Skip to contentSkip to conversation

Patch "x86/entry/64: Add missing irqflags tracing to native_load_gs_index()" has been added to the 4.9-stable tree

PATCH [email protected] 3 participants 8 messages Nov 27, 2017 – Dec 4, 2017

Subsystems (3)
lore
Share & tools
lore

[email protected]

author maintainer Applied
#0 lore raw

This is a note to let you know that I've just added the patch titled

x86/entry/64: Add missing irqflags tracing to native_load_gs_index()

to the 4.9-stable tree which can be found at:
http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary

The filename of the patch is:
x86-entry-64-add-missing-irqflags-tracing-to-native_load_gs_index.patch
and it can be found in the queue-4.9 subdirectory.

If you, or anyone else, feels it should not be added to the stable tree, please let <[email protected]> know about it.

Quoted 1 line

From ca37e57bbe0cf1455ea3e84eb89ed04a132d59e1 Mon Sep 17 00:00:00 2001

From: Andy Lutomirski <[email protected]>
Date: Wed, 22 Nov 2017 20:39:16 -0800
Subject: x86/entry/64: Add missing irqflags tracing to native_load_gs_index()

From: Andy Lutomirski <[email protected]>

commit ca37e57bbe0cf1455ea3e84eb89ed04a132d59e1 upstream.

Running this code with IRQs enabled (where dummy_lock is a spinlock):

static void check_load_gs_index(void)

{
	/* This will fail. */
	load_gs_index(0xffff);

	spin_lock(&dummy_lock);
	spin_unlock(&dummy_lock);
}

Will generate a lockdep warning. The issue is that the actual write to %gs would cause an exception with IRQs disabled, and the exception handler would, as an inadvertent side effect, update irqflag tracing to reflect the IRQs-off status. native_load_gs_index() would then turn IRQs back on and return with irqflag tracing still thinking that IRQs were off. The dummy lock-and-unlock causes lockdep to notice the error and warn.

Fix it by adding the missing tracing.

Apparently nothing did this in a context where it mattered. I haven't tried to find a code path that would actually exhibit the warning if appropriately nasty user code were running.

I suspect that the security impact of this bug is very, very low -- production systems don't run with lockdep enabled, and the warning is mostly harmless anyway.

Found during a quick audit of the entry code to try to track down an unrelated bug that Ingo found in some still-in-development code.

---
 arch/x86/entry/entry_64.S |   10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
arch/x86/entry/entry_64.S +8 -2
@@ -54,15 +54,19 @@ ENTRY(native_usergs_sysret64)
 ENDPROC(native_usergs_sysret64)
 #endif /* CONFIG_PARAVIRT */
 
-.macro TRACE_IRQS_IRETQ
+.macro TRACE_IRQS_FLAGS flags:req
 #ifdef CONFIG_TRACE_IRQFLAGS
-	bt	$9, EFLAGS(%rsp)		/* interrupts off? */
+	bt	$9, \flags		/* interrupts off? */
 	jnc	1f
 	TRACE_IRQS_ON
 1:
 #endif
 .endm
 
+.macro TRACE_IRQS_IRETQ
+	TRACE_IRQS_FLAGS EFLAGS(%rsp)
+.endm
+
 /*
  * When dynamic function tracer is enabled it will add a breakpoint
  * to all locations that it is about to modify, sync CPUs, update
@@ -868,11 +872,13 @@ idtentry simd_coprocessor_error		do_simd
 ENTRY(native_load_gs_index)
 	pushfq
 	DISABLE_INTERRUPTS(CLBR_ANY & ~CLBR_RDI)
+	TRACE_IRQS_OFF
 	SWAPGS
 .Lgs_change:
 	movl	%edi, %gs
 2:	ALTERNATIVE "", "mfence", X86_BUG_SWAPGS_FENCE
 	SWAPGS
+	TRACE_IRQS_FLAGS (%rsp)
 	popfq
 	ret
 END(native_load_gs_index)

Patches currently in stable-queue which might be from [email protected] are

queue-4.9/x86-entry-64-add-missing-irqflags-tracing-to-native_load_gs_index.patch

maintainer
#1 lore raw

If this hasn't been released yet, please hold off. It's buggy.

Quoted 116 lines · On Mon, Nov 27, 2017 at 4:41 AM, <[email protected]> wrote:

This is a note to let you know that I've just added the patch titled

x86/entry/64: Add missing irqflags tracing to native_load_gs_index()

to the 4.9-stable tree which can be found at:
http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary

The filename of the patch is:
x86-entry-64-add-missing-irqflags-tracing-to-native_load_gs_index.patch
and it can be found in the queue-4.9 subdirectory.

If you, or anyone else, feels it should not be added to the stable tree, please let <[email protected]> know about it.

From ca37e57bbe0cf1455ea3e84eb89ed04a132d59e1 Mon Sep 17 00:00:00 2001
From: Andy Lutomirski <[email protected]>
Date: Wed, 22 Nov 2017 20:39:16 -0800
Subject: x86/entry/64: Add missing irqflags tracing to native_load_gs_index()

From: Andy Lutomirski <[email protected]>

commit ca37e57bbe0cf1455ea3e84eb89ed04a132d59e1 upstream.

Running this code with IRQs enabled (where dummy_lock is a spinlock):

static void check_load_gs_index(void)

{
        /* This will fail. */
        load_gs_index(0xffff);

        spin_lock(&dummy_lock);
        spin_unlock(&dummy_lock);
}

Will generate a lockdep warning. The issue is that the actual write to %gs would cause an exception with IRQs disabled, and the exception handler would, as an inadvertent side effect, update irqflag tracing to reflect the IRQs-off status. native_load_gs_index() would then turn IRQs back on and return with irqflag tracing still thinking that IRQs were off. The dummy lock-and-unlock causes lockdep to notice the error and warn.

Fix it by adding the missing tracing.

Apparently nothing did this in a context where it mattered. I haven't tried to find a code path that would actually exhibit the warning if appropriately nasty user code were running.

I suspect that the security impact of this bug is very, very low -- production systems don't run with lockdep enabled, and the warning is mostly harmless anyway.

Found during a quick audit of the entry code to try to track down an unrelated bug that Ingo found in some still-in-development code.

---
 arch/x86/entry/entry_64.S |   10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
arch/x86/entry/entry_64.S +8 -2
@@ -54,15 +54,19 @@ ENTRY(native_usergs_sysret64)
 ENDPROC(native_usergs_sysret64)
 #endif /* CONFIG_PARAVIRT */
 
-.macro TRACE_IRQS_IRETQ
+.macro TRACE_IRQS_FLAGS flags:req
 #ifdef CONFIG_TRACE_IRQFLAGS
-       bt      $9, EFLAGS(%rsp)                /* interrupts off? */
+       bt      $9, \flags              /* interrupts off? */
        jnc     1f
        TRACE_IRQS_ON
 1:
 #endif
 .endm
 
+.macro TRACE_IRQS_IRETQ
+       TRACE_IRQS_FLAGS EFLAGS(%rsp)
+.endm
+
 /*
  * When dynamic function tracer is enabled it will add a breakpoint
  * to all locations that it is about to modify, sync CPUs, update
@@ -868,11 +872,13 @@ idtentry simd_coprocessor_error           do_simd
 ENTRY(native_load_gs_index)
        pushfq
        DISABLE_INTERRUPTS(CLBR_ANY & ~CLBR_RDI)
+       TRACE_IRQS_OFF
        SWAPGS
 .Lgs_change:
        movl    %edi, %gs
 2:     ALTERNATIVE "", "mfence", X86_BUG_SWAPGS_FENCE
        SWAPGS
+       TRACE_IRQS_FLAGS (%rsp)
        popfq
        ret
 END(native_load_gs_index)

Patches currently in stable-queue which might be from [email protected] are

queue-4.9/x86-entry-64-add-missing-irqflags-tracing-to-native_load_gs_index.patch

author maintainer
#2 lore raw
Quoted 1 line · Andy Lutomirski

If this hasn't been released yet, please hold off. It's buggy.

Ah crap, I missed this and just released it. Is there a fix for this in Linus's tree? How "buggy" is it?

thanks,

greg k-h

#3 lore raw
Quoted 9 lines · Greg KH
Quoted 3 lines · Andy Lutomirski

If this hasn't been released yet, please hold off. It's buggy.

Ah crap, I missed this and just released it. Is there a fix for this in Linus's tree? How "buggy" is it?

It breaks suspend/resume on some configuration that I haven't quite figured out. I think it only happens if (lockdep is on and some other unknown condition occurs) or (ftrace is actively tracing irq flags). I have a fix, but I was waiting to for the OP to test it.

I'll send the fix Linusward later today without the Tested-by if I don't hear anything.

Quoted 4 lines

thanks,

greg k-h

maintainer
#4 lore raw
Quoted 7 lines · Andy Lutomirski
Quoted 7 lines · Greg KH
Quoted 3 lines · Andy Lutomirski

If this hasn't been released yet, please hold off. It's buggy.

Quoted 6 lines
Quoted 2 lines

Ah crap, I missed this and just released it. Is there a fix for this in Linus's tree? How "buggy" is it?

It breaks suspend/resume on some configuration that I haven't quite figured out. I think it only happens if (lockdep is on and some other unknown condition occurs) or (ftrace is actively tracing irq flags). I have a fix, but I was waiting to for the OP to test it.

I'll send the fix Linusward later today without the Tested-by if I don't hear anything.

I think the thing to do is to revert the patch from -stable. The bug it fixes is very minor, and the regression is that it made a pre-existing bug in some nearly-undebuggable core resume code much easier to hit. I don't feel comfortable with a backport of the latter fix until it has a good long soak in Linus' tree.

author maintainer
#5 lore raw
Quoted 13 lines · Andy Lutomirski
Quoted 12 lines · Andy Lutomirski
Quoted 9 lines · Greg KH
Quoted 3 lines · Andy Lutomirski

If this hasn't been released yet, please hold off. It's buggy.

Ah crap, I missed this and just released it. Is there a fix for this in Linus's tree? How "buggy" is it?

It breaks suspend/resume on some configuration that I haven't quite figured out. I think it only happens if (lockdep is on and some other unknown condition occurs) or (ftrace is actively tracing irq flags). I have a fix, but I was waiting to for the OP to test it.

Quoted 7 lines
Quoted 1 line

I'll send the fix Linusward later today without the Tested-by if I don't hear anything.

I think the thing to do is to revert the patch from -stable. The bug it fixes is very minor, and the regression is that it made a pre-existing bug in some nearly-undebuggable core resume code much easier to hit. I don't feel comfortable with a backport of the latter fix until it has a good long soak in Linus' tree.

Ok, now reverted. Should I do the same thing for the patch in 4.14-stable as well?

thanks,

greg k-h

#6 lore raw
Quoted 20 lines · Greg KH
Quoted 17 lines · Andy Lutomirski
Quoted 14 lines · Andy Lutomirski
Quoted 9 lines
Quoted 6 lines · Greg KH

On Wed, Nov 29, 2017 at 08:23:10AM -0800, Andy Lutomirski wrote:
If this hasn't been released yet, please hold off. It's buggy.

Ah crap, I missed this and just released it. Is there a fix for this in Linus's tree? How "buggy" is it?

It breaks suspend/resume on some configuration that I haven't quite figured out. I think it only happens if (lockdep is on and some other unknown condition occurs) or (ftrace is actively tracing irq flags). I have a fix, but I was waiting to for the OP to test it.

I'll send the fix Linusward later today without the Tested-by if I don't hear anything.

Quoted 8 lines
Quoted 5 lines

I think the thing to do is to revert the patch from -stable. The bug it fixes is very minor, and the regression is that it made a pre-existing bug in some nearly-undebuggable core resume code much easier to hit. I don't feel comfortable with a backport of the latter fix until it has a good long soak in Linus' tree.

Ok, now reverted. Should I do the same thing for the patch in 4.14-stable as well?

Yes please.

Quoted 4 lines

thanks,

greg k-h

author maintainer
#7 lore raw
Quoted 26 lines · Andy Lutomirski
Quoted 26 lines · Greg KH
Quoted 22 lines · Andy Lutomirski
Quoted 8 lines · Andy Lutomirski
Quoted 7 lines
Quoted 6 lines · Greg KH

On Wed, Nov 29, 2017 at 08:23:10AM -0800, Andy Lutomirski wrote:
If this hasn't been released yet, please hold off. It's buggy.

Quoted 6 lines
Quoted 2 lines

Ah crap, I missed this and just released it. Is there a fix for this in Linus's tree? How "buggy" is it?

It breaks suspend/resume on some configuration that I haven't quite figured out. I think it only happens if (lockdep is on and some other unknown condition occurs) or (ftrace is actively tracing irq flags). I have a fix, but I was waiting to for the OP to test it.

I'll send the fix Linusward later today without the Tested-by if I don't hear anything.

I think the thing to do is to revert the patch from -stable. The bug it fixes is very minor, and the regression is that it made a pre-existing bug in some nearly-undebuggable core resume code much easier to hit. I don't feel comfortable with a backport of the latter fix until it has a good long soak in Linus' tree.

Quoted 4 lines
Quoted 2 lines

Ok, now reverted. Should I do the same thing for the patch in 4.14-stable as well?

Yes please.

Great, now done.

greg k-h