Skip to contentSkip to conversation

arm64: entry: Place an SB sequence following an ERET instruction

PATCH Applied Florian Fainelli 4 participants 12 messages Jul 9, 2020 – Aug 24, 2020

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 arm64: entry: Place an SB sequence following an ERET instruction

From: Will Deacon <[email protected]>

commit 679db70801da9fda91d26caf13bf5b5ccc74e8e8 upstream

Some CPUs can speculate past an ERET instruction and potentially perform speculative accesses to memory before processing the exception return. Since the register state is often controlled by a lower privilege level at the point of an ERET, this could potentially be used as part of a side-channel attack.

This patch emits an SB sequence after each ERET so that speculation is held up on exception return.

[florian: Adjust hyp-entry.S to account for the label
added change to hyp/entry.S]

---
Changes in v2:

- added missing hunk in hyp/entry.S per Will's feedback

 arch/arm64/kernel/entry.S      | 2 ++
 arch/arm64/kvm/hyp/entry.S     | 2 ++
 arch/arm64/kvm/hyp/hyp-entry.S | 4 ++++
 3 files changed, 8 insertions(+)
arch/arm64/kernel/entry.S +2 -0
index ca978d7d98eb..3408c782702c 100644
@@ -255,6 +255,7 @@ alternative_insn eret, nop, ARM64_UNMAP_KERNEL_AT_EL0
 	.else
 	eret
 	.endif
+	sb
 	.endm
 
 	.macro	get_thread_info, rd
@@ -945,6 +946,7 @@ __ni_sys_trace:
 	mrs	x30, far_el1
 	.endif
 	eret
+	sb
 	.endm
 
 	.align	11
arch/arm64/kvm/hyp/entry.S +2 -0
index a360ac6e89e9..93704e6894d2 100644
@@ -83,6 +83,7 @@ ENTRY(__guest_enter)
 
 	// Do not touch any register after this!
 	eret
+	sb
 ENDPROC(__guest_enter)
 
 ENTRY(__guest_exit)
@@ -195,4 +196,5 @@ alternative_endif
 	ldp	x0, x1, [sp], #16
 
 	eret
+	sb
 ENDPROC(__fpsimd_guest_restore)
arch/arm64/kvm/hyp/hyp-entry.S +4 -0
index bf4988f9dae8..3675e7f0ab72 100644
@@ -97,6 +97,7 @@ el1_sync:				// Guest trapped into EL2
 	do_el2_call
 
 2:	eret
+	sb
 
 el1_hvc_guest:
 	/*
@@ -147,6 +148,7 @@ wa_epilogue:
 	mov	x0, xzr
 	add	sp, sp, #16
 	eret
+	sb
 
 el1_trap:
 	get_vcpu_ptr	x1, x0
@@ -198,6 +200,7 @@ el2_error:
 	b.ne	__hyp_panic
 	mov	x0, #(1 << ARM_EXIT_WITH_SERROR_BIT)
 	eret
+	sb
 
 ENTRY(__hyp_do_panic)
 	mov	lr, #(PSR_F_BIT | PSR_I_BIT | PSR_A_BIT | PSR_D_BIT |\
@@ -206,6 +209,7 @@ ENTRY(__hyp_do_panic)
 	ldr	lr, =panic
 	msr	elr_el2, lr
 	eret
+	sb
 ENDPROC(__hyp_do_panic)
 
 ENTRY(__hyp_panic)
Signature · 6 lines
2.17.1

_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
maintainer Applied
#1 lore raw
Quoted 10 lines · Florian Fainelli

From: Will Deacon <[email protected]>

commit 679db70801da9fda91d26caf13bf5b5ccc74e8e8 upstream

Some CPUs can speculate past an ERET instruction and potentially perform speculative accesses to memory before processing the exception return. Since the register state is often controlled by a lower privilege level at the point of an ERET, this could potentially be used as part of a side-channel attack.

Quoted 7 lines

This patch emits an SB sequence after each ERET so that speculation is held up on exception return.

[florian: Adjust hyp-entry.S to account for the label added change to hyp/entry.S]

I've queued it up, thanks!

Signature · 2 lines
Thanks,
Sasha
maintainer
#2 lore raw
In reply to Florian Fainelli
Quoted 13 lines · Florian Fainelli

From: Will Deacon <[email protected]>

commit 679db70801da9fda91d26caf13bf5b5ccc74e8e8 upstream

Some CPUs can speculate past an ERET instruction and potentially perform speculative accesses to memory before processing the exception return. Since the register state is often controlled by a lower privilege level at the point of an ERET, this could potentially be used as part of a side-channel attack.

This patch emits an SB sequence after each ERET so that speculation is held up on exception return.

Quoted 8 lines

[florian: Adjust hyp-entry.S to account for the label
added change to hyp/entry.S]

---
Changes in v2:

- added missing hunk in hyp/entry.S per Will's feedback

What about 4.19.y and 4.14.y trees? I can't take something for 4.9.y and then have a regression if someone moves to a newer release, right?

thanks,

greg k-h

#3 lore raw
Quoted 21 lines · Greg KH
Quoted 20 lines · Florian Fainelli

From: Will Deacon <[email protected]>

commit 679db70801da9fda91d26caf13bf5b5ccc74e8e8 upstream

Some CPUs can speculate past an ERET instruction and potentially perform speculative accesses to memory before processing the exception return. Since the register state is often controlled by a lower privilege level at the point of an ERET, this could potentially be used as part of a side-channel attack.

This patch emits an SB sequence after each ERET so that speculation is held up on exception return.

[florian: Adjust hyp-entry.S to account for the label
added change to hyp/entry.S]

---
Changes in v2:

Quoted 4 lines
Quoted 1 line

- added missing hunk in hyp/entry.S per Will's feedback

What about 4.19.y and 4.14.y trees? I can't take something for 4.9.y and then have a regression if someone moves to a newer release, right?

Sure, send you candidates for 4.14 and 4.19.

Signature · 1 line
Florian
#4 lore raw
Quoted 22 lines · Florian Fainelli
Quoted 21 lines · Greg KH
Quoted 20 lines · Florian Fainelli

From: Will Deacon <[email protected]>

commit 679db70801da9fda91d26caf13bf5b5ccc74e8e8 upstream

Some CPUs can speculate past an ERET instruction and potentially perform speculative accesses to memory before processing the exception return. Since the register state is often controlled by a lower privilege level at the point of an ERET, this could potentially be used as part of a side-channel attack.

This patch emits an SB sequence after each ERET so that speculation is held up on exception return.

[florian: Adjust hyp-entry.S to account for the label
added change to hyp/entry.S]

---
Changes in v2:

Quoted 6 lines
Quoted 4 lines
Quoted 1 line

- added missing hunk in hyp/entry.S per Will's feedback

What about 4.19.y and 4.14.y trees? I can't take something for 4.9.y and then have a regression if someone moves to a newer release, right?

Sure, send you candidates for 4.14 and 4.19.

Greg, did you have a chance to queue those changes for 4.9, 4.14 and 4.19?

https://lore.kernel.org/linux-arm-kernel/[email protected]/
https://lore.kernel.org/linux-arm-kernel/[email protected]/
https://lore.kernel.org/linux-arm-kernel/[email protected]/

Thanks

Signature · 5 lines
Florian
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
maintainer
#5 lore raw
Quoted 30 lines · Florian Fainelli
Quoted 27 lines · Florian Fainelli
Quoted 25 lines · Greg KH
Quoted 13 lines · Florian Fainelli

From: Will Deacon <[email protected]>

commit 679db70801da9fda91d26caf13bf5b5ccc74e8e8 upstream

Some CPUs can speculate past an ERET instruction and potentially perform speculative accesses to memory before processing the exception return. Since the register state is often controlled by a lower privilege level at the point of an ERET, this could potentially be used as part of a side-channel attack.

This patch emits an SB sequence after each ERET so that speculation is held up on exception return.

Quoted 8 lines

[florian: Adjust hyp-entry.S to account for the label
added change to hyp/entry.S]

---
Changes in v2:

- added missing hunk in hyp/entry.S per Will's feedback

What about 4.19.y and 4.14.y trees? I can't take something for 4.9.y and then have a regression if someone moves to a newer release, right?

Quoted 7 lines
Quoted 1 line

Sure, send you candidates for 4.14 and 4.19.

Greg, did you have a chance to queue those changes for 4.9, 4.14 and 4.19?

https://lore.kernel.org/linux-arm-kernel/[email protected]/
https://lore.kernel.org/linux-arm-kernel/[email protected]/
https://lore.kernel.org/linux-arm-kernel/[email protected]/

Nope, I was waiting for Will's "ack" for these.

thanks,

greg k-h

#6 lore raw
Quoted 33 lines · Greg KH
Quoted 32 lines · Florian Fainelli
Quoted 28 lines · Florian Fainelli
Quoted 21 lines · Greg KH
Quoted 20 lines · Florian Fainelli

From: Will Deacon <[email protected]>

commit 679db70801da9fda91d26caf13bf5b5ccc74e8e8 upstream

Some CPUs can speculate past an ERET instruction and potentially perform speculative accesses to memory before processing the exception return. Since the register state is often controlled by a lower privilege level at the point of an ERET, this could potentially be used as part of a side-channel attack.

This patch emits an SB sequence after each ERET so that speculation is held up on exception return.

[florian: Adjust hyp-entry.S to account for the label
added change to hyp/entry.S]

---
Changes in v2:

Quoted 4 lines
Quoted 1 line

- added missing hunk in hyp/entry.S per Will's feedback

What about 4.19.y and 4.14.y trees? I can't take something for 4.9.y and then have a regression if someone moves to a newer release, right?

Sure, send you candidates for 4.14 and 4.19.

Quoted 7 lines
Quoted 5 lines

Greg, did you have a chance to queue those changes for 4.9, 4.14 and 4.19?

https://lore.kernel.org/linux-arm-kernel/[email protected]/
https://lore.kernel.org/linux-arm-kernel/[email protected]/
https://lore.kernel.org/linux-arm-kernel/[email protected]/

Nope, I was waiting for Will's "ack" for these.

OK, Will, can you review those? Thanks

Signature · 1 line
Florian
#7 lore raw
Quoted 38 lines · Florian Fainelli
Quoted 35 lines · Greg KH
Quoted 34 lines · Florian Fainelli
Quoted 22 lines · Florian Fainelli
Quoted 21 lines · Greg KH
Quoted 20 lines · Florian Fainelli

From: Will Deacon <[email protected]>

commit 679db70801da9fda91d26caf13bf5b5ccc74e8e8 upstream

Some CPUs can speculate past an ERET instruction and potentially perform speculative accesses to memory before processing the exception return. Since the register state is often controlled by a lower privilege level at the point of an ERET, this could potentially be used as part of a side-channel attack.

This patch emits an SB sequence after each ERET so that speculation is held up on exception return.

[florian: Adjust hyp-entry.S to account for the label
added change to hyp/entry.S]

---
Changes in v2:

Quoted 6 lines
Quoted 4 lines
Quoted 1 line

- added missing hunk in hyp/entry.S per Will's feedback

What about 4.19.y and 4.14.y trees? I can't take something for 4.9.y and then have a regression if someone moves to a newer release, right?

Sure, send you candidates for 4.14 and 4.19.

Greg, did you have a chance to queue those changes for 4.9, 4.14 and 4.19?

Quoted 7 lines
Quoted 5 lines
Quoted 3 lines

https://lore.kernel.org/linux-arm-kernel/[email protected]/
https://lore.kernel.org/linux-arm-kernel/[email protected]/
https://lore.kernel.org/linux-arm-kernel/[email protected]/

Nope, I was waiting for Will's "ack" for these.

OK, Will, can you review those? Thanks

Will, can you please review those patches?

Signature · 1 line
Florian
#8 lore raw
In reply to Greg KH
Quoted 33 lines · Greg KH
Quoted 32 lines · Florian Fainelli
Quoted 28 lines · Florian Fainelli
Quoted 21 lines · Greg KH
Quoted 20 lines · Florian Fainelli

From: Will Deacon <[email protected]>

commit 679db70801da9fda91d26caf13bf5b5ccc74e8e8 upstream

Some CPUs can speculate past an ERET instruction and potentially perform speculative accesses to memory before processing the exception return. Since the register state is often controlled by a lower privilege level at the point of an ERET, this could potentially be used as part of a side-channel attack.

This patch emits an SB sequence after each ERET so that speculation is held up on exception return.

[florian: Adjust hyp-entry.S to account for the label
added change to hyp/entry.S]

---
Changes in v2:

Quoted 4 lines
Quoted 1 line

- added missing hunk in hyp/entry.S per Will's feedback

What about 4.19.y and 4.14.y trees? I can't take something for 4.9.y and then have a regression if someone moves to a newer release, right?

Sure, send you candidates for 4.14 and 4.19.

Quoted 7 lines
Quoted 5 lines

Greg, did you have a chance to queue those changes for 4.9, 4.14 and 4.19?

https://lore.kernel.org/linux-arm-kernel/[email protected]/
https://lore.kernel.org/linux-arm-kernel/[email protected]/
https://lore.kernel.org/linux-arm-kernel/[email protected]/

Nope, I was waiting for Will's "ack" for these.

This patch doesn't even build for me (the 'sb' macro is not defined in 4.9), and I really wonder why we bother backporting it at all. Nobody's ever shown it to be a problem in practice, and it's clear that this is just being submitted to tick a box rather than anything else (otherwise it would build, right?).

So I'm not going to Ack any of them. As with a lot of this side-channel stuff the cure is far worse than the disease.

Will
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm

#9 lore raw
Quoted 40 lines · Will Deacon
Quoted 39 lines · Greg KH
Quoted 37 lines · Florian Fainelli
Quoted 22 lines · Florian Fainelli
Quoted 21 lines · Greg KH
Quoted 20 lines · Florian Fainelli

From: Will Deacon <[email protected]>

commit 679db70801da9fda91d26caf13bf5b5ccc74e8e8 upstream

Some CPUs can speculate past an ERET instruction and potentially perform speculative accesses to memory before processing the exception return. Since the register state is often controlled by a lower privilege level at the point of an ERET, this could potentially be used as part of a side-channel attack.

This patch emits an SB sequence after each ERET so that speculation is held up on exception return.

[florian: Adjust hyp-entry.S to account for the label
added change to hyp/entry.S]

---
Changes in v2:

Quoted 6 lines
Quoted 4 lines
Quoted 1 line

- added missing hunk in hyp/entry.S per Will's feedback

What about 4.19.y and 4.14.y trees? I can't take something for 4.9.y and then have a regression if someone moves to a newer release, right?

Sure, send you candidates for 4.14 and 4.19.

Greg, did you have a chance to queue those changes for 4.9, 4.14 and 4.19?

https://lore.kernel.org/linux-arm-kernel/[email protected]/
https://lore.kernel.org/linux-arm-kernel/[email protected]/
https://lore.kernel.org/linux-arm-kernel/[email protected]/

Quoted 7 lines
Quoted 1 line

Nope, I was waiting for Will's "ack" for these.

This patch doesn't even build for me (the 'sb' macro is not defined in 4.9), and I really wonder why we bother backporting it at all. Nobody's ever shown it to be a problem in practice, and it's clear that this is just being submitted to tick a box rather than anything else (otherwise it would build, right?).

Doh, I completely missed submitting the patch this depended on that's why I did not notice the build failure locally, sorry about that, what a shame.

Would not be the same "tick a box" argument be used against your original submission then? Sure, I have not been able to demonstrate in real life this was a problem, however the same can be said about a lot security related fixes.

What if it becomes exploitable in the future, would not it be nice to have it in a 6 year LTS kernel?

Quoted 3 lines

So I'm not going to Ack any of them. As with a lot of this side-channel stuff the cure is far worse than the disease.

Assuming that my v3 does build correctly, which it will, would you be keen on changing your position?

Signature · 5 lines
Florian
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
#10 lore raw

Hi Florian,

Quoted 17 lines · Florian Fainelli
Quoted 8 lines · Will Deacon
Quoted 7 lines · Greg KH
Quoted 5 lines · Florian Fainelli

Greg, did you have a chance to queue those changes for 4.9, 4.14 and 4.19?

https://lore.kernel.org/linux-arm-kernel/[email protected]/
https://lore.kernel.org/linux-arm-kernel/[email protected]/
https://lore.kernel.org/linux-arm-kernel/[email protected]/

Quoted 7 lines
Quoted 1 line

Nope, I was waiting for Will's "ack" for these.

This patch doesn't even build for me (the 'sb' macro is not defined in 4.9), and I really wonder why we bother backporting it at all. Nobody's ever shown it to be a problem in practice, and it's clear that this is just being submitted to tick a box rather than anything else (otherwise it would build, right?).

Quoted 8 lines

Doh, I completely missed submitting the patch this depended on that's why I did not notice the build failure locally, sorry about that, what a shame.

Would not be the same "tick a box" argument be used against your original submission then? Sure, I have not been able to demonstrate in real life this was a problem, however the same can be said about a lot security related fixes.

Sort of, although I wrote the original patch because it was dead easy to do and saved having to think too much about the problem, whereas the complexity of backporting largerly diminishes that imo.

Quoted 2 lines

What if it becomes exploitable in the future, would not it be nice to have it in a 6 year LTS kernel?

Even if people are stuck on an old LTS, they should still be taking the regular updates for it, and we would obviously need to backport the fix if it turned out to be exploitable (and hey, we could even test it then!).

Quoted 4 lines
Quoted 2 lines

So I'm not going to Ack any of them. As with a lot of this side-channel stuff the cure is far worse than the disease.

Assuming that my v3 does build correctly, which it will, would you be keen on changing your position?

Note that I'm not trying to block this patch from going in, I'm just saying that I'm not supportive of it. Perhaps somebody from Arm can review it if they think it's worth the effort.

Will
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm

#11 lore raw
Quoted 40 lines · Will Deacon

Hi Florian,

Quoted 17 lines · Florian Fainelli
Quoted 8 lines · Will Deacon
Quoted 7 lines · Greg KH
Quoted 5 lines · Florian Fainelli

Greg, did you have a chance to queue those changes for 4.9, 4.14 and 4.19?

https://lore.kernel.org/linux-arm-kernel/[email protected]/
https://lore.kernel.org/linux-arm-kernel/[email protected]/
https://lore.kernel.org/linux-arm-kernel/[email protected]/

Quoted 7 lines
Quoted 1 line

Nope, I was waiting for Will's "ack" for these.

This patch doesn't even build for me (the 'sb' macro is not defined in 4.9), and I really wonder why we bother backporting it at all. Nobody's ever shown it to be a problem in practice, and it's clear that this is just being submitted to tick a box rather than anything else (otherwise it would build, right?).

Quoted 8 lines

Doh, I completely missed submitting the patch this depended on that's why I did not notice the build failure locally, sorry about that, what a shame.

Would not be the same "tick a box" argument be used against your original submission then? Sure, I have not been able to demonstrate in real life this was a problem, however the same can be said about a lot security related fixes.

Sort of, although I wrote the original patch because it was dead easy to do and saved having to think too much about the problem, whereas the complexity of backporting largerly diminishes that imo.

Quoted 2 lines

What if it becomes exploitable in the future, would not it be nice to have it in a 6 year LTS kernel?

Even if people are stuck on an old LTS, they should still be taking the regular updates for it, and we would obviously need to backport the fix if it turned out to be exploitable (and hey, we could even test it then!).

Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

So I'm not going to Ack any of them. As with a lot of this side-channel stuff the cure is far worse than the disease.

Assuming that my v3 does build correctly, which it will, would you be keen on changing your position?

Note that I'm not trying to block this patch from going in, I'm just saying that I'm not supportive of it. Perhaps somebody from Arm can review it if they think it's worth the effort.

How about I submit the actual full series (two patches) and we take the discussion from there?

Thanks for responding!

Signature · 5 lines
Florian
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm