Thread (19 messages) 19 messages, 9 authors, 5d ago

Re: [PATCH v4 5/8] powerpc: add exit_flags field in pt_regs

From: Dmitry V. Levin <hidden>
Date: 2026-07-23 11:10:23
Also in: lkml

On Thu, Jul 23, 2026 at 03:02:38PM +0530, Mukesh Kumar Chaurasiya wrote:
On Wed, Jul 22, 2026 at 10:01:55AM +0300, Dmitry V. Levin wrote:
quoted
On Fri, Jan 23, 2026 at 01:09:13PM +0530, Mukesh Kumar Chaurasiya wrote:
quoted
From: Mukesh Kumar Chaurasiya <redacted>

Add a new field `exit_flags` in the pt_regs structure. This field will hold
the flags set during interrupt or syscall execution that are required during
exit to user mode.
Apparently, this change breaks PTRACE_GETREGS as the latter has no means
to communicate the size of struct pt_regs being written to userspace.

As result, the kernel with this patch will write to userpsace 32 more
bytes than before, and unless the userspace is re-compiled with the latest
ptrace.h, 32 bytes of userspace memory after struct pt_regs will be
overwritten.

This issue was detected by strace test suite, specifically, the test that
checks whether the actual number of bytes written by
ptrace(PTRACE_GETREGS) matches sizeof(struct pt_regs), and that test failed
with runtime version v7.1 and headers version v7.2-rc1.
[...]

Hey,

Can you try this fix? This should fix the ABI breakage.
[...]
quoted hunk ↗ jump to hunk
diff --git a/arch/powerpc/include/uapi/asm/ptrace.h b/arch/powerpc/include/uapi/asm/ptrace.h
index a393b7f2760a..01e630149d48 100644
--- a/arch/powerpc/include/uapi/asm/ptrace.h
+++ b/arch/powerpc/include/uapi/asm/ptrace.h
@@ -55,8 +55,6 @@ struct pt_regs
 	unsigned long dar;		/* Fault registers */
 	unsigned long dsisr;		/* on 4xx/Book-E used for ESR */
 	unsigned long result;		/* Result of a system call */
-	unsigned long exit_flags;	/* System call exit flags */
-	unsigned long __pt_regs_pad[3];	/* Maintain 16 byte interrupt stack alignment */
 };
 
 #endif /* __ASSEMBLER__ */
@@ -116,12 +114,10 @@ struct pt_regs
 #define PT_DAR	41
 #define PT_DSISR 42
 #define PT_RESULT 43
-#define PT_EXIT_FLAGS 44
-#define PT_PAD 47 /* 3 times */
-#define PT_DSCR 48
-#define PT_REGS_COUNT 48
+#define PT_DSCR 44
+#define PT_REGS_COUNT 44
 
-#define PT_FPR0	(PT_REGS_COUNT + 4)	/* each FP reg occupies 2 slots in this space */
+#define PT_FPR0	48	/* each FP reg occupies 2 slots in this space */
 
 #ifndef __powerpc64__
 
@@ -133,7 +129,7 @@ struct pt_regs
 #define PT_FPSCR (PT_FPR0 + 32)	/* each FP reg occupies 1 slot in 64-bit space */
 
 
-#define PT_VR0	(PT_FPSCR + 2)	/* <82> each Vector reg occupies 2 slots in 64-bit */
+#define PT_VR0 82	/* each Vector reg occupies 2 slots in 64-bit */
 #define PT_VSCR (PT_VR0 + 32*2 + 1)
 #define PT_VRSAVE (PT_VR0 + 33*2)
 
@@ -141,7 +137,7 @@ struct pt_regs
 /*
  * Only store first 32 VSRs here. The second 32 VSRs in VR0-31
  */
-#define PT_VSR0	(PT_VRSAVE + 2)	/* each VSR reg occupies 2 slots in 64-bit */
+#define PT_VSR0 150	/* each VSR reg occupies 2 slots in 64-bit */
 #define PT_VSR31 (PT_VSR0 + 2*31)
 #endif /* __powerpc64__ */
 
As this patch essentially reverts the uapi header to its v7.1 state,
the uapi part of the patch is definitely fine.


-- 
ldv
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help