[Qemu-devel] [PATCH] gdbstub: fix sThreadInfo handler

[Qemu-devel] [PATCH] gdbstub: fix sThreadInfo handler

[Max Filippov]

After the commit 182735e cpu: Make first_cpu and next_cpu CPUState
we can no longer blindly use cpu->next_cpu->env_ptr to get CPUArchState
of the next CPU, as the next_cpu is NULL in the last CPU.

This fixes segfault caused by gdb command 'info threads'.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 gdbstub.c |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/gdbstub.c b/gdbstub.c
index 0ee82a9..deb8159 100644
--- a/gdbstub.c
+++ b/gdbstub.c
@@ -2406,10 +2406,11 @@ static int gdb_handle_packet(GDBState *s, const char *line_buf)
         } else if (strcmp(p,"sThreadInfo") == 0) {
         report_cpuinfo:
             if (s->query_cpu) {
+                CPUState *cpu = ENV_GET_CPU(s->query_cpu);
                 snprintf(buf, sizeof(buf), "m%x",
-                         cpu_index(ENV_GET_CPU(s->query_cpu)));
+                         cpu_index(cpu));
                 put_packet(s, buf);
-                s->query_cpu = ENV_GET_CPU(s->query_cpu)->next_cpu->env_ptr;
+                s->query_cpu = cpu->next_cpu ? cpu->next_cpu->env_ptr : NULL;
             } else
                 put_packet(s, "l");
             break;
-- 
1.7.7.6

Re: [Qemu-devel] [PATCH] gdbstub: fix sThreadInfo handler

[Andreas Färber]

Am 22.07.2013 07:24, schrieb Max Filippov:
> After the commit 182735e cpu: Make first_cpu and next_cpu CPUState
> we can no longer blindly use cpu->next_cpu->env_ptr to get CPUArchState
> of the next CPU, as the next_cpu is NULL in the last CPU.
> 
> This fixes segfault caused by gdb command 'info threads'.
> 
> Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>

This should be obsoleted by the patch on qom-cpu that I pointed you to,
no? I so far didn't send a pull for that alone since I was hoping to get
more gdbstub patches in - I added a comment explaining vaddr type as
requested by Peter, so a Reviewed-by/Acked-by for the modified Xtensa
subclasses would help push that forward, and no one bothered to review
the ..._read_register() changes yet IIRC, split up on rth's request. I
could just apply the initial set_pc stuff of course and wait some longer
with the rest.

Andreas

> ---
>  gdbstub.c |    5 +++--
>  1 files changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/gdbstub.c b/gdbstub.c
> index 0ee82a9..deb8159 100644
> --- a/gdbstub.c
> +++ b/gdbstub.c
> @@ -2406,10 +2406,11 @@ static int gdb_handle_packet(GDBState *s, const char *line_buf)
>          } else if (strcmp(p,"sThreadInfo") == 0) {
>          report_cpuinfo:
>              if (s->query_cpu) {
> +                CPUState *cpu = ENV_GET_CPU(s->query_cpu);
>                  snprintf(buf, sizeof(buf), "m%x",
> -                         cpu_index(ENV_GET_CPU(s->query_cpu)));
> +                         cpu_index(cpu));
>                  put_packet(s, buf);
> -                s->query_cpu = ENV_GET_CPU(s->query_cpu)->next_cpu->env_ptr;
> +                s->query_cpu = cpu->next_cpu ? cpu->next_cpu->env_ptr : NULL;
>              } else
>                  put_packet(s, "l");
>              break;
> 


-- 
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg

Re: [Qemu-devel] [PATCH] gdbstub: fix sThreadInfo handler

[Max Filippov]

On Mon, Jul 22, 2013 at 2:54 PM, Andreas Färber <afaerber@suse.de> wrote:
> Am 22.07.2013 07:24, schrieb Max Filippov:
>> After the commit 182735e cpu: Make first_cpu and next_cpu CPUState
>> we can no longer blindly use cpu->next_cpu->env_ptr to get CPUArchState
>> of the next CPU, as the next_cpu is NULL in the last CPU.
>>
>> This fixes segfault caused by gdb command 'info threads'.
>>
>> Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
>
> This should be obsoleted by the patch on qom-cpu that I pointed you to,
> no? I so far didn't send a pull for that alone since I was hoping to get
> more gdbstub patches in - I added a comment explaining vaddr type as

I've noticed that the other gdbstub fix is in, but debugging is still broken
in the mainline. In fact it had been broken just a few commits after that
fix, in the same pull request.

Holding a fix in order to make a bigger pull request later is not the right
thing IMHO.

> requested by Peter, so a Reviewed-by/Acked-by for the modified Xtensa
> subclasses would help push that forward, and no one bothered to review

Ok, I will review it.

> the ..._read_register() changes yet IIRC, split up on rth's request. I
> could just apply the initial set_pc stuff of course and wait some longer
> with the rest.

-- 
Thanks.
-- Max