/arch/x86/kernel/vmlinux_64.lds.S
+++ b/arch/x86/kernel/vmlinux_64.lds.S
@@ -220,8 +220,7 @@ SECTIONS
* so that it can be accessed as a percpu variable.
*/
. = ALIGN(PAGE_SIZE);
- PERCPU_VADDR_PREALLOC(0, :percpu, pda_size)
- per_cpu____pda = __per_cpu_start;
+ PERCPU_VADDR(0, :percpu)
#else
PERCPU(PAGE_SIZE)
#endif
@@ -262,3 +261,8 @@ SECTIONS
*/
ASSERT((_end - _text <= KERNEL_IMAGE_SIZE),
"kernel image bigger than KERNEL_IMAGE_SIZE")
+
+#ifdef CONFIG_SMP
+ASSERT((per_cpu__irq_stack_union == 0),
+ "irq_stack_union is not at start of per-cpu area");
+#endif
--
1.6.1.rc1
next prev parent reply other threads:[~2009-01-19 0:53 UTC|newest]
Thread overview: 39+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-01-18 16:39 [PATCHSET x86:core/percpu] percpu: move PDA fields to percpu Tejun Heo
2009-01-18 16:39 ` [PATCH 01/12] x86-64: Move irq stats from PDA to per-cpu and consolidate with 32-bit Tejun Heo
2009-01-18 16:39 ` [PATCH 02/12] x86-64: Move TLB state " Tejun Heo
2009-01-18 16:39 ` [PATCH 03/12] x86-64: Convert irqstacks to per-cpu Tejun Heo
2009-01-18 18:16 ` Brian Gerst
2009-01-18 23:38 ` Tejun Heo
2009-01-18 23:43 ` Ingo Molnar
2009-01-19 0:52 ` Tejun Heo
2009-01-18 16:39 ` [PATCH 04/12] x86-64: Convert exception stacks " Tejun Heo
2009-01-18 16:39 ` [PATCH 05/12] x86-64: Move cpu number from PDA to per-cpu and consolidate with 32-bit Tejun Heo
2009-01-18 16:39 ` [PATCH 06/12] x86-64: Move current task " Tejun Heo
2009-01-18 16:39 ` [PATCH 07/12] x86-64: Move kernelstack from PDA to per-cpu Tejun Heo
2009-01-18 16:39 ` [PATCH 08/12] x86-64: Move oldrsp " Tejun Heo
2009-01-18 16:39 ` [PATCH 09/12] x86-64: Move irqcount " Tejun Heo
2009-01-18 16:39 ` [PATCH 10/12] x86-64: Move nodenumber " Tejun Heo
2009-01-18 16:39 ` [PATCH 11/12] x86-64: Move isidle " Tejun Heo
2009-01-18 16:39 ` [PATCH 12/12] x86-64: Use absolute displacements for per-cpu accesses Tejun Heo
2009-01-18 16:49 ` [PATCHSET x86:core/percpu] percpu: move PDA fields to percpu Ingo Molnar
2009-01-19 0:51 ` Brian Gerst
2009-01-19 0:52 ` [PATCH 1/5] x86-64: Remove pda_init() Brian Gerst
2009-01-19 0:52 ` [PATCH 2/5] percpu: Refactor percpu.h Brian Gerst
2009-01-19 0:52 ` [PATCH 3/5] x86-64: Rework __per_cpu_load adjustments Brian Gerst
2009-01-19 0:52 ` Brian Gerst [this message]
2009-01-19 0:52 ` [PATCH 5/5] x86-64: Remove pda.h Brian Gerst
2009-01-19 2:46 ` [PATCH 6/6] linker script: kill PERCPU_VADDR_PREALLOC() Tejun Heo
2009-01-19 2:18 ` [PATCH 4/5] x86-64: Remove the PDA Tejun Heo
2009-01-19 2:52 ` Brian Gerst
2009-01-19 3:05 ` Tejun Heo
2009-01-19 2:46 ` [PATCH UPDATED 4/5] x86-64: Move stack_canary into irq_stack Tejun Heo
2009-02-05 22:30 ` [PATCH 2/5] percpu: Refactor percpu.h Tony Luck
2009-02-06 11:11 ` Brian Gerst
2009-02-06 19:06 ` Luck, Tony
2009-01-19 9:53 ` [PATCH 1/5] x86-64: Remove pda_init() Peter Zijlstra
2009-01-19 2:45 ` [PATCHSET x86:core/percpu] percpu: move PDA fields to percpu Tejun Heo
2009-01-19 11:19 ` Ingo Molnar
2009-01-19 11:35 ` Ingo Molnar
2009-01-19 13:43 ` Tejun Heo
2009-01-20 3:34 ` Tejun Heo
2009-01-20 7:24 ` Ingo Molnar
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1232326345-3534-4-git-send-email-brgerst@gmail.com \
--to=brgerst@gmail.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=tj@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.