Skip to contentSkip to conversation

x86, traps: Enable DEBUG_STACK after cpu_init() for TRAP_DB/BP.

PATCH In review Wang Nan 4 participants 11 messages Feb 26, 2015 – Mar 4, 2015 x86

lore
Share & tools
lore

Wang Nan

author
#0 lore raw

PATCH v2 1/1 x86, traps: Enable DEBUG_STACK after cpu_init() for TRAP_DB/BP.

Before this patch early_trap_init() installs DEBUG_STACK for X86_TRAP_BP and X86_TRAP_DB. However, DEBUG_STACK doesn't work correctly until cpu_init() <-- trap_init().

This patch passes 0 to set_intr_gate_ist() and set_system_intr_gate_ist() instead of DEBUG_STACK to let it use same stack as kernel, and installs DEBUG_STACK for them in trap_init().

As core runs at ring 0 between early_trap_init() and trap_init(), there is no chance to get a bad stack before trap_init().

As NMI is also enabled in trap_init(), we don't need to care about is_debug_stack() and related things used in arch/x86/kernel/nmi.c.

---
v1 -> v2: Correct grammar issues in comments.

---
 arch/x86/kernel/traps.c | 21 +++++++++++++++++++--
 1 file changed, 19 insertions(+), 2 deletions(-)
arch/x86/kernel/traps.c +19 -2
index 9d2073e..4281988 100644
@@ -925,9 +925,17 @@ dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code)
 /* Set of traps needed for early debugging. */
 void __init early_trap_init(void)
 {
-	set_intr_gate_ist(X86_TRAP_DB, &debug, DEBUG_STACK);
+	/*
+	 * Don't set ist to DEBUG_STACK as it doesn't work until TSS is
+	 * ready in cpu_init() <-- trap_init(). Before trap_init(), CPU
+	 * runs at ring 0 so it is impossible to hit an invalid stack.
+	 * Using the original stack works well enough at this early
+	 * stage. DEBUG_STACK will be equipped after cpu_init() in
+	 * trap_init().
+	 */
+	set_intr_gate_ist(X86_TRAP_DB, &debug, 0);
 	/* int3 can be called from all */
-	set_system_intr_gate_ist(X86_TRAP_BP, &int3, DEBUG_STACK);
+	set_system_intr_gate_ist(X86_TRAP_BP, &int3, 0);
 #ifdef CONFIG_X86_32
 	set_intr_gate(X86_TRAP_PF, page_fault);
 #endif
@@ -1005,6 +1013,15 @@ void __init trap_init(void)
 	 */
 	cpu_init();
 
+	/*
+	 * X86_TRAP_DB and X86_TRAP_BP have been set
+	 * in early_trap_init(). However, DEBUG_STACK works only after
+	 * cpu_init() loads TSS. See comments in early_trap_init().
+	 */
+	set_intr_gate_ist(X86_TRAP_DB, &debug, DEBUG_STACK);
+	/* int3 can be called from all */
+	set_system_intr_gate_ist(X86_TRAP_BP, &int3, DEBUG_STACK);
+
 	x86_init.irqs.trap_init();
 
 #ifdef CONFIG_X86_64
Signature · 1 line
1.8.4
#1 lore raw

[tip:x86/asm] x86/traps: Enable DEBUG_STACK after cpu_init() for TRAP_DB/BP

x86/traps: Enable DEBUG_STACK after cpu_init() for TRAP_DB/BP
Commit-ID:  b4d8327024637cb2a1f7910dcb5d0ad7a096f473
Gitweb:     http://git.kernel.org/tip/b4d8327024637cb2a1f7910dcb5d0ad7a096f473
Author:     Wang Nan <[email protected]>
AuthorDate: Thu, 26 Feb 2015 13:49:39 +0800
Committer:  Ingo Molnar <[email protected]>
CommitDate: Thu, 26 Feb 2015 12:29:20 +0100

x86/traps: Enable DEBUG_STACK after cpu_init() for TRAP_DB/BP

Before this patch early_trap_init() installs DEBUG_STACK for X86_TRAP_BP and X86_TRAP_DB. However, DEBUG_STACK doesn't work correctly until cpu_init() <-- trap_init().

This patch passes 0 to set_intr_gate_ist() and set_system_intr_gate_ist() instead of DEBUG_STACK to let it use same stack as kernel, and installs DEBUG_STACK for them in trap_init().

As core runs at ring 0 between early_trap_init() and trap_init(), there is no chance to get a bad stack before trap_init().

As NMI is also enabled in trap_init(), we don't need to care about is_debug_stack() and related things used in arch/x86/kernel/nmi.c.

---
 arch/x86/kernel/traps.c | 21 +++++++++++++++++++--
 1 file changed, 19 insertions(+), 2 deletions(-)
arch/x86/kernel/traps.c +19 -2
index 9d2073e..4281988 100644
@@ -925,9 +925,17 @@ dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code)
 /* Set of traps needed for early debugging. */
 void __init early_trap_init(void)
 {
-	set_intr_gate_ist(X86_TRAP_DB, &debug, DEBUG_STACK);
+	/*
+	 * Don't set ist to DEBUG_STACK as it doesn't work until TSS is
+	 * ready in cpu_init() <-- trap_init(). Before trap_init(), CPU
+	 * runs at ring 0 so it is impossible to hit an invalid stack.
+	 * Using the original stack works well enough at this early
+	 * stage. DEBUG_STACK will be equipped after cpu_init() in
+	 * trap_init().
+	 */
+	set_intr_gate_ist(X86_TRAP_DB, &debug, 0);
 	/* int3 can be called from all */
-	set_system_intr_gate_ist(X86_TRAP_BP, &int3, DEBUG_STACK);
+	set_system_intr_gate_ist(X86_TRAP_BP, &int3, 0);
 #ifdef CONFIG_X86_32
 	set_intr_gate(X86_TRAP_PF, page_fault);
 #endif
@@ -1005,6 +1013,15 @@ void __init trap_init(void)
 	 */
 	cpu_init();
 
+	/*
+	 * X86_TRAP_DB and X86_TRAP_BP have been set
+	 * in early_trap_init(). However, DEBUG_STACK works only after
+	 * cpu_init() loads TSS. See comments in early_trap_init().
+	 */
+	set_intr_gate_ist(X86_TRAP_DB, &debug, DEBUG_STACK);
+	/* int3 can be called from all */
+	set_system_intr_gate_ist(X86_TRAP_BP, &int3, DEBUG_STACK);
+
 	x86_init.irqs.trap_init();
 
 #ifdef CONFIG_X86_64
#2 lore raw

Re: [tip:x86/asm] x86/traps: Enable DEBUG_STACK after cpu_init() for TRAP_DB/BP

Quoted 9 lines · tip-bot for Wang Nan
Commit-ID:  b4d8327024637cb2a1f7910dcb5d0ad7a096f473
Gitweb:     http://git.kernel.org/tip/b4d8327024637cb2a1f7910dcb5d0ad7a096f473
Author:     Wang Nan <[email protected]>
AuthorDate: Thu, 26 Feb 2015 13:49:39 +0800
Committer:  Ingo Molnar <[email protected]>
CommitDate: Thu, 26 Feb 2015 12:29:20 +0100

x86/traps: Enable DEBUG_STACK after cpu_init() for TRAP_DB/BP

Quoted 9 lines

Before this patch early_trap_init() installs DEBUG_STACK for X86_TRAP_BP and X86_TRAP_DB. However, DEBUG_STACK doesn't work correctly until cpu_init() <-- trap_init().

This patch passes 0 to set_intr_gate_ist() and set_system_intr_gate_ist() instead of DEBUG_STACK to let it use same stack as kernel, and installs DEBUG_STACK for them in trap_init().

Sorry, I'm late to the party. This patch, while technically correct AFAICT, is really ugly, because the whole point is that it *doesn't* use ist. In other words:

Quoted 1 line

+ set_intr_gate_ist(X86_TRAP_DB, &debug, 0);

That should be set_intr_gate(X86_TRAP_DB, &debug);

arch/x86/kernel/traps.c dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code) · line 938

Quoted 3 lines
        /* int3 can be called from all */
arch/x86/kernel/traps.c +1 -1
-       set_system_intr_gate_ist(X86_TRAP_BP, &int3, DEBUG_STACK);
+       set_system_intr_gate_ist(X86_TRAP_BP, &int3, 0);

Similarly, this should be set_system_gate.

arch/x86/kernel/traps.c void __init trap_init(void) · line 1019

Quoted 11 lines
 #ifdef CONFIG_X86_32
        set_intr_gate(X86_TRAP_PF, page_fault);

#endif

arch/x86/kernel/traps.c +4 -0
@@ -1005,6 +1013,15 @@ void __init trap_init(void)
         */
        cpu_init();
 
+       /*
+        * X86_TRAP_DB and X86_TRAP_BP have been set
+        * in early_trap_init(). However, DEBUG_STACK works only after
+        * cpu_init() loads TSS. See comments in early_trap_init().

It's not that DEBUG_STACK only works after the TSS is loaded; it's that the IST mechanism only works after TSS is loaded.

--Andy

author
#3 lore raw

Re: [tip:x86/asm] x86/traps: Enable DEBUG_STACK after cpu_init() for TRAP_DB/BP

Quoted 20 lines · Andy Lutomirski
Quoted 9 lines · tip-bot for Wang Nan
Commit-ID:  b4d8327024637cb2a1f7910dcb5d0ad7a096f473
Gitweb:     http://git.kernel.org/tip/b4d8327024637cb2a1f7910dcb5d0ad7a096f473
Author:     Wang Nan <[email protected]>
AuthorDate: Thu, 26 Feb 2015 13:49:39 +0800
Committer:  Ingo Molnar <[email protected]>
CommitDate: Thu, 26 Feb 2015 12:29:20 +0100

x86/traps: Enable DEBUG_STACK after cpu_init() for TRAP_DB/BP

Quoted 9 lines

Before this patch early_trap_init() installs DEBUG_STACK for X86_TRAP_BP and X86_TRAP_DB. However, DEBUG_STACK doesn't work correctly until cpu_init() <-- trap_init().

This patch passes 0 to set_intr_gate_ist() and set_system_intr_gate_ist() instead of DEBUG_STACK to let it use same stack as kernel, and installs DEBUG_STACK for them in trap_init().

Quoted 8 lines

Sorry, I'm late to the party. This patch, while technically correct AFAICT, is really ugly, because the whole point is that it *doesn't* use ist. In other words:

Quoted 1 line

+ set_intr_gate_ist(X86_TRAP_DB, &debug, 0);

That should be set_intr_gate(X86_TRAP_DB, &debug);

I considered this, but found that set_intr_gate() and set_intr_gate_ist(..., 0) behaviors differently: set_intr_gate() -> _trace_set_gate() requires 'trace_##addr' to be installed to trace_idt_table, while set_intr_gate_ist(..., 0) puts 'addr'. into trace_idt_table() through _set_gate().
Therefore, if we want to use set_intr_gate() we need to provide a trace_debug entry for it, which will be overwritten in trap_init() so it is in fact a useless symbol.

Anyway, I'd like to post a v3 patch follow your advise. Please keep an eye on it.

Thank you!

Quoted 23 lines

arch/x86/kernel/traps.c dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code) · line 938

Quoted 3 lines
        /* int3 can be called from all */
arch/x86/kernel/traps.c +1 -1
-       set_system_intr_gate_ist(X86_TRAP_BP, &int3, DEBUG_STACK);
+       set_system_intr_gate_ist(X86_TRAP_BP, &int3, 0);

Similarly, this should be set_system_gate.

arch/x86/kernel/traps.c void __init trap_init(void) · line 1019

Quoted 11 lines
 #ifdef CONFIG_X86_32
        set_intr_gate(X86_TRAP_PF, page_fault);

#endif

arch/x86/kernel/traps.c +4 -0
@@ -1005,6 +1013,15 @@ void __init trap_init(void)
         */
        cpu_init();
 
+       /*
+        * X86_TRAP_DB and X86_TRAP_BP have been set
+        * in early_trap_init(). However, DEBUG_STACK works only after
+        * cpu_init() loads TSS. See comments in early_trap_init().

It's not that DEBUG_STACK only works after the TSS is loaded; it's that the IST mechanism only works after TSS is loaded.

--Andy

#4 lore raw

Re: [tip:x86/asm] x86/traps: Enable DEBUG_STACK after cpu_init() for TRAP_DB/BP

Quoted 30 lines · Wang Nan
Quoted 20 lines · Andy Lutomirski
Quoted 9 lines · tip-bot for Wang Nan
Commit-ID:  b4d8327024637cb2a1f7910dcb5d0ad7a096f473
Gitweb:     http://git.kernel.org/tip/b4d8327024637cb2a1f7910dcb5d0ad7a096f473
Author:     Wang Nan <[email protected]>
AuthorDate: Thu, 26 Feb 2015 13:49:39 +0800
Committer:  Ingo Molnar <[email protected]>
CommitDate: Thu, 26 Feb 2015 12:29:20 +0100

x86/traps: Enable DEBUG_STACK after cpu_init() for TRAP_DB/BP

Quoted 9 lines

Before this patch early_trap_init() installs DEBUG_STACK for X86_TRAP_BP and X86_TRAP_DB. However, DEBUG_STACK doesn't work correctly until cpu_init() <-- trap_init().

This patch passes 0 to set_intr_gate_ist() and set_system_intr_gate_ist() instead of DEBUG_STACK to let it use same stack as kernel, and installs DEBUG_STACK for them in trap_init().

Quoted 8 lines

Sorry, I'm late to the party. This patch, while technically correct AFAICT, is really ugly, because the whole point is that it *doesn't* use ist. In other words:

Quoted 1 line

+ set_intr_gate_ist(X86_TRAP_DB, &debug, 0);

That should be set_intr_gate(X86_TRAP_DB, &debug);

Quoted 6 lines

I considered this, but found that set_intr_gate() and set_intr_gate_ist(..., 0) behaviors differently: set_intr_gate() -> _trace_set_gate() requires 'trace_##addr' to be installed to trace_idt_table, while set_intr_gate_ist(..., 0) puts 'addr'. into trace_idt_table() through _set_gate().
Therefore, if we want to use set_intr_gate() we need to provide a trace_debug entry for it, which will be overwritten in trap_init() so it is in fact a useless symbol.

Yuck. Then IMO we should have a separate helper for this. Better yet, we should rename set_intr_gate, because this distinction could easily be overlooked.

Quoted 30 lines

Anyway, I'd like to post a v3 patch follow your advise. Please keep an eye on it.

Thank you!

Quoted 23 lines

arch/x86/kernel/traps.c dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code) · line 938

Quoted 3 lines
        /* int3 can be called from all */
arch/x86/kernel/traps.c +1 -1
-       set_system_intr_gate_ist(X86_TRAP_BP, &int3, DEBUG_STACK);
+       set_system_intr_gate_ist(X86_TRAP_BP, &int3, 0);

Similarly, this should be set_system_gate.

arch/x86/kernel/traps.c void __init trap_init(void) · line 1019

Quoted 11 lines
 #ifdef CONFIG_X86_32
        set_intr_gate(X86_TRAP_PF, page_fault);

#endif

arch/x86/kernel/traps.c +4 -0
@@ -1005,6 +1013,15 @@ void __init trap_init(void)
         */
        cpu_init();
 
+       /*
+        * X86_TRAP_DB and X86_TRAP_BP have been set
+        * in early_trap_init(). However, DEBUG_STACK works only after
+        * cpu_init() loads TSS. See comments in early_trap_init().

It's not that DEBUG_STACK only works after the TSS is loaded; it's that the IST mechanism only works after TSS is loaded.

--Andy

Signature · 2 lines
Andy Lutomirski
AMA Capital Management, LLC
author
#5 lore raw

PATCH 1/1 x86, traps: early_trap_init() cleanup.

As early_trap_init() doesn't use IST, replace set_intr_gate_ist(..., 0) and set_system_intr_gate_ist(..., 0) with their standard counterparts.

set_intr_gate() requires a trace_debug symbol which we don't have and won't use. Use a small macro trick as a workaround.

---

Hi Andy Lutomirski,

This patch is a bit tricky, but I think we don't need to define another helper for such a small problem. What's your opinion?

Thank you!

---
 arch/x86/kernel/traps.c | 23 ++++++++++++++---------
 1 file changed, 14 insertions(+), 9 deletions(-)
arch/x86/kernel/traps.c +14 -9
index 4281988..c24434a 100644
@@ -925,17 +925,22 @@ dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code)
 /* Set of traps needed for early debugging. */
 void __init early_trap_init(void)
 {
+
 	/*
-	 * Don't set ist to DEBUG_STACK as it doesn't work until TSS is
-	 * ready in cpu_init() <-- trap_init(). Before trap_init(), CPU
-	 * runs at ring 0 so it is impossible to hit an invalid stack.
-	 * Using the original stack works well enough at this early
-	 * stage. DEBUG_STACK will be equipped after cpu_init() in
-	 * trap_init().
+	 * Don't use IST to set DEBUG_STACK as it doesn't work until TSS is
+	 * ready in cpu_init() <-- trap_init(). Before trap_init(), CPU runs at
+	 * ring 0 so it is impossible to hit an invalid stack.  Using the
+	 * original stack works well enough at this early stage. DEBUG_STACK
+	 * will be equipped after cpu_init() in trap_init().
+	 *
+	 * Since set_intr_gate() needs a trace_debug but we don't have it,
+	 * use the following #define as a workaround.
 	 */
-	set_intr_gate_ist(X86_TRAP_DB, &debug, 0);
+#define trace_debug debug
+	set_intr_gate(X86_TRAP_DB, debug);
+#undef trace_debug
 	/* int3 can be called from all */
-	set_system_intr_gate_ist(X86_TRAP_BP, &int3, 0);
+	set_system_intr_gate(X86_TRAP_BP, &int3);
 #ifdef CONFIG_X86_32
 	set_intr_gate(X86_TRAP_PF, page_fault);
 #endif
@@ -1015,7 +1020,7 @@ void __init trap_init(void)
 
 	/*
 	 * X86_TRAP_DB and X86_TRAP_BP have been set
-	 * in early_trap_init(). However, DEBUG_STACK works only after
+	 * in early_trap_init(). However, ITS works only after
 	 * cpu_init() loads TSS. See comments in early_trap_init().
 	 */
 	set_intr_gate_ist(X86_TRAP_DB, &debug, DEBUG_STACK);
Signature · 1 line
1.8.4
maintainer
#6 lore raw

Re: [PATCH] x86, traps: early_trap_init() cleanup.

Quoted 37 lines · Wang Nan

As early_trap_init() doesn't use IST, replace set_intr_gate_ist(..., 0) and set_system_intr_gate_ist(..., 0) with their standard counterparts.

set_intr_gate() requires a trace_debug symbol which we don't have and won't use. Use a small macro trick as a workaround.

---

Hi Andy Lutomirski,

This patch is a bit tricky, but I think we don't need to define another helper for such a small problem. What's your opinion?

Thank you!

---
 arch/x86/kernel/traps.c | 23 ++++++++++++++---------
 1 file changed, 14 insertions(+), 9 deletions(-)
arch/x86/kernel/traps.c +2 -6
index 4281988..c24434a 100644
@@ -925,17 +925,22 @@ dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code)
 /* Set of traps needed for early debugging. */
 void __init early_trap_init(void)
 {
+
 	/*
-	 * Don't set ist to DEBUG_STACK as it doesn't work until TSS is
-	 * ready in cpu_init() <-- trap_init(). Before trap_init(), CPU
-	 * runs at ring 0 so it is impossible to hit an invalid stack.
-	 * Using the original stack works well enough at this early
-	 * stage. DEBUG_STACK will be equipped after cpu_init() in
-	 * trap_init().
+	 * Don't use IST to set DEBUG_STACK as it doesn't work until TSS is

arch/x86/kernel/traps.c dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code) · line 939

Quoted 10 lines
arch/x86/kernel/traps.c +8 -1
+	 * ready in cpu_init() <-- trap_init(). Before trap_init(), CPU runs at
+	 * ring 0 so it is impossible to hit an invalid stack.  Using the
+	 * original stack works well enough at this early stage. DEBUG_STACK
+	 * will be equipped after cpu_init() in trap_init().
+	 *
+	 * Since set_intr_gate() needs a trace_debug but we don't have it,
+	 * use the following #define as a workaround.
 	 */
-	set_intr_gate_ist(X86_TRAP_DB, &debug, 0);
+#define trace_debug debug

This goes to arch/x86/include/asm/traps.h like the rest of the trace_* defines.

Quoted 8 lines
arch/x86/kernel/traps.c +3 -1
+	set_intr_gate(X86_TRAP_DB, debug);
+#undef trace_debug
 	/* int3 can be called from all */
-	set_system_intr_gate_ist(X86_TRAP_BP, &int3, 0);
+	set_system_intr_gate(X86_TRAP_BP, &int3);
 #ifdef CONFIG_X86_32
 	set_intr_gate(X86_TRAP_PF, page_fault);
 #endif
Signature · 5 lines
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--
author
#7 lore raw

PATCH v2 1/1 x86, traps: separate set_intr_gate() and cleanup early_trap_init().

In reply to Andy Lutomirski

As early_trap_init() doesn't use IST, replace set_intr_gate_ist() and set_system_intr_gate_ist() with their standard counterparts.

set_intr_gate() requires a trace_debug symbol which we don't have and won't use. This patch seprates set_intr_gate() into 2 parts, and uses base version in early_trap_init().

---

Hi Andy Lutomirski,

This patch should be less tricky than previous one [1]. I also tried to renaming all set_intr_gate(), but it causes too many code changes, so I think you will be satisfied with this one.

Thank you!

[1] https://lkml.org/lkml/2015/2/26/770

---
 arch/x86/include/asm/desc.h |  7 ++++++-
 arch/x86/kernel/traps.c     | 20 ++++++++++++--------
 2 files changed, 18 insertions(+), 9 deletions(-)
arch/x86/include/asm/desc.h +6 -1
index a94b82e..a0bf89f 100644
@@ -376,11 +376,16 @@ static inline void _set_gate(int gate, unsigned type, void *addr,
  * Pentium F0 0F bugfix can have resulted in the mapped
  * IDT being write-protected.
  */
-#define set_intr_gate(n, addr)						\
+#define set_intr_gate_notrace(n, addr)					\
 	do {								\
 		BUG_ON((unsigned)n > 0xFF);				\
 		_set_gate(n, GATE_INTERRUPT, (void *)addr, 0, 0,	\
 			  __KERNEL_CS);					\
+	} while (0)
+
+#define set_intr_gate(n, addr)						\
+	do {								\
+		set_intr_gate_notrace(n, addr);				\
 		_trace_set_gate(n, GATE_INTERRUPT, (void *)trace_##addr,\
 				0, 0, __KERNEL_CS);			\
 	} while (0)
arch/x86/kernel/traps.c +12 -8
index 4281988..9965bd1 100644
@@ -926,16 +926,20 @@ dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code)
 void __init early_trap_init(void)
 {
 	/*
-	 * Don't set ist to DEBUG_STACK as it doesn't work until TSS is
-	 * ready in cpu_init() <-- trap_init(). Before trap_init(), CPU
-	 * runs at ring 0 so it is impossible to hit an invalid stack.
-	 * Using the original stack works well enough at this early
-	 * stage. DEBUG_STACK will be equipped after cpu_init() in
+	 * Don't use IST to set DEBUG_STACK as it doesn't work until TSS
+	 * is ready in cpu_init() <-- trap_init(). Before trap_init(),
+	 * CPU runs at ring 0 so it is impossible to hit an invalid
+	 * stack.  Using the original stack works well enough at this
+	 * early stage. DEBUG_STACK will be equipped after cpu_init() in
 	 * trap_init().
+	 *
+	 * We don't need to set trace_idt_table like set_intr_gate(),
+	 * since we don't have trace_debug and it will be reset to
+	 * 'debug' in trap_init() by set_intr_gate_ist().
 	 */
-	set_intr_gate_ist(X86_TRAP_DB, &debug, 0);
+	set_intr_gate_notrace(X86_TRAP_DB, debug);
 	/* int3 can be called from all */
-	set_system_intr_gate_ist(X86_TRAP_BP, &int3, 0);
+	set_system_intr_gate(X86_TRAP_BP, &int3);
 #ifdef CONFIG_X86_32
 	set_intr_gate(X86_TRAP_PF, page_fault);
 #endif
@@ -1015,7 +1019,7 @@ void __init trap_init(void)
 
 	/*
 	 * X86_TRAP_DB and X86_TRAP_BP have been set
-	 * in early_trap_init(). However, DEBUG_STACK works only after
+	 * in early_trap_init(). However, ITS works only after
 	 * cpu_init() loads TSS. See comments in early_trap_init().
 	 */
 	set_intr_gate_ist(X86_TRAP_DB, &debug, DEBUG_STACK);
Signature · 1 line
1.8.4
author
#8 lore raw

Re: [PATCH v2] x86, traps: separate set_intr_gate() and cleanup early_trap_init().

Hi Andy Lutomirski,

Do you have any comment on this patch?

Thank you.

Quoted 89 lines · Wang Nan

As early_trap_init() doesn't use IST, replace set_intr_gate_ist() and set_system_intr_gate_ist() with their standard counterparts.

set_intr_gate() requires a trace_debug symbol which we don't have and won't use. This patch seprates set_intr_gate() into 2 parts, and uses base version in early_trap_init().

---

Hi Andy Lutomirski,

This patch should be less tricky than previous one [1]. I also tried to renaming all set_intr_gate(), but it causes too many code changes, so I think you will be satisfied with this one.

Thank you!

[1] https://lkml.org/lkml/2015/2/26/770

---
 arch/x86/include/asm/desc.h |  7 ++++++-
 arch/x86/kernel/traps.c     | 20 ++++++++++++--------
 2 files changed, 18 insertions(+), 9 deletions(-)
arch/x86/include/asm/desc.h +6 -1
index a94b82e..a0bf89f 100644
@@ -376,11 +376,16 @@ static inline void _set_gate(int gate, unsigned type, void *addr,
  * Pentium F0 0F bugfix can have resulted in the mapped
  * IDT being write-protected.
  */
-#define set_intr_gate(n, addr)						\
+#define set_intr_gate_notrace(n, addr)					\
 	do {								\
 		BUG_ON((unsigned)n > 0xFF);				\
 		_set_gate(n, GATE_INTERRUPT, (void *)addr, 0, 0,	\
 			  __KERNEL_CS);					\
+	} while (0)
+
+#define set_intr_gate(n, addr)						\
+	do {								\
+		set_intr_gate_notrace(n, addr);				\
 		_trace_set_gate(n, GATE_INTERRUPT, (void *)trace_##addr,\
 				0, 0, __KERNEL_CS);			\
 	} while (0)
arch/x86/kernel/traps.c +12 -8
index 4281988..9965bd1 100644
@@ -926,16 +926,20 @@ dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code)
 void __init early_trap_init(void)
 {
 	/*
-	 * Don't set ist to DEBUG_STACK as it doesn't work until TSS is
-	 * ready in cpu_init() <-- trap_init(). Before trap_init(), CPU
-	 * runs at ring 0 so it is impossible to hit an invalid stack.
-	 * Using the original stack works well enough at this early
-	 * stage. DEBUG_STACK will be equipped after cpu_init() in
+	 * Don't use IST to set DEBUG_STACK as it doesn't work until TSS
+	 * is ready in cpu_init() <-- trap_init(). Before trap_init(),
+	 * CPU runs at ring 0 so it is impossible to hit an invalid
+	 * stack.  Using the original stack works well enough at this
+	 * early stage. DEBUG_STACK will be equipped after cpu_init() in
 	 * trap_init().
+	 *
+	 * We don't need to set trace_idt_table like set_intr_gate(),
+	 * since we don't have trace_debug and it will be reset to
+	 * 'debug' in trap_init() by set_intr_gate_ist().
 	 */
-	set_intr_gate_ist(X86_TRAP_DB, &debug, 0);
+	set_intr_gate_notrace(X86_TRAP_DB, debug);
 	/* int3 can be called from all */
-	set_system_intr_gate_ist(X86_TRAP_BP, &int3, 0);
+	set_system_intr_gate(X86_TRAP_BP, &int3);
 #ifdef CONFIG_X86_32
 	set_intr_gate(X86_TRAP_PF, page_fault);
 #endif
@@ -1015,7 +1019,7 @@ void __init trap_init(void)
 
 	/*
 	 * X86_TRAP_DB and X86_TRAP_BP have been set
-	 * in early_trap_init(). However, DEBUG_STACK works only after
+	 * in early_trap_init(). However, ITS works only after
 	 * cpu_init() loads TSS. See comments in early_trap_init().
 	 */
 	set_intr_gate_ist(X86_TRAP_DB, &debug, DEBUG_STACK);
#9 lore raw

Re: [PATCH v2] x86, traps: separate set_intr_gate() and cleanup early_trap_init().

In reply to Wang Nan
Quoted 8 lines · Wang Nan

As early_trap_init() doesn't use IST, replace set_intr_gate_ist() and set_system_intr_gate_ist() with their standard counterparts.

set_intr_gate() requires a trace_debug symbol which we don't have and won't use. This patch seprates set_intr_gate() into 2 parts, and uses base version in early_trap_init().

Looks good to me. Ingo?

--Andy

Quoted 83 lines

---

Hi Andy Lutomirski,

This patch should be less tricky than previous one [1]. I also tried to renaming all set_intr_gate(), but it causes too many code changes, so I think you will be satisfied with this one.

Thank you!

[1] https://lkml.org/lkml/2015/2/26/770

---
 arch/x86/include/asm/desc.h |  7 ++++++-
 arch/x86/kernel/traps.c     | 20 ++++++++++++--------
 2 files changed, 18 insertions(+), 9 deletions(-)
arch/x86/include/asm/desc.h +6 -1
index a94b82e..a0bf89f 100644
@@ -376,11 +376,16 @@ static inline void _set_gate(int gate, unsigned type, void *addr,
  * Pentium F0 0F bugfix can have resulted in the mapped
  * IDT being write-protected.
  */
-#define set_intr_gate(n, addr)                                         \
+#define set_intr_gate_notrace(n, addr)                                 \
        do {                                                            \
                BUG_ON((unsigned)n > 0xFF);                             \
                _set_gate(n, GATE_INTERRUPT, (void *)addr, 0, 0,        \
                          __KERNEL_CS);                                 \
+       } while (0)
+
+#define set_intr_gate(n, addr)                                         \
+       do {                                                            \
+               set_intr_gate_notrace(n, addr);                         \
                _trace_set_gate(n, GATE_INTERRUPT, (void *)trace_##addr,\
                                0, 0, __KERNEL_CS);                     \
        } while (0)
arch/x86/kernel/traps.c +12 -9
index 4281988..9965bd1 100644
@@ -926,16 +926,20 @@ dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code)
 void __init early_trap_init(void)
 {
        /*
-        * Don't set ist to DEBUG_STACK as it doesn't work until TSS is
-        * ready in cpu_init() <-- trap_init(). Before trap_init(), CPU
-        * runs at ring 0 so it is impossible to hit an invalid stack.
-        * Using the original stack works well enough at this early
-        * stage. DEBUG_STACK will be equipped after cpu_init() in
+        * Don't use IST to set DEBUG_STACK as it doesn't work until TSS
+        * is ready in cpu_init() <-- trap_init(). Before trap_init(),
+        * CPU runs at ring 0 so it is impossible to hit an invalid
+        * stack.  Using the original stack works well enough at this
+        * early stage. DEBUG_STACK will be equipped after cpu_init() in
         * trap_init().
+        *
+        * We don't need to set trace_idt_table like set_intr_gate(),
+        * since we don't have trace_debug and it will be reset to
+        * 'debug' in trap_init() by set_intr_gate_ist().
         */
-       set_intr_gate_ist(X86_TRAP_DB, &debug, 0);
+       set_intr_gate_notrace(X86_TRAP_DB, debug);
        /* int3 can be called from all */
-       set_system_intr_gate_ist(X86_TRAP_BP, &int3, 0);
+       set_system_intr_gate(X86_TRAP_BP, &int3);
 #ifdef CONFIG_X86_32
        set_intr_gate(X86_TRAP_PF, page_fault);
 #endif
@@ -1015,7 +1019,7 @@ void __init trap_init(void)
 
        /*
         * X86_TRAP_DB and X86_TRAP_BP have been set
-        * in early_trap_init(). However, DEBUG_STACK works only after
+        * in early_trap_init(). However, ITS works only after
         * cpu_init() loads TSS. See comments in early_trap_init().
         */
        set_intr_gate_ist(X86_TRAP_DB, &debug, DEBUG_STACK);
--

1.8.4

Signature · 2 lines
Andy Lutomirski
AMA Capital Management, LLC
#10 lore raw

[tip:x86/asm] x86/traps: Separate set_intr_gate() and clean up early_trap_init()

In reply to Wang Nan
x86/traps: Separate set_intr_gate() and clean up early_trap_init()
Commit-ID:  5eca7453d61003bf886992388f8cb407e6f0d051
Gitweb:     http://git.kernel.org/tip/5eca7453d61003bf886992388f8cb407e6f0d051
Author:     Wang Nan <[email protected]>
AuthorDate: Fri, 27 Feb 2015 12:19:49 +0800
Committer:  Ingo Molnar <[email protected]>
CommitDate: Thu, 5 Mar 2015 00:47:29 +0100

x86/traps: Separate set_intr_gate() and clean up early_trap_init()

As early_trap_init() doesn't use IST, replace set_intr_gate_ist() and set_system_intr_gate_ist() with their standard counterparts.

set_intr_gate() requires a trace_debug symbol which we don't have and won't use. This patch separates set_intr_gate() into two parts, and uses base version in early_trap_init().

---
 arch/x86/include/asm/desc.h |  7 ++++++-
 arch/x86/kernel/traps.c     | 20 ++++++++++++--------
 2 files changed, 18 insertions(+), 9 deletions(-)
arch/x86/include/asm/desc.h +6 -1
index a94b82e..a0bf89f 100644
@@ -376,11 +376,16 @@ static inline void _set_gate(int gate, unsigned type, void *addr,
  * Pentium F0 0F bugfix can have resulted in the mapped
  * IDT being write-protected.
  */
-#define set_intr_gate(n, addr)						\
+#define set_intr_gate_notrace(n, addr)					\
 	do {								\
 		BUG_ON((unsigned)n > 0xFF);				\
 		_set_gate(n, GATE_INTERRUPT, (void *)addr, 0, 0,	\
 			  __KERNEL_CS);					\
+	} while (0)
+
+#define set_intr_gate(n, addr)						\
+	do {								\
+		set_intr_gate_notrace(n, addr);				\
 		_trace_set_gate(n, GATE_INTERRUPT, (void *)trace_##addr,\
 				0, 0, __KERNEL_CS);			\
 	} while (0)
arch/x86/kernel/traps.c +12 -8
index 4281988..9965bd1 100644
@@ -926,16 +926,20 @@ dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code)
 void __init early_trap_init(void)
 {
 	/*
-	 * Don't set ist to DEBUG_STACK as it doesn't work until TSS is
-	 * ready in cpu_init() <-- trap_init(). Before trap_init(), CPU
-	 * runs at ring 0 so it is impossible to hit an invalid stack.
-	 * Using the original stack works well enough at this early
-	 * stage. DEBUG_STACK will be equipped after cpu_init() in
+	 * Don't use IST to set DEBUG_STACK as it doesn't work until TSS
+	 * is ready in cpu_init() <-- trap_init(). Before trap_init(),
+	 * CPU runs at ring 0 so it is impossible to hit an invalid
+	 * stack.  Using the original stack works well enough at this
+	 * early stage. DEBUG_STACK will be equipped after cpu_init() in
 	 * trap_init().
+	 *
+	 * We don't need to set trace_idt_table like set_intr_gate(),
+	 * since we don't have trace_debug and it will be reset to
+	 * 'debug' in trap_init() by set_intr_gate_ist().
 	 */
-	set_intr_gate_ist(X86_TRAP_DB, &debug, 0);
+	set_intr_gate_notrace(X86_TRAP_DB, debug);
 	/* int3 can be called from all */
-	set_system_intr_gate_ist(X86_TRAP_BP, &int3, 0);
+	set_system_intr_gate(X86_TRAP_BP, &int3);
 #ifdef CONFIG_X86_32
 	set_intr_gate(X86_TRAP_PF, page_fault);
 #endif
@@ -1015,7 +1019,7 @@ void __init trap_init(void)
 
 	/*
 	 * X86_TRAP_DB and X86_TRAP_BP have been set
-	 * in early_trap_init(). However, DEBUG_STACK works only after
+	 * in early_trap_init(). However, ITS works only after
 	 * cpu_init() loads TSS. See comments in early_trap_init().
 	 */
 	set_intr_gate_ist(X86_TRAP_DB, &debug, DEBUG_STACK);