| Message ID | 1444093820-6934-1-git-send-email-wonhong.kwon@lge.com (mailing list archive) |
|---|---|
| State | Accepted, archived |
| Delegated to: | Rafael Wysocki |
| Headers | show |
On Tue 2015-10-06 10:10:20, Wonhong Kwon wrote: > pm_init is being invoked by core_initcall and hibernate_image_size_init > calculates preferred image size (image_size) based on total pages > (totalram_pages). This totalram_pages can be modified during various > initcall-s phase and this can cause miscalculated image_size. > > For example, when CMA is being used, init_cma_reserved_pageblock tries > to change the totalram_pages and this job is done during core_initcall. > In order words, the totalram_pages doesn't take CMA reserved pages into > account when image_size is calculated and it can be too small. > > Move pm_init and pm_disk_init to late_initcall_sync so that it happens > after all other initcall-s change the totalram_pages. Hmm, someone else hit same (similar) problem, but had way more complex solution. So it is nice this is easy. But is it enough? Can CMA be modular? Can CMA allocate memory after boot? THanks, Pavel
> On Tue 2015-10-06 10:10:20, Wonhong Kwon wrote: > > pm_init is being invoked by core_initcall and > > hibernate_image_size_init calculates preferred image size (image_size) > > based on total pages (totalram_pages). This totalram_pages can be > > modified during various initcall-s phase and this can cause > miscalculated image_size. > > > > For example, when CMA is being used, init_cma_reserved_pageblock tries > > to change the totalram_pages and this job is done during core_initcall. > > In order words, the totalram_pages doesn't take CMA reserved pages > > into account when image_size is calculated and it can be too small. > > > > Move pm_init and pm_disk_init to late_initcall_sync so that it happens > > after all other initcall-s change the totalram_pages. > > Hmm, someone else hit same (similar) problem, but had way more complex > solution. > > So it is nice this is easy. > > But is it enough? I'm not a MM expert, but AFAIU, lots of things including CMA try to change totalram_pages during and after boot. For instance, memory is being plugged out after boot, __offline_pages invokes adjust_managed_page_count and totalram_pages can be changed. However, image_size can be changed by user via sysfs after boot, and I think user's preferred size should have a high priority than one calculated with totalram_pages. So, my patch was intended to calculate default image_size based on more accurate totalram_pages during boot, not based on real one on any time after boot. > Can CMA be modular? Can CMA allocate memory after boot? AFAIK, CMA cannot intercept any memory after boot. Reserving memory for CMA area should be required on early stage. Thanks, Wonhong > > THanks, > Pavel > > -- > (english) http://www.livejournal.com/~pavelmachek > (cesky, pictures) > http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html -- To unsubscribe from this list: send the line "unsubscribe linux-pm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tuesday, October 06, 2015 10:10:20 AM Wonhong Kwon wrote: > pm_init is being invoked by core_initcall and hibernate_image_size_init > calculates preferred image size (image_size) based on total pages > (totalram_pages). This totalram_pages can be modified during various > initcall-s phase and this can cause miscalculated image_size. > > For example, when CMA is being used, init_cma_reserved_pageblock tries > to change the totalram_pages and this job is done during core_initcall. > In order words, the totalram_pages doesn't take CMA reserved pages into > account when image_size is calculated and it can be too small. > > Move pm_init and pm_disk_init to late_initcall_sync so that it happens > after all other initcall-s change the totalram_pages. > > Reported-by: Sangseok Lee <sangseok.lee@lge.com> > Signed-off-by: Wonhong Kwon <wonhong.kwon@lge.com> Applied, thanks! Rafael -- To unsubscribe from this list: send the line "unsubscribe linux-pm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/kernel/power/hibernate.c b/kernel/power/hibernate.c index 690f78f..c7ff849 100644 --- a/kernel/power/hibernate.c +++ b/kernel/power/hibernate.c @@ -1087,7 +1087,7 @@ static int __init pm_disk_init(void) return sysfs_create_group(power_kobj, &attr_group); } -core_initcall(pm_disk_init); +late_initcall_sync(pm_disk_init); static int __init resume_setup(char *str) diff --git a/kernel/power/main.c b/kernel/power/main.c index 63d395b..c3e223b 100644 --- a/kernel/power/main.c +++ b/kernel/power/main.c @@ -643,4 +643,4 @@ static int __init pm_init(void) return pm_autosleep_init(); } -core_initcall(pm_init); +late_initcall_sync(pm_init);
pm_init is being invoked by core_initcall and hibernate_image_size_init calculates preferred image size (image_size) based on total pages (totalram_pages). This totalram_pages can be modified during various initcall-s phase and this can cause miscalculated image_size. For example, when CMA is being used, init_cma_reserved_pageblock tries to change the totalram_pages and this job is done during core_initcall. In order words, the totalram_pages doesn't take CMA reserved pages into account when image_size is calculated and it can be too small. Move pm_init and pm_disk_init to late_initcall_sync so that it happens after all other initcall-s change the totalram_pages. Reported-by: Sangseok Lee <sangseok.lee@lge.com> Signed-off-by: Wonhong Kwon <wonhong.kwon@lge.com> --- kernel/power/hibernate.c | 2 +- kernel/power/main.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)