linux-sh.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH] sh: Fix ptrace FPU state initialisation
@ 2011-02-01 17:37 Phil Edworthy
  2011-03-18 14:16 ` [PATCH] sh: Fix ptrace fpu " Phil Edworthy
  2011-03-23 13:25 ` Paul Mundt
  0 siblings, 2 replies; 3+ messages in thread
From: Phil Edworthy @ 2011-02-01 17:37 UTC (permalink / raw)
  To: linux-sh

RFC:

Commit 0ea820cf introduced the PTRACE_GETFPREGS/SETFPREGS cmds,
but gdb-server still accesses the FPU state using the
PTRACE_PEEKUSR/POKEUSR commands. In this case, xstate was not
initialised. The patch below against 2.6.35 fixes this. I'll
rework this for head if it's ok.

However, should we still allow PEEKUSR/POKEUSR for FPU registers,
or should we return an error?

---
 arch/sh/kernel/ptrace_32.c |    9 ++++++++-
 1 files changed, 8 insertions(+), 1 deletions(-)

diff --git a/arch/sh/kernel/ptrace_32.c b/arch/sh/kernel/ptrace_32.c
index 6c4bbba..df0144c 100644
--- a/arch/sh/kernel/ptrace_32.c
+++ b/arch/sh/kernel/ptrace_32.c
@@ -363,9 +363,13 @@ long arch_ptrace(struct task_struct *child, long
request, long addr, long data)
 					tmp = FPSCR_INIT;
 				else
 					tmp = 0;
-			} else
+			} else {
+				ret = init_fpu(child);
+				if (ret)
+					break;
 				tmp = ((long *)child->thread.xstate)
 					[(addr - (long)&dummy->fpu) >> 2];
+			}
 		} else if (addr = (long) &dummy->u_fpvalid)
 			tmp = !!tsk_used_math(child);
 		else if (addr = PT_TEXT_ADDR)
@@ -392,6 +396,9 @@ long arch_ptrace(struct task_struct *child, long
request, long addr, long data)
 			ret = put_stack_long(child, addr, data);
 		else if (addr >= (long) &dummy->fpu &&
 			 addr < (long) &dummy->u_fpvalid) {
+			ret = init_fpu(child);
+			if (ret)
+				break;
 			set_stopped_child_used_math(child);
 			((long *)child->thread.xstate)
 				[(addr - (long)&dummy->fpu) >> 2] = data;
-- 
1.7.1

^ permalink raw reply related	[flat|nested] 3+ messages in thread

* [PATCH] sh: Fix ptrace fpu state initialisation
  2011-02-01 17:37 [PATCH] sh: Fix ptrace FPU state initialisation Phil Edworthy
@ 2011-03-18 14:16 ` Phil Edworthy
  2011-03-23 13:25 ` Paul Mundt
  1 sibling, 0 replies; 3+ messages in thread
From: Phil Edworthy @ 2011-03-18 14:16 UTC (permalink / raw)
  To: linux-sh

Commit 0ea820cf introduced the PTRACE_GETFPREGS/SETFPREGS cmds,
but gdb-server still accesses the FPU state using the
PTRACE_PEEKUSR/POKEUSR commands. In this case, xstate was not
initialised.

Signed-off-by: Phil Edworthy <phil.edworthy@renesas.com>
---
 arch/sh/kernel/ptrace_32.c |    6 ++++++
 arch/sh/kernel/ptrace_64.c |    6 ++++++
 2 files changed, 12 insertions(+), 0 deletions(-)

diff --git a/arch/sh/kernel/ptrace_32.c b/arch/sh/kernel/ptrace_32.c
index 90a15d2..b53664e 100644
--- a/arch/sh/kernel/ptrace_32.c
+++ b/arch/sh/kernel/ptrace_32.c
@@ -392,6 +392,9 @@ long arch_ptrace(struct task_struct *child, long request,
 					tmp = 0;
 			} else {
 				unsigned long index;
+				ret = init_fpu(child);
+				if (ret)
+					break;
 				index = addr - offsetof(struct user, fpu);
 				tmp = ((unsigned long *)child->thread.xstate)
 					[index >> 2];
@@ -423,6 +426,9 @@ long arch_ptrace(struct task_struct *child, long request,
 		else if (addr >= offsetof(struct user, fpu) &&
 			 addr < offsetof(struct user, u_fpvalid)) {
 			unsigned long index;
+			ret = init_fpu(child);
+			if (ret)
+				break;
 			index = addr - offsetof(struct user, fpu);
 			set_stopped_child_used_math(child);
 			((unsigned long *)child->thread.xstate)
diff --git a/arch/sh/kernel/ptrace_64.c b/arch/sh/kernel/ptrace_64.c
index 4436eac..c8f9764 100644
--- a/arch/sh/kernel/ptrace_64.c
+++ b/arch/sh/kernel/ptrace_64.c
@@ -403,6 +403,9 @@ long arch_ptrace(struct task_struct *child, long request,
 		else if ((addr >= offsetof(struct user, fpu)) &&
 			 (addr <  offsetof(struct user, u_fpvalid))) {
 			unsigned long index;
+			ret = init_fpu(child);
+			if (ret)
+				break;
 			index = addr - offsetof(struct user, fpu);
 			tmp = get_fpu_long(child, index);
 		} else if (addr = offsetof(struct user, u_fpvalid)) {
@@ -442,6 +445,9 @@ long arch_ptrace(struct task_struct *child, long request,
 		else if ((addr >= offsetof(struct user, fpu)) &&
 			 (addr <  offsetof(struct user, u_fpvalid))) {
 			unsigned long index;
+			ret = init_fpu(child);
+			if (ret)
+				break;
 			index = addr - offsetof(struct user, fpu);
 			ret = put_fpu_long(child, index, data);
 		}
-- 
1.7.1


^ permalink raw reply related	[flat|nested] 3+ messages in thread

* Re: [PATCH] sh: Fix ptrace fpu state initialisation
  2011-02-01 17:37 [PATCH] sh: Fix ptrace FPU state initialisation Phil Edworthy
  2011-03-18 14:16 ` [PATCH] sh: Fix ptrace fpu " Phil Edworthy
@ 2011-03-23 13:25 ` Paul Mundt
  1 sibling, 0 replies; 3+ messages in thread
From: Paul Mundt @ 2011-03-23 13:25 UTC (permalink / raw)
  To: linux-sh

On Fri, Mar 18, 2011 at 02:16:31PM +0000, Phil Edworthy wrote:
> Commit 0ea820cf introduced the PTRACE_GETFPREGS/SETFPREGS cmds,
> but gdb-server still accesses the FPU state using the
> PTRACE_PEEKUSR/POKEUSR commands. In this case, xstate was not
> initialised.
> 
> Signed-off-by: Phil Edworthy <phil.edworthy@renesas.com>

Applied, thanks.

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2011-03-23 13:25 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2011-02-01 17:37 [PATCH] sh: Fix ptrace FPU state initialisation Phil Edworthy
2011-03-18 14:16 ` [PATCH] sh: Fix ptrace fpu " Phil Edworthy
2011-03-23 13:25 ` Paul Mundt

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).