Skip to contentSkip to conversation

x86_64 debug_stack nested patch

PATCH In review bibo,mao 4 participants 6 messages May 10, 2006

lore
Share & tools
lore

bibo,mao

author
#0 lore raw

PATCH 1/1 x86_64 debug_stack nested patch

hi,
In x86_64 platform, INT1 and INT3 trap stack is IST stack called DEBUG_STACK, when INT1/INT3 trap happens, system will switch to DEBUG_STACK by hardware. Current DEBUG_STACK size is 4K, when int1/int3 trap happens, kernel will minus current DEBUG_STACK IST value by 4k. But if int3/int1 trap is nested, it will destroy other vector's IST stack. This patch modifies this, it sets DEBUG_STACK size as 8K and allows two level of nested int1/int3 trap.
Kprobe DEBUG_STACK may be nested, because kprobe hanlder may be probed by other kprobes. This patch is against 2.6.17-rc3.

Thanks
bibo,mao

arch/x86_64/kernel/traps.c +7 -33
diff -Nruap 2.6.17-rc3.org/arch/x86_64/kernel/traps.c 2.6.17-rc3/arch/x86_64/kernel/traps.c
@@ -141,50 +141,24 @@ static unsigned long *in_exception_stack
 		[DOUBLEFAULT_STACK - 1] = "#DF",
 		[STACKFAULT_STACK - 1] = "#SS",
 		[MCE_STACK - 1] = "#MC",
-#if DEBUG_STKSZ > EXCEPTION_STKSZ
-		[N_EXCEPTION_STACKS ... N_EXCEPTION_STACKS + DEBUG_STKSZ / EXCEPTION_STKSZ - 2] = "#DB[?]"
-#endif
 	};
-	unsigned k;
+	unsigned stack_size, end, k;
 
 	for (k = 0; k < N_EXCEPTION_STACKS; k++) {
-		unsigned long end;
-
-		switch (k + 1) {
-#if DEBUG_STKSZ > EXCEPTION_STKSZ
-		case DEBUG_STACK:
-			end = cpu_pda(cpu)->debugstack + DEBUG_STKSZ;
-			break;
-#endif
-		default:
-			end = per_cpu(init_tss, cpu).ist[k];
-			break;
-		}
+		end = per_cpu(init_tss, cpu).ist[k];
 		if (stack >= end)
 			continue;
-		if (stack >= end - EXCEPTION_STKSZ) {
+		if (k == (DEBUG_STACK - 1))
+			stack_size = DEBUG_STKSZ;
+		else stack_size = EXCEPTION_STKSZ;
+
+		if (stack >= end - stack_size) {
 			if (*usedp & (1U << k))
 				break;
 			*usedp |= 1U << k;
 			*idp = ids[k];
 			return (unsigned long *)end;
 		}
-#if DEBUG_STKSZ > EXCEPTION_STKSZ
-		if (k == DEBUG_STACK - 1 && stack >= end - DEBUG_STKSZ) {
-			unsigned j = N_EXCEPTION_STACKS - 1;
-
-			do {
-				++j;
-				end -= EXCEPTION_STKSZ;
-				ids[j][4] = '1' + (j - N_EXCEPTION_STACKS);
-			} while (stack < end - EXCEPTION_STKSZ);
-			if (*usedp & (1U << j))
-				break;
-			*usedp |= 1U << j;
-			*idp = ids[j];
-			return (unsigned long *)end;
-		}
-#endif
 	}
 	return NULL;
 }
include/asm-x86_64/page.h +1 -1
diff -Nruap 2.6.17-rc3.org/include/asm-x86_64/page.h 2.6.17-rc3/include/asm-x86_64/page.h
@@ -20,7 +20,7 @@
 #define EXCEPTION_STACK_ORDER 0
 #define EXCEPTION_STKSZ (PAGE_SIZE << EXCEPTION_STACK_ORDER)
 
-#define DEBUG_STACK_ORDER EXCEPTION_STACK_ORDER
+#define DEBUG_STACK_ORDER 1
 #define DEBUG_STKSZ (PAGE_SIZE << DEBUG_STACK_ORDER)
 
 #define IRQSTACK_ORDER 2
#1 lore raw

Would you mind explaining why you
- needed to replace the handling of the DEBUG_STKSZ > EXCEPTION_STKSZ in in_exception_stack()?
- used a hard coded 1 instead of (EXCEPTION_STACK_ORDER + 1) for defining DEBUG_STACK_ORDER?

Thanks, Jan

Quoted 1 line
Quoted 1 line
Quoted 1 line

"bibo,mao" <[email protected]> 10.05.06 08:45 >>>

hi,
In x86_64 platform, INT1 and INT3 trap stack is IST stack called DEBUG_STACK, when INT1/INT3 trap happens, system will switch to DEBUG_STACK by hardware. Current DEBUG_STACK size is 4K, when int1/int3 trap happens, kernel will minus current DEBUG_STACK IST value by 4k. But if int3/int1 trap is nested, it will destroy other vector's IST stack. This patch modifies this, it sets DEBUG_STACK size as 8K and allows two level of nested int1/int3 trap.
Kprobe DEBUG_STACK may be nested, because kprobe hanlder may be probed by other kprobes. This patch is against 2.6.17-rc3.

Thanks
bibo,mao

#2 lore raw

ok, EXCEPTION_STACK_ORDER + 1 is better for me.
in_exception_stack() function is to judge which IST stack by parameter stack value, if DEBUG_STKSZ value is set as 8K. The original function can not judge whether it is within DEBUG_STACK space.

Thanks
bibo,mao

Quoted 22 lines · Jan Beulich

Would you mind explaining why you
- needed to replace the handling of the DEBUG_STKSZ > EXCEPTION_STKSZ in in_exception_stack()?
- used a hard coded 1 instead of (EXCEPTION_STACK_ORDER + 1) for defining DEBUG_STACK_ORDER?

Thanks, Jan

Quoted 1 line
Quoted 1 line
Quoted 1 line

"bibo,mao" <[email protected]> 10.05.06 08:45 >>>

hi,
In x86_64 platform, INT1 and INT3 trap stack is IST stack called DEBUG_STACK, when INT1/INT3 trap happens, system will switch to DEBUG_STACK by hardware. Current DEBUG_STACK size is 4K, when int1/int3 trap happens, kernel will minus current DEBUG_STACK IST value by 4k. But if int3/int1 trap is nested, it will destroy other vector's IST stack. This patch modifies this, it sets DEBUG_STACK size as 8K and allows two level of nested int1/int3 trap.
Kprobe DEBUG_STACK may be nested, because kprobe hanlder may be probed by other kprobes. This patch is against 2.6.17-rc3.

Thanks
bibo,mao
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#3 lore raw
Quoted 3 lines

in_exception_stack() function is to judge which IST stack by parameter stack value, if DEBUG_STKSZ value is set as 8K. The original function can not judge whether it is within DEBUG_STACK space.

I rather think that the new code can't work properly. Since the pointer in the TSS gets decreased while the handler is running, using that value is not going to tell you the end of the stack, but you'd rather get the end of the stack the next (nested) invocation of the handler would use. Further, treating the entire DEBUG_STKSZ range as a single piece is wrong, too, because it is not being used as a contiguous stack (but rather as 2 stacks EXCEPTION_STKSZ in size); the new code shouldn't be able to properly deal with nested invocations because of this.

Jan

#4 lore raw

yes, I am wrong. And I will modify this. And then only need define DEBUG_STACK_ORDER as (EXCEPTION_STACK_ORDER + 1)

thanks
bibo,mao

Quoted 15 lines · Jan Beulich
Quoted 1 line

in_exception_stack() function is to judge which IST stack by parameter stack value, if DEBUG_STKSZ value is set as 8K. The original function can not judge whether it is within DEBUG_STACK space.

I rather think that the new code can't work properly. Since the pointer in the TSS gets decreased while the handler is running, using that value is not going to tell you the end of the stack, but you'd rather get the end of the stack the next (nested) invocation of the handler would use. Further, treating the entire DEBUG_STKSZ range as a single piece is wrong, too, because it is not being used as a contiguous stack (but rather as 2 stacks EXCEPTION_STKSZ in size); the new code shouldn't be able to properly deal with nested invocations because of this.

Jan
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#5 lore raw
Quoted 2 lines · bibo,mao

yes, I am wrong. And I will modify this. And then only need define DEBUG_STACK_ORDER as (EXCEPTION_STACK_ORDER + 1)

If the kprobes code really wants to nest (in my opinion it's a kprobes bug) you should reduce its stack to 2K or so. Don't want to waste space for stupid code like this.

-Andi