Skip to contentSkip to conversation

microblaze: fix signal-frame handling for NPTL

PATCH Superseded Ramin Moussavi 2 participants 9 messages Jun 4 – Jun 5 1 message not in the archive Microblaze architecture

lore
Share & tools
lore
#0 lore raw

cover microblaze: fix signal-frame handling for NPTL

Two independent fixes to the microblaze signal-delivery path, both found while bringing the uClibc-ng NPTL test suite up on microblazeel under qemu-system (petalogix-s3adsp1800).

Patch 1 wires up sigaltstack(), which is currently routed to sys_ni_syscall even though the signal code fully supports an alternate stack - microblaze is the only architecture leaving it unimplemented.

Patch 2 reserves the ABI argument-home area at the top of the signal frame. The MicroBlaze calling convention lets a handler store its incoming register arguments into [r1+4]..[r1+28]; since r1 points at struct rt_sigframe on entry, those stores land in siginfo/ucontext and corrupt the signal state. An eight-word leading gap moves them into scratch space.

Both were tested with the uClibc-ng NPTL tests; the affected tests (tst-cancel20/21, tst-cancelx20/21, tst-signal6, tst-timer4/5, tst-mqueue5) now pass. checkpatch --strict is clean and both apply to current linux-next.

Ramin Moussavi (2):
microblaze: wire up sigaltstack
microblaze: reserve the ABI argument-home area in the signal frame

 arch/microblaze/kernel/signal.c             | 7 +++++++
 arch/microblaze/kernel/syscalls/syscall.tbl | 2 +-
 2 files changed, 8 insertions(+), 1 deletion(-)
Signature · 1 line
2.53.0
#1 lore raw

PATCH 1/2 microblaze: wire up sigaltstack

sigaltstack is wired to sys_ni_syscall - microblaze is the only architecture without it - although the microblaze signal delivery code fully supports the alternate signal stack: get_sigframe() picks the stack via sigsp(), setup_rt_frame() saves it with __save_altstack() and sys_rt_sigreturn() calls restore_altstack().

Wire it up to sys_sigaltstack.

Tested on qemu petalogix-s3adsp1800 (microblazeel) with the uClibc-ng test suite: the five sigaltstack-dependent NPTL tests (tst-cancel20/21, tst-cancelx20/21, tst-signal6) pass; before this change sigaltstack() returned ENOSYS.

---
 arch/microblaze/kernel/syscalls/syscall.tbl | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/microblaze/kernel/syscalls/syscall.tbl +1 -1
index 223d26303..b4ce48e8a 100644
@@ -193,7 +193,7 @@
 183	common	getcwd				sys_getcwd
 184	common	capget				sys_capget
 185	common	capset				sys_capset
-186	common	sigaltstack			sys_ni_syscall
+186	common	sigaltstack			sys_sigaltstack
 187	common	sendfile			sys_sendfile
 188	common	getpmsg				sys_ni_syscall
 189	common	putpmsg				sys_ni_syscall
Signature · 1 line
2.53.0
#2 lore raw

PATCH 2/2 microblaze: reserve the ABI argument-home area in the signal frame

In reply to Ramin Moussavi

The MicroBlaze procedure call standard lets a callee store its incoming register arguments r5..r10 into the caller-provided home slots at [r1+4]..[r1+28]. When the kernel enters a signal handler it sets r1 to point at struct rt_sigframe, whose leading members are the siginfo and ucontext prepared for the handler. A handler that homes its arguments - which an unoptimised (-O0) build always does - therefore overwrites the start of siginfo/ucontext, corrupting the signal state that the handler and sys_rt_sigreturn() depend on.

Reserve the home area by making an eight-word gap the first member of struct rt_sigframe, so the handler's argument stores land in scratch space instead of clobbering siginfo.

Tested on qemu-system-microblazeel (petalogix-s3adsp1800) with the uClibc-ng NPTL test suite: tst-timer4, tst-timer5, tst-mqueue5 and tst-signal6 pass; before this change they failed because the handler clobbered the signal frame.

---
 arch/microblaze/kernel/signal.c | 7 +++++++
 1 file changed, 7 insertions(+)
arch/microblaze/kernel/signal.c +7 -0
index c78a0ff48..df9c07d77 100644
@@ -49,6 +49,13 @@ struct sigframe {
 };
 
 struct rt_sigframe {
+	/*
+	 * Home area for the handler's register arguments: the MicroBlaze
+	 * ABI lets the callee store r5..r10 at [r1+4]..[r1+28], and r1
+	 * points at this frame when the handler is entered.  Without the
+	 * gap those stores corrupt info/uc.
+	 */
+	unsigned long abi_gap[8];
 	struct siginfo info;
 	struct ucontext uc;
 	unsigned long tramp[2];	/* signal trampoline */
Signature · 1 line
2.53.0
maintainer
#3 lore raw

Re: [PATCH 2/2] microblaze: reserve the ABI argument-home area in the signal frame

Quoted 28 lines · Ramin Moussavi

The MicroBlaze procedure call standard lets a callee store its incoming register arguments r5..r10 into the caller-provided home slots at [r1+4]..[r1+28]. When the kernel enters a signal handler it sets r1 to point at struct rt_sigframe, whose leading members are the siginfo and ucontext prepared for the handler. A handler that homes its arguments - which an unoptimised (-O0) build always does - therefore overwrites the start of siginfo/ucontext, corrupting the signal state that the handler and sys_rt_sigreturn() depend on.

Reserve the home area by making an eight-word gap the first member of struct rt_sigframe, so the handler's argument stores land in scratch space instead of clobbering siginfo.

Tested on qemu-system-microblazeel (petalogix-s3adsp1800) with the uClibc-ng NPTL test suite: tst-timer4, tst-timer5, tst-mqueue5 and tst-signal6 pass; before this change they failed because the handler clobbered the signal frame.

---
  arch/microblaze/kernel/signal.c | 7 +++++++
  1 file changed, 7 insertions(+)
arch/microblaze/kernel/signal.c +0 -0
index c78a0ff48..df9c07d77 100644

arch/microblaze/kernel/signal.c struct sigframe { · line 54

Quoted 6 lines
arch/microblaze/kernel/signal.c +3 -0
@@ -49,6 +49,13 @@ struct sigframe {
  };
 
  struct rt_sigframe {
+	/*
+	 * Home area for the handler's register arguments: the MicroBlaze
+	 * ABI lets the callee store r5..r10 at [r1+4]..[r1+28], and r1

Actually I have created similar patch a month ago but I have only 7 spaces here.

r15 - r1 + 0
r5 - r1+4
r6 - r1+8
r7 - r1+12
r8 - r1+16
r9 - r1+20
r10 - r1+24

It means 7 should be enough. Can you please retest it?

Thanks,
Michal

Signature · 6 lines
Michal Simek, Ing. (M.Eng), OpenPGP -> KeyID: FE3D1F91
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel - Xilinx Microblaze
Maintainer of Linux kernel - Xilinx Zynq ARM and ZynqMP/Versal ARM64 SoCs
U-Boot custodian - Xilinx Microblaze/Zynq/ZynqMP/Versal/Versal NET SoCs
TF-A maintainer - Xilinx ZynqMP/Versal/Versal NET SoCs
maintainer
#5 lore raw

Re: [PATCH 2/2] microblaze: reserve the ABI argument-home area in the signal frame

In reply to a message not in the archive

Hi,

Quoted 2 lines · Ramin Moussavi

Hi Michal,

please avoid top posting next time.

Quoted 4 lines

  You're right, 7 is enough - the home area is exactly r15 at r1+0 and r5..r10 at r1+4..r1+24. I only picked 8 because it looked like a rounder number than 7; the 8th word is never written by the ABI, so 7 is the correct minimal size. Please go ahead with your version.

you are sending 2 patches. Just send a v2 version that should be fine with updated value.

Thanks,
Michal

Signature · 6 lines
Michal Simek, Ing. (M.Eng), OpenPGP -> KeyID: FE3D1F91
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel - Xilinx Microblaze
Maintainer of Linux kernel - Xilinx Zynq ARM and ZynqMP/Versal ARM64 SoCs
U-Boot custodian - Xilinx Microblaze/Zynq/ZynqMP/Versal/Versal NET SoCs
TF-A maintainer - Xilinx ZynqMP/Versal/Versal NET SoCs
#6 lore raw

v2 cover microblaze: fix signal-frame handling for NPTL

In reply to Ramin Moussavi

Two independent fixes to the microblaze signal-delivery path, both found while bringing the uClibc-ng NPTL test suite up on microblazeel under qemu-system (petalogix-s3adsp1800).

Patch 1 wires up sigaltstack(), which is currently routed to sys_ni_syscall even though the signal code fully supports an alternate stack - microblaze is the only architecture leaving it unimplemented.

Patch 2 reserves the ABI argument-home area at the top of the signal frame. The MicroBlaze calling convention reserves [r1+0] for the return address and lets a handler store its incoming register arguments r5..r10 into [r1+4]..[r1+24]; since r1 points at struct rt_sigframe on entry, those stores land in siginfo/ucontext and corrupt the signal state. A seven-word leading gap moves them into scratch space.

Both were tested with the uClibc-ng NPTL tests; the affected tests (tst-cancel20/21, tst-cancelx20/21, tst-signal6, tst-timer4/5, tst-mqueue5) pass. checkpatch --strict is clean and both apply to current linux-next.

Changes since v1 [1]:
- Patch 2: shrink the gap from eight to seven words. Michal pointed out that the home area is exactly seven words (r15 at r1+0, r5..r10 at r1+4..r1+24); the eighth word was never written by the ABI. Retested on qemu - tst-timer4/5, tst-mqueue5, tst-signal6 still pass.
- Patch 2: fix the offset range in the commit message and the in-code comment ([r1+4]..[r1+24], not [r1+28]).
- Patch 1: unchanged.

[1] https://lore.kernel.org/all/[email protected]/

Ramin Moussavi (2):
microblaze: wire up sigaltstack
microblaze: reserve the ABI argument-home area in the signal frame

 arch/microblaze/kernel/signal.c             | 8 ++++++++
 arch/microblaze/kernel/syscalls/syscall.tbl | 2 +-
 2 files changed, 9 insertions(+), 1 deletion(-)
Signature · 1 line
2.53.0
#7 lore raw

PATCH v2 1/2 microblaze: wire up sigaltstack

sigaltstack is wired to sys_ni_syscall - microblaze is the only architecture without it - although the microblaze signal delivery code fully supports the alternate signal stack: get_sigframe() picks the stack via sigsp(), setup_rt_frame() saves it with __save_altstack() and sys_rt_sigreturn() calls restore_altstack().

Wire it up to sys_sigaltstack.

Tested on qemu petalogix-s3adsp1800 (microblazeel, kernel 6.5.10 and syscall table unchanged in current mainline) with the uClibc-ng test
suite: the five sigaltstack-dependent NPTL tests (tst-cancel20/21, tst-cancelx20/21, tst-signal6) pass; before this change sigaltstack() returned ENOSYS.

---
 arch/microblaze/kernel/syscalls/syscall.tbl | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/microblaze/kernel/syscalls/syscall.tbl +1 -1
index 223d26303..b4ce48e8a 100644
@@ -193,7 +193,7 @@
 183	common	getcwd				sys_getcwd
 184	common	capget				sys_capget
 185	common	capset				sys_capset
-186	common	sigaltstack			sys_ni_syscall
+186	common	sigaltstack			sys_sigaltstack
 187	common	sendfile			sys_sendfile
 188	common	getpmsg				sys_ni_syscall
 189	common	putpmsg				sys_ni_syscall
Signature · 1 line
2.53.0
#8 lore raw

PATCH v2 2/2 microblaze: reserve the ABI argument-home area in the signal frame

In reply to Ramin Moussavi

The MicroBlaze procedure call standard reserves [r1+0] for the return address and lets a callee store its incoming register arguments r5..r10 into the caller-provided home slots at [r1+4]..[r1+24]. When the kernel enters a signal handler it sets r1 to point at struct rt_sigframe, whose leading members are the siginfo and ucontext prepared for the handler. A handler that homes its arguments - which an unoptimised (-O0) build always does - therefore overwrites the start of siginfo/ucontext, corrupting the signal state that the handler and sys_rt_sigreturn() depend on.

Reserve the home area by making a seven-word gap the first member of struct rt_sigframe, so the handler's argument stores land in scratch space instead of clobbering siginfo.

Tested on qemu-system-microblazeel (petalogix-s3adsp1800) with the uClibc-ng NPTL test suite: tst-timer4, tst-timer5, tst-mqueue5 and tst-signal6 pass; before this change they failed because the handler clobbered the signal frame.

---
 arch/microblaze/kernel/signal.c | 8 ++++++++
 1 file changed, 8 insertions(+)
arch/microblaze/kernel/signal.c +8 -0
index c78a0ff48..9cca06d60 100644
@@ -49,6 +49,14 @@ struct sigframe {
 };
 
 struct rt_sigframe {
+	/*
+	 * Home area for the handler's register arguments: the MicroBlaze
+	 * ABI reserves [r1+0] for the return address and lets the callee
+	 * store r5..r10 at [r1+4]..[r1+24], and r1 points at this frame
+	 * when the handler is entered.  Without the gap those stores
+	 * corrupt info/uc.
+	 */
+	unsigned long abi_gap[7];
 	struct siginfo info;
 	struct ucontext uc;
 	unsigned long tramp[2];	/* signal trampoline */
Signature · 1 line
2.53.0
maintainer
#9 lore raw
In reply to Ramin Moussavi
Quoted 31 lines · Ramin Moussavi

Two independent fixes to the microblaze signal-delivery path, both found while bringing the uClibc-ng NPTL test suite up on microblazeel under qemu-system (petalogix-s3adsp1800).

Patch 1 wires up sigaltstack(), which is currently routed to sys_ni_syscall even though the signal code fully supports an alternate stack - microblaze is the only architecture leaving it unimplemented.

Patch 2 reserves the ABI argument-home area at the top of the signal frame. The MicroBlaze calling convention reserves [r1+0] for the return address and lets a handler store its incoming register arguments r5..r10 into [r1+4]..[r1+24]; since r1 points at struct rt_sigframe on entry, those stores land in siginfo/ucontext and corrupt the signal state. A seven-word leading gap moves them into scratch space.

Both were tested with the uClibc-ng NPTL tests; the affected tests (tst-cancel20/21, tst-cancelx20/21, tst-signal6, tst-timer4/5, tst-mqueue5) pass. checkpatch --strict is clean and both apply to current linux-next.

Changes since v1 [1]:
- Patch 2: shrink the gap from eight to seven words. Michal pointed out that the home area is exactly seven words (r15 at r1+0, r5..r10 at r1+4..r1+24); the eighth word was never written by the ABI. Retested on qemu - tst-timer4/5, tst-mqueue5, tst-signal6 still pass.
- Patch 2: fix the offset range in the commit message and the in-code

    comment ([r1+4]..[r1+24], not [r1+28]).

- Patch 1: unchanged.

[1] https://lore.kernel.org/all/[email protected]/

Quoted 8 lines

Ramin Moussavi (2):
microblaze: wire up sigaltstack
microblaze: reserve the ABI argument-home area in the signal frame

  arch/microblaze/kernel/signal.c             | 8 ++++++++
  arch/microblaze/kernel/syscalls/syscall.tbl | 2 +-
  2 files changed, 9 insertions(+), 1 deletion(-)

2 issues here.

1. Do not send it as reply to v1 but separately.
2. Author of patch is not matching SOB in the email You are sending it from
Ramin Moussavi <[email protected]>
and SOB has
Ramin Moussavi <[email protected]>

Please fix it and send v3

Thanks,
Michal

Signature · 6 lines
Michal Simek, Ing. (M.Eng), OpenPGP -> KeyID: FE3D1F91
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel - Xilinx Microblaze
Maintainer of Linux kernel - Xilinx Zynq ARM and ZynqMP/Versal ARM64 SoCs
U-Boot custodian - Xilinx Microblaze/Zynq/ZynqMP/Versal/Versal NET SoCs
TF-A maintainer - Xilinx ZynqMP/Versal/Versal NET SoCs