Messages in this thread | | | From | "Rafael J. Wysocki" <> | Subject | Re: [PATCH] PM/Hibernate: Thaw kernel threads in hibernation_snapshot() in error/test path | Date | Thu, 2 Feb 2012 21:33:39 +0100 |
| |
On Thursday, February 02, 2012, Srivatsa S. Bhat wrote: > On 02/03/2012 01:48 AM, Rafael J. Wysocki wrote: > > > On Thursday, February 02, 2012, Srivatsa S. Bhat wrote: > >> On 02/03/2012 12:41 AM, Rafael J. Wysocki wrote: > >> > >>> On Thursday, February 02, 2012, Srivatsa S. Bhat wrote: > >>>> In the hibernation call path, the kernel threads are frozen inside > >>>> hibernation_snapshot(). If we happen to encounter an error further down > >>>> the road or if we are exiting early due to a successful freezer test, > >>>> then thaw kernel threads before returning to the caller. > >>>> > >>>> Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> > >>>> --- > >>>> > >>>> kernel/power/hibernate.c | 6 ++++-- > >>>> kernel/power/user.c | 8 ++------ > >>>> 2 files changed, 6 insertions(+), 8 deletions(-) > >>>> > >>>> diff --git a/kernel/power/hibernate.c b/kernel/power/hibernate.c > >>>> index a5d4cf0..c6dee73 100644 > >>>> --- a/kernel/power/hibernate.c > >>>> +++ b/kernel/power/hibernate.c > >>>> @@ -343,13 +343,13 @@ int hibernation_snapshot(int platform_mode) > >>>> * successful freezer test. > >>>> */ > >>>> freezer_test_done = true; > >>>> - goto Cleanup; > >>>> + goto Thaw; > >>>> } > >>>> > >>>> error = dpm_prepare(PMSG_FREEZE); > >>>> if (error) { > >>>> dpm_complete(PMSG_RECOVER); > >>>> - goto Cleanup; > >>>> + goto Thaw; > >>>> } > >>>> > >>>> suspend_console(); > >>>> @@ -385,6 +385,8 @@ int hibernation_snapshot(int platform_mode) > >>>> platform_end(platform_mode); > >>>> return error; > >>>> > >>>> + Thaw: > >>>> + thaw_kernel_threads(); > >>> > >>> Actaully, no. You have to do swsusp_free() before thawing, otherwise > >>> some allocations made by the just thawed kernel threads may fail. > >>> > >> But then what about the case (in the existing code) when > >> freeze_kernel_threads() fails? It would first thaw kernel threads (in > >> fact it used to thaw even userspace tasks earlier!) before calling > >> swsusp_free(). So, the existing code itself seems to be brittle, considering > >> the point you raised. Right? > > > > Well, that's why freeze_kernel_threads() originally hadn't tried to thaw anything > > and started to do that after one of the Tejun's commits (and I forgot about > > this particular issue back then). > > > >>>> Cleanup: > >>>> swsusp_free(); > >>>> goto Close; > >>>> diff --git a/kernel/power/user.c b/kernel/power/user.c > >>>> index 3e10007..7bee91f 100644 > >>>> --- a/kernel/power/user.c > >>>> +++ b/kernel/power/user.c > >>>> @@ -249,16 +249,12 @@ static long snapshot_ioctl(struct file *filp, unsigned int cmd, > >>>> } > >>>> pm_restore_gfp_mask(); > >>>> error = hibernation_snapshot(data->platform_support); > >>>> - if (error) { > >>>> - thaw_kernel_threads(); > >>>> - } else { > >>>> + if (!error) { > >>>> error = put_user(in_suspend, (int __user *)arg); > >>>> if (!error && !freezer_test_done) > >>>> data->ready = 1; > >>>> - if (freezer_test_done) { > >>>> + if (freezer_test_done) > >>>> freezer_test_done = false; > >>>> - thaw_kernel_threads(); > >>>> - } > >>>> } > >>>> break; > >>> > >>> Overall, this seems to be a cleanup, or is it a bug fix? > >>> > >> > >> This was intended as a cleanup only, not a bug fix. But now, (see my point > >> above), I am beginning to feel that the existing code itself is not robust > >> enough... > > > > Well, let's pretend that we think it is safe to thaw kernel threads before > > freeing memory (actually, they are frozen after the preallocation, so it > > really should be OK). > > > > > Yeah, even I had the same reasoning in mind when writing this patch. > > > > So I think I'll apply your patch after all. > > > > > :-)
However, this one should probably be regarded as a regression fix:
http://marc.info/?l=linux-kernel&m=132813572708843&w=4
because thawing all processes before calling swsusp_free() is definitely not a good idea.
Thanks, Rafael
| |