* AW: strace not working with latest git
2008-09-10 15:52 strace not working with latest git Nylund, Martin
@ 2008-09-11 9:43 ` Nylund, Martin
2008-09-11 14:00 ` Nylund, Martin
` (3 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Nylund, Martin @ 2008-09-11 9:43 UTC (permalink / raw)
To: linux-sh
> Hello all,
>
> I haven't bee able to get strace working on my sh7722 target:
>
> $> uname -a
> Linux 192.168.105.52 2.6.27-rc6-00008-g3ed6d13-dirty #184 Wed
> Sep 10 17:31:44 CEST 2008 sh4al-dsp unknown
>
> $> strace ls
> restart_syscall(<... resuming interrupted call ...>) = 0
> .
> . <repeating many times>
> .
> restart_syscall(<... resuming interrupted call ...>) = 0
> restart_syscall(<... resuming interrupted call ...>) = 0
> bin etc linuxrc mplayer sbin tmp
> restart_syscall(<... resuming interrupted call ...>) = 0
> boot include man opt share usr
> restart_syscall(<... resuming interrupted call ...>) = 0
> dev lib mnt proc sys var
> restart_syscall(<... resuming interrupted call ...>) = 0
> ~ #
>
>
> I traced the problem a bit and in strace sources, the ptrace
> call PTRACE_PEEKUSR always returns 0 when fetching the
> syscall number. I also added a printk to PTRACE_PEEKUSR in
> arch/sh/kernel/ptrace_32.c and the returned value is always 0.
>
> What might be wrong here?
>
> Martin
I got strace now working with following hack. Still don't know why this
is needed though. get_stack_long() seemst to read data from wrong
offset...
diff --git a/arch/sh/kernel/ptrace_32.c b/arch/sh/kernel/ptrace_32.c
index 035cb30..7be388a 100644
--- a/arch/sh/kernel/ptrace_32.c
+++ b/arch/sh/kernel/ptrace_32.c
@@ -107,6 +107,7 @@ long arch_ptrace(struct task_struct *child, long
request, long addr, long data)
struct user * dummy = NULL;
int ret;
+ addr=(addr+0x34);
switch (request) {
/* read the word at location addr in the USER area. */
case PTRACE_PEEKUSR: {
^ permalink raw reply related [flat|nested] 6+ messages in thread* AW: strace not working with latest git
2008-09-10 15:52 strace not working with latest git Nylund, Martin
2008-09-11 9:43 ` AW: " Nylund, Martin
@ 2008-09-11 14:00 ` Nylund, Martin
2008-09-12 9:35 ` Paul Mundt
` (2 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Nylund, Martin @ 2008-09-11 14:00 UTC (permalink / raw)
To: linux-sh
> I haven't bee able to get strace working on my sh7722 target:
>
> $> uname -a
> Linux 192.168.105.52 2.6.27-rc6-00008-g3ed6d13-dirty #184 Wed
> Sep 10 17:31:44 CEST 2008 sh4al-dsp unknown
>
> $> strace ls
> restart_syscall(<... resuming interrupted call ...>) = 0
> .
> . <repeating many times>
> .
> restart_syscall(<... resuming interrupted call ...>) = 0
> restart_syscall(<... resuming interrupted call ...>) = 0
> bin etc linuxrc mplayer sbin tmp
> restart_syscall(<... resuming interrupted call ...>) = 0
> boot include man opt share usr
> restart_syscall(<... resuming interrupted call ...>) = 0
> dev lib mnt proc sys var
> restart_syscall(<... resuming interrupted call ...>) = 0
> ~ #
>
>
> I traced the problem a bit and in strace sources, the ptrace
> call PTRACE_PEEKUSR always returns 0 when fetching the
> syscall number. I also added a printk to PTRACE_PEEKUSR in
> arch/sh/kernel/ptrace_32.c and the returned value is always 0.
>
I got strace now working with following hack. Still don't know why this
is needed though. get_stack_long() seemst to read data from wrong
offset...
diff --git a/arch/sh/kernel/ptrace_32.c b/arch/sh/kernel/ptrace_32.c
index 035cb30..950b7e2 100644
--- a/arch/sh/kernel/ptrace_32.c
+++ b/arch/sh/kernel/ptrace_32.c
@@ -107,6 +107,9 @@ long arch_ptrace(struct task_struct *child, long
request, long addr, long data)
struct user * dummy = NULL;
int ret;
+ if(addr<0xff)
+ addr=(addr+0x34);
+
switch (request) {
/* read the word at location addr in the USER area. */
case PTRACE_PEEKUSR: {
^ permalink raw reply related [flat|nested] 6+ messages in thread* Re: strace not working with latest git
2008-09-10 15:52 strace not working with latest git Nylund, Martin
2008-09-11 9:43 ` AW: " Nylund, Martin
2008-09-11 14:00 ` Nylund, Martin
@ 2008-09-12 9:35 ` Paul Mundt
2008-09-12 9:54 ` AW: " Nylund, Martin
2008-09-12 10:39 ` Paul Mundt
4 siblings, 0 replies; 6+ messages in thread
From: Paul Mundt @ 2008-09-12 9:35 UTC (permalink / raw)
To: linux-sh
On Thu, Sep 11, 2008 at 04:00:16PM +0200, Nylund, Martin wrote:
>
> > I haven't bee able to get strace working on my sh7722 target:
> >
> > $> uname -a
> > Linux 192.168.105.52 2.6.27-rc6-00008-g3ed6d13-dirty #184 Wed
> > Sep 10 17:31:44 CEST 2008 sh4al-dsp unknown
> >
> > $> strace ls
> > restart_syscall(<... resuming interrupted call ...>) = 0
> > .
> > . <repeating many times>
> > .
> > restart_syscall(<... resuming interrupted call ...>) = 0
> > restart_syscall(<... resuming interrupted call ...>) = 0
> > bin etc linuxrc mplayer sbin tmp
> > restart_syscall(<... resuming interrupted call ...>) = 0
> > boot include man opt share usr
> > restart_syscall(<... resuming interrupted call ...>) = 0
> > dev lib mnt proc sys var
> > restart_syscall(<... resuming interrupted call ...>) = 0
> > ~ #
> >
> >
> > I traced the problem a bit and in strace sources, the ptrace
> > call PTRACE_PEEKUSR always returns 0 when fetching the
> > syscall number. I also added a printk to PTRACE_PEEKUSR in
> > arch/sh/kernel/ptrace_32.c and the returned value is always 0.
> >
>
> I got strace now working with following hack. Still don't know why this
> is needed though. get_stack_long() seemst to read data from wrong
> offset...
>
> diff --git a/arch/sh/kernel/ptrace_32.c b/arch/sh/kernel/ptrace_32.c
> index 035cb30..950b7e2 100644
> --- a/arch/sh/kernel/ptrace_32.c
> +++ b/arch/sh/kernel/ptrace_32.c
> @@ -107,6 +107,9 @@ long arch_ptrace(struct task_struct *child, long
> request, long addr, long data)
> struct user * dummy = NULL;
> int ret;
>
> + if(addr<0xff)
> + addr=(addr+0x34);
> +
> switch (request) {
> /* read the word at location addr in the USER area. */
> case PTRACE_PEEKUSR: {
>
This looks very much like the sizeof struct pt_dspregs. What is your
CONFIG_SH_DSP setting? Have you tried changing it and comparing?
^ permalink raw reply [flat|nested] 6+ messages in thread* AW: strace not working with latest git
2008-09-10 15:52 strace not working with latest git Nylund, Martin
` (2 preceding siblings ...)
2008-09-12 9:35 ` Paul Mundt
@ 2008-09-12 9:54 ` Nylund, Martin
2008-09-12 10:39 ` Paul Mundt
4 siblings, 0 replies; 6+ messages in thread
From: Nylund, Martin @ 2008-09-12 9:54 UTC (permalink / raw)
To: linux-sh
> On Thu, Sep 11, 2008 at 04:00:16PM +0200, Nylund, Martin wrote:
> >
> > > I haven't bee able to get strace working on my sh7722 target:
> > >
> > > $> uname -a
> > > Linux 192.168.105.52 2.6.27-rc6-00008-g3ed6d13-dirty #184 Wed
> > > Sep 10 17:31:44 CEST 2008 sh4al-dsp unknown
> > >
> > > $> strace ls
> > > restart_syscall(<... resuming interrupted call ...>) = 0
> > > .
> > > . <repeating many times>
> > > .
> > > restart_syscall(<... resuming interrupted call ...>) = 0
> > > restart_syscall(<... resuming interrupted call ...>) = 0
> > > bin etc linuxrc mplayer sbin tmp
> > > restart_syscall(<... resuming interrupted call ...>) = 0
> > > boot include man opt share usr
> > > restart_syscall(<... resuming interrupted call ...>) = 0
> > > dev lib mnt proc sys var
> > > restart_syscall(<... resuming interrupted call ...>) = 0
> > > ~ #
> > >
> > >
> > > I traced the problem a bit and in strace sources, the ptrace
> > > call PTRACE_PEEKUSR always returns 0 when fetching the
> > > syscall number. I also added a printk to PTRACE_PEEKUSR in
> > > arch/sh/kernel/ptrace_32.c and the returned value is always 0.
> > >
> >
> > I got strace now working with following hack. Still don't know why
> > this is needed though. get_stack_long() seemst to read data
> from wrong
> > offset...
> >
> > diff --git a/arch/sh/kernel/ptrace_32.c
> b/arch/sh/kernel/ptrace_32.c
> > index 035cb30..950b7e2 100644
> > --- a/arch/sh/kernel/ptrace_32.c
> > +++ b/arch/sh/kernel/ptrace_32.c
> > @@ -107,6 +107,9 @@ long arch_ptrace(struct task_struct
> *child, long
> > request, long addr, long data)
> > struct user * dummy = NULL;
> > int ret;
> >
> > + if(addr<0xff)
> > + addr=(addr+0x34);
> > +
> > switch (request) {
> > /* read the word at location addr in the USER area. */
> > case PTRACE_PEEKUSR: {
> >
> This looks very much like the sizeof struct pt_dspregs. What
> is your CONFIG_SH_DSP setting? Have you tried changing it and
> comparing?
Yes I had CONFIG_SH_DSP set. I now switched it off and strace works
without the above offset hack.
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: strace not working with latest git
2008-09-10 15:52 strace not working with latest git Nylund, Martin
` (3 preceding siblings ...)
2008-09-12 9:54 ` AW: " Nylund, Martin
@ 2008-09-12 10:39 ` Paul Mundt
4 siblings, 0 replies; 6+ messages in thread
From: Paul Mundt @ 2008-09-12 10:39 UTC (permalink / raw)
To: linux-sh
On Fri, Sep 12, 2008 at 11:54:48AM +0200, Nylund, Martin wrote:
> > On Thu, Sep 11, 2008 at 04:00:16PM +0200, Nylund, Martin wrote:
> > > diff --git a/arch/sh/kernel/ptrace_32.c
> > b/arch/sh/kernel/ptrace_32.c
> > > index 035cb30..950b7e2 100644
> > > --- a/arch/sh/kernel/ptrace_32.c
> > > +++ b/arch/sh/kernel/ptrace_32.c
> > > @@ -107,6 +107,9 @@ long arch_ptrace(struct task_struct
> > *child, long
> > > request, long addr, long data)
> > > struct user * dummy = NULL;
> > > int ret;
> > >
> > > + if(addr<0xff)
> > > + addr=(addr+0x34);
> > > +
> > > switch (request) {
> > > /* read the word at location addr in the USER area. */
> > > case PTRACE_PEEKUSR: {
> > >
> > This looks very much like the sizeof struct pt_dspregs. What
> > is your CONFIG_SH_DSP setting? Have you tried changing it and
> > comparing?
>
> Yes I had CONFIG_SH_DSP set. I now switched it off and strace works
> without the above offset hack.
That's interesting. It looks like CONFIG_SH_DSP needs some more attention
for SH4AL-DSP anyways. I'm currently rewriting the ptrace bits for
tracehook/regset conversion anyways, so we should get this fixed properly
there.
^ permalink raw reply [flat|nested] 6+ messages in thread