Just found out that by 'echo sth_wrong > /sys/power/state' the kernel would
oops in a fatal way (no clean exit from there).
The oops suggested that the code would enter an invalid fn.
You may apply the included patch to solve the bug. IMHO doing a clean exit is
much preferrable than having BUG() there.
In the patch I also added some double checking of the fn. You may disable that
(although sleeping is a one time fn which won't be slowed dow by that).
Status report: (HP Omnibook XE3GC)
S1 ('standby') works
S3, S4bios will suspend, but NOT resume.
that's the same as previous kernels I've tested so far.
I still have the ALSA-artsd deadlock I reported w. -test1 (in short, artsd
resumes from S1 in a 'disk I/O' process state, which deadlocks maestro3)
see: http://marc.theaimsgroup.com/?l=linux-kernel&m=105836807605512&w=2
Bernd Petrovitsch wrote:
> On Tue, 2003-08-26 at 00:25, P. Christeas wrote:
> > Just found out that by 'echo sth_wrong > /sys/power/state' the kernel
> > would oops in a fatal way (no clean exit from there).
> > The oops suggested that the code would enter an invalid fn.
> >
> > You may apply the included patch to solve the bug. IMHO doing a clean
> > exit is much preferrable than having BUG() there.
>
> > diff -Bbur /diskb/users/panos/linux-off/kernel/power/main.c
> > /usr/src/linux/kernel/power/main.c ---
> > /diskb/users/panos/linux-off/kernel/power/main.c 2003-08-23
> > 12:13:17.000000000 +0300 +++
> > /usr/src/linux/kernel/power/main.c 2003-08-26 00:59:34.000000000 +0300 @@
> > -500,7 +514,7 @@
> > if (s->name && !strcmp(buf,s->name))
> > break;
> > }
> > - if (s)
> > + if ( (s) && (state < PM_SUSPEND_MAX) )
> > error = enter_state(state);
> > else
> > error = -EINVAL;
>
> What do you think about the attached patch to solve the bug and remove a
> warning?
>
> Bernd
Already tried that. If you look more closely, the s will receive the state
*before* the name *in it* is strcmp()'ed. This means it won't be NULL anyway.
> if (s->name && !strcmp(buf,s->name))
> break;
I also thought of checking "( (s) && (s->name !=NULL) )" , but IMHO the
'state' check is cleaner (no dereference).
--- linux-2.6.0-test4/kernel/power/main.c Sat Aug 23 01:53:13 2003
+++ linux-2.6.0-test4-patched/kernel/power/main.c Mon Aug 25 21:16:50
2003
@@ -492,7 +492,7 @@
static ssize_t state_store(struct subsystem * subsys, const char * buf,
size_t n)
{
u32 state;
- struct pm_state * s;
+ struct pm_state * s = NULL;
int error;
for (state = 0; state < PM_SUSPEND_MAX; state++) {
On Tue, 2003-08-26 at 00:25, P. Christeas wrote:
> Just found out that by 'echo sth_wrong > /sys/power/state' the kernel would
> oops in a fatal way (no clean exit from there).
> The oops suggested that the code would enter an invalid fn.
>
> You may apply the included patch to solve the bug. IMHO doing a clean exit is
> much preferrable than having BUG() there.
[...]
> ----
[...]
> diff -Bbur /diskb/users/panos/linux-off/kernel/power/main.c /usr/src/linux/kernel/power/main.c
> --- /diskb/users/panos/linux-off/kernel/power/main.c 2003-08-23 12:13:17.000000000 +0300
> +++ /usr/src/linux/kernel/power/main.c 2003-08-26 00:59:34.000000000 +0300
> @@ -500,7 +514,7 @@
> if (s->name && !strcmp(buf,s->name))
> break;
> }
> - if (s)
> + if ( (s) && (state < PM_SUSPEND_MAX) )
> error = enter_state(state);
> else
> error = -EINVAL;
What do you think about the attached patch to solve the bug and remove a
warning?
Bernd
--
Firmix Software GmbH http://www.firmix.at/
mobil: +43 664 4416156 fax: +43 1 7890849-55
Embedded Linux Development and Services
On Tue, 2003-08-26 at 00:47, P. Christeas wrote:
> Bernd Petrovitsch wrote:
> > On Tue, 2003-08-26 at 00:25, P. Christeas wrote:
> > > Just found out that by 'echo sth_wrong > /sys/power/state' the kernel
> > > would oops in a fatal way (no clean exit from there).
> > > The oops suggested that the code would enter an invalid fn.
> > >
> > > You may apply the included patch to solve the bug. IMHO doing a clean
> > > exit is much preferrable than having BUG() there.
> >
> > > diff -Bbur /diskb/users/panos/linux-off/kernel/power/main.c
> > > /usr/src/linux/kernel/power/main.c ---
> > > /diskb/users/panos/linux-off/kernel/power/main.c 2003-08-23
> > > 12:13:17.000000000 +0300 +++
> > > /usr/src/linux/kernel/power/main.c 2003-08-26 00:59:34.000000000 +0300 @@
> > > -500,7 +514,7 @@
> > > if (s->name && !strcmp(buf,s->name))
> > > break;
> > > }
> > > - if (s)
> > > + if ( (s) && (state < PM_SUSPEND_MAX) )
> > > error = enter_state(state);
> > > else
> > > error = -EINVAL;
> >
> > What do you think about the attached patch to solve the bug and remove a
> > warning?
[...]
> Already tried that. If you look more closely, the s will receive the state
> *before* the name *in it* is strcmp()'ed. This means it won't be NULL anyway.
> > if (s->name && !strcmp(buf,s->name))
> > break;
*arrgl*, yes. Could only change somthing if the PM_SUSPEND_MAX == 0.
Hmm, the check on `s != NULL' seems also pretty superflous since
s loops over the elements in the array.
> I also thought of checking "( (s) && (s->name !=NULL) )" , but IMHO the
> 'state' check is cleaner (no dereference).
Yup.
Bernd
--
Firmix Software GmbH http://www.firmix.at/
mobil: +43 664 4416156 fax: +43 1 7890849-55
Embedded Linux Development and Services