| Message ID | 20220531100007.174649-4-christian.koenig@amd.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [01/13] fs: add OOM badness callback to file_operatrations struct | expand |
On Tue 31-05-22 11:59:57, Christian König wrote: > This gives the OOM killer an additional hint which processes are > referencing shmem files with potentially no other accounting for them. > > Signed-off-by: Christian König <christian.koenig@amd.com> > --- > mm/shmem.c | 6 ++++++ > 1 file changed, 6 insertions(+) > > diff --git a/mm/shmem.c b/mm/shmem.c > index 4b2fea33158e..a4ad92a16968 100644 > --- a/mm/shmem.c > +++ b/mm/shmem.c > @@ -2179,6 +2179,11 @@ unsigned long shmem_get_unmapped_area(struct file *file, > return inflated_addr; > } > > +static long shmem_oom_badness(struct file *file) > +{ > + return i_size_read(file_inode(file)) >> PAGE_SHIFT; > +} This doesn't really represent the in memory size of the file, does it? Also the memcg oom handling could be considerably skewed if the file was shared between more memcgs.
Am 09.06.22 um 11:18 schrieb Michal Hocko: > On Tue 31-05-22 11:59:57, Christian König wrote: >> This gives the OOM killer an additional hint which processes are >> referencing shmem files with potentially no other accounting for them. >> >> Signed-off-by: Christian König <christian.koenig@amd.com> >> --- >> mm/shmem.c | 6 ++++++ >> 1 file changed, 6 insertions(+) >> >> diff --git a/mm/shmem.c b/mm/shmem.c >> index 4b2fea33158e..a4ad92a16968 100644 >> --- a/mm/shmem.c >> +++ b/mm/shmem.c >> @@ -2179,6 +2179,11 @@ unsigned long shmem_get_unmapped_area(struct file *file, >> return inflated_addr; >> } >> >> +static long shmem_oom_badness(struct file *file) >> +{ >> + return i_size_read(file_inode(file)) >> PAGE_SHIFT; >> +} > This doesn't really represent the in memory size of the file, does it? Well the file could be partially or fully swapped out as anonymous memory or the address space only sparse populated, but even then just using the file size as OOM badness sounded like the most straightforward approach to me. What could happen is that the file is also mmaped and we double account. > Also the memcg oom handling could be considerably skewed if the file was > shared between more memcgs. Yes, and that's one of the reasons why I didn't touched the memcg by this and only affected the classic OOM killer. Thanks for the comments, Christian.
On Thu 09-06-22 14:16:56, Christian König wrote: > Am 09.06.22 um 11:18 schrieb Michal Hocko: > > On Tue 31-05-22 11:59:57, Christian König wrote: > > > This gives the OOM killer an additional hint which processes are > > > referencing shmem files with potentially no other accounting for them. > > > > > > Signed-off-by: Christian König <christian.koenig@amd.com> > > > --- > > > mm/shmem.c | 6 ++++++ > > > 1 file changed, 6 insertions(+) > > > > > > diff --git a/mm/shmem.c b/mm/shmem.c > > > index 4b2fea33158e..a4ad92a16968 100644 > > > --- a/mm/shmem.c > > > +++ b/mm/shmem.c > > > @@ -2179,6 +2179,11 @@ unsigned long shmem_get_unmapped_area(struct file *file, > > > return inflated_addr; > > > } > > > +static long shmem_oom_badness(struct file *file) > > > +{ > > > + return i_size_read(file_inode(file)) >> PAGE_SHIFT; > > > +} > > This doesn't really represent the in memory size of the file, does it? > > Well the file could be partially or fully swapped out as anonymous memory or > the address space only sparse populated, but even then just using the file > size as OOM badness sounded like the most straightforward approach to me. It covers hole as well, right? > What could happen is that the file is also mmaped and we double account. > > > Also the memcg oom handling could be considerably skewed if the file was > > shared between more memcgs. > > Yes, and that's one of the reasons why I didn't touched the memcg by this > and only affected the classic OOM killer. oom_badness is for all oom handlers, including memcg. Maybe I have misread an earlier patch but I do not see anything specific to global oom handling.
Am 09.06.22 um 14:57 schrieb Michal Hocko: > On Thu 09-06-22 14:16:56, Christian König wrote: >> Am 09.06.22 um 11:18 schrieb Michal Hocko: >>> On Tue 31-05-22 11:59:57, Christian König wrote: >>>> This gives the OOM killer an additional hint which processes are >>>> referencing shmem files with potentially no other accounting for them. >>>> >>>> Signed-off-by: Christian König <christian.koenig@amd.com> >>>> --- >>>> mm/shmem.c | 6 ++++++ >>>> 1 file changed, 6 insertions(+) >>>> >>>> diff --git a/mm/shmem.c b/mm/shmem.c >>>> index 4b2fea33158e..a4ad92a16968 100644 >>>> --- a/mm/shmem.c >>>> +++ b/mm/shmem.c >>>> @@ -2179,6 +2179,11 @@ unsigned long shmem_get_unmapped_area(struct file *file, >>>> return inflated_addr; >>>> } >>>> +static long shmem_oom_badness(struct file *file) >>>> +{ >>>> + return i_size_read(file_inode(file)) >> PAGE_SHIFT; >>>> +} >>> This doesn't really represent the in memory size of the file, does it? >> Well the file could be partially or fully swapped out as anonymous memory or >> the address space only sparse populated, but even then just using the file >> size as OOM badness sounded like the most straightforward approach to me. > It covers hole as well, right? Yes, exactly. > >> What could happen is that the file is also mmaped and we double account. >> >>> Also the memcg oom handling could be considerably skewed if the file was >>> shared between more memcgs. >> Yes, and that's one of the reasons why I didn't touched the memcg by this >> and only affected the classic OOM killer. > oom_badness is for all oom handlers, including memcg. Maybe I have > misread an earlier patch but I do not see anything specific to global > oom handling. As far as I can see the oom_badness() function is only used in oom_kill.c and in procfs to return the oom score. Did I missed something? Regards, Christian.
On Thu 09-06-22 16:10:33, Christian König wrote: > Am 09.06.22 um 14:57 schrieb Michal Hocko: > > On Thu 09-06-22 14:16:56, Christian König wrote: > > > Am 09.06.22 um 11:18 schrieb Michal Hocko: > > > > On Tue 31-05-22 11:59:57, Christian König wrote: > > > > > This gives the OOM killer an additional hint which processes are > > > > > referencing shmem files with potentially no other accounting for them. > > > > > > > > > > Signed-off-by: Christian König <christian.koenig@amd.com> > > > > > --- > > > > > mm/shmem.c | 6 ++++++ > > > > > 1 file changed, 6 insertions(+) > > > > > > > > > > diff --git a/mm/shmem.c b/mm/shmem.c > > > > > index 4b2fea33158e..a4ad92a16968 100644 > > > > > --- a/mm/shmem.c > > > > > +++ b/mm/shmem.c > > > > > @@ -2179,6 +2179,11 @@ unsigned long shmem_get_unmapped_area(struct file *file, > > > > > return inflated_addr; > > > > > } > > > > > +static long shmem_oom_badness(struct file *file) > > > > > +{ > > > > > + return i_size_read(file_inode(file)) >> PAGE_SHIFT; > > > > > +} > > > > This doesn't really represent the in memory size of the file, does it? > > > Well the file could be partially or fully swapped out as anonymous memory or > > > the address space only sparse populated, but even then just using the file > > > size as OOM badness sounded like the most straightforward approach to me. > > It covers hole as well, right? > > Yes, exactly. So let's say I have a huge sparse shmem file. I will get killed because the oom_badness of such a file would be large as well... > > > What could happen is that the file is also mmaped and we double account. > > > > > > > Also the memcg oom handling could be considerably skewed if the file was > > > > shared between more memcgs. > > > Yes, and that's one of the reasons why I didn't touched the memcg by this > > > and only affected the classic OOM killer. > > oom_badness is for all oom handlers, including memcg. Maybe I have > > misread an earlier patch but I do not see anything specific to global > > oom handling. > > As far as I can see the oom_badness() function is only used in > oom_kill.c and in procfs to return the oom score. Did I missed > something? oom_kill.c implements most of the oom killer functionality. Memcg oom killing is a part of that. Have a look at select_bad_process.
Am 09.06.22 um 16:21 schrieb Michal Hocko: > On Thu 09-06-22 16:10:33, Christian König wrote: >> Am 09.06.22 um 14:57 schrieb Michal Hocko: >>> On Thu 09-06-22 14:16:56, Christian König wrote: >>>> Am 09.06.22 um 11:18 schrieb Michal Hocko: >>>>> On Tue 31-05-22 11:59:57, Christian König wrote: >>>>>> This gives the OOM killer an additional hint which processes are >>>>>> referencing shmem files with potentially no other accounting for them. >>>>>> >>>>>> Signed-off-by: Christian König <christian.koenig@amd.com> >>>>>> --- >>>>>> mm/shmem.c | 6 ++++++ >>>>>> 1 file changed, 6 insertions(+) >>>>>> >>>>>> diff --git a/mm/shmem.c b/mm/shmem.c >>>>>> index 4b2fea33158e..a4ad92a16968 100644 >>>>>> --- a/mm/shmem.c >>>>>> +++ b/mm/shmem.c >>>>>> @@ -2179,6 +2179,11 @@ unsigned long shmem_get_unmapped_area(struct file *file, >>>>>> return inflated_addr; >>>>>> } >>>>>> +static long shmem_oom_badness(struct file *file) >>>>>> +{ >>>>>> + return i_size_read(file_inode(file)) >> PAGE_SHIFT; >>>>>> +} >>>>> This doesn't really represent the in memory size of the file, does it? >>>> Well the file could be partially or fully swapped out as anonymous memory or >>>> the address space only sparse populated, but even then just using the file >>>> size as OOM badness sounded like the most straightforward approach to me. >>> It covers hole as well, right? >> Yes, exactly. > So let's say I have a huge sparse shmem file. I will get killed because > the oom_badness of such a file would be large as well... Yes, correct. But I of hand don't see how we could improve that accounting. >>>> What could happen is that the file is also mmaped and we double account. >>>> >>>>> Also the memcg oom handling could be considerably skewed if the file was >>>>> shared between more memcgs. >>>> Yes, and that's one of the reasons why I didn't touched the memcg by this >>>> and only affected the classic OOM killer. >>> oom_badness is for all oom handlers, including memcg. Maybe I have >>> misread an earlier patch but I do not see anything specific to global >>> oom handling. >> As far as I can see the oom_badness() function is only used in >> oom_kill.c and in procfs to return the oom score. Did I missed >> something? > oom_kill.c implements most of the oom killer functionality. Memcg oom > killing is a part of that. Have a look at select_bad_process. Ah! So mem_cgroup_scan_tasks() calls oom_evaluate_task for each task in the control group. Thanks for pointing that out, that was absolutely not obvious to me. Is that a show stopper? How should we address this? Christian.
On Thu 09-06-22 16:29:46, Christian König wrote:
[...]
> Is that a show stopper? How should we address this?
This is a hard problem to deal with and I am not sure this simple
solution is really a good fit. Not only because of the memcg side of
things. I have my doubts that sparse files handling is ok as well.
I do realize this is a long term problem and there is a demand for some
solution at least. I am not sure how to deal with shared resources
myself. The best approximation I can come up with is to limit the scope
of the damage into a memcg context. One idea I was playing with (but
never convinced myself it is really a worth) is to allow a new mode of
the oom victim selection for the global oom event. It would be an opt in
and the victim would be selected from the biggest leaf memcg (or kill
the whole memcg if it has group_oom configured.
That would address at least some of the accounting issue because charges
are better tracked than per process memory consumption. It is a crude
and ugly hack and it doesn't solve the underlying problem as shared
resources are not guaranteed to be freed when processes die but maybe it
would be just slightly better than the existing scheme which is clearly
lacking behind existing userspace.
Am 2022-06-09 um 10:21 schrieb Michal Hocko: > On Thu 09-06-22 16:10:33, Christian König wrote: >> Am 09.06.22 um 14:57 schrieb Michal Hocko: >>> On Thu 09-06-22 14:16:56, Christian König wrote: >>>> Am 09.06.22 um 11:18 schrieb Michal Hocko: >>>>> On Tue 31-05-22 11:59:57, Christian König wrote: >>>>>> This gives the OOM killer an additional hint which processes are >>>>>> referencing shmem files with potentially no other accounting for them. >>>>>> >>>>>> Signed-off-by: Christian König <christian.koenig@amd.com> >>>>>> --- >>>>>> mm/shmem.c | 6 ++++++ >>>>>> 1 file changed, 6 insertions(+) >>>>>> >>>>>> diff --git a/mm/shmem.c b/mm/shmem.c >>>>>> index 4b2fea33158e..a4ad92a16968 100644 >>>>>> --- a/mm/shmem.c >>>>>> +++ b/mm/shmem.c >>>>>> @@ -2179,6 +2179,11 @@ unsigned long shmem_get_unmapped_area(struct file *file, >>>>>> return inflated_addr; >>>>>> } >>>>>> +static long shmem_oom_badness(struct file *file) >>>>>> +{ >>>>>> + return i_size_read(file_inode(file)) >> PAGE_SHIFT; >>>>>> +} >>>>> This doesn't really represent the in memory size of the file, does it? >>>> Well the file could be partially or fully swapped out as anonymous memory or >>>> the address space only sparse populated, but even then just using the file >>>> size as OOM badness sounded like the most straightforward approach to me. >>> It covers hole as well, right? >> Yes, exactly. > So let's say I have a huge sparse shmem file. I will get killed because > the oom_badness of such a file would be large as well... Would killing processes free shmem files, though? Aren't those persistent anyway? In that case, shmem files should not contribute to oom_badness at all. I guess a special case would be files that were removed from the filesystem but are still open in some processes. Regards, Felix > >>>> What could happen is that the file is also mmaped and we double account. >>>> >>>>> Also the memcg oom handling could be considerably skewed if the file was >>>>> shared between more memcgs. >>>> Yes, and that's one of the reasons why I didn't touched the memcg by this >>>> and only affected the classic OOM killer. >>> oom_badness is for all oom handlers, including memcg. Maybe I have >>> misread an earlier patch but I do not see anything specific to global >>> oom handling. >> As far as I can see the oom_badness() function is only used in >> oom_kill.c and in procfs to return the oom score. Did I missed >> something? > oom_kill.c implements most of the oom killer functionality. Memcg oom > killing is a part of that. Have a look at select_bad_process. >
Am 09.06.22 um 17:19 schrieb Felix Kuehling: > > Am 2022-06-09 um 10:21 schrieb Michal Hocko: >> On Thu 09-06-22 16:10:33, Christian König wrote: >>> Am 09.06.22 um 14:57 schrieb Michal Hocko: >>>> On Thu 09-06-22 14:16:56, Christian König wrote: >>>>> Am 09.06.22 um 11:18 schrieb Michal Hocko: >>>>>> On Tue 31-05-22 11:59:57, Christian König wrote: >>>>>>> This gives the OOM killer an additional hint which processes are >>>>>>> referencing shmem files with potentially no other accounting for >>>>>>> them. >>>>>>> >>>>>>> Signed-off-by: Christian König <christian.koenig@amd.com> >>>>>>> --- >>>>>>> mm/shmem.c | 6 ++++++ >>>>>>> 1 file changed, 6 insertions(+) >>>>>>> >>>>>>> diff --git a/mm/shmem.c b/mm/shmem.c >>>>>>> index 4b2fea33158e..a4ad92a16968 100644 >>>>>>> --- a/mm/shmem.c >>>>>>> +++ b/mm/shmem.c >>>>>>> @@ -2179,6 +2179,11 @@ unsigned long >>>>>>> shmem_get_unmapped_area(struct file *file, >>>>>>> return inflated_addr; >>>>>>> } >>>>>>> +static long shmem_oom_badness(struct file *file) >>>>>>> +{ >>>>>>> + return i_size_read(file_inode(file)) >> PAGE_SHIFT; >>>>>>> +} >>>>>> This doesn't really represent the in memory size of the file, >>>>>> does it? >>>>> Well the file could be partially or fully swapped out as anonymous >>>>> memory or >>>>> the address space only sparse populated, but even then just using >>>>> the file >>>>> size as OOM badness sounded like the most straightforward approach >>>>> to me. >>>> It covers hole as well, right? >>> Yes, exactly. >> So let's say I have a huge sparse shmem file. I will get killed because >> the oom_badness of such a file would be large as well... > > Would killing processes free shmem files, though? Aren't those > persistent anyway? In that case, shmem files should not contribute to > oom_badness at all. At least for the memfd_create() case they do, yes. Those files were never part of any filesystem in the first place, so by killing all the process referencing them you can indeed free the memory locked by them. Regards, Christian. > > I guess a special case would be files that were removed from the > filesystem but are still open in some processes. > > Regards, > Felix > > >> >>>>> What could happen is that the file is also mmaped and we double >>>>> account. >>>>> >>>>>> Also the memcg oom handling could be considerably skewed if the >>>>>> file was >>>>>> shared between more memcgs. >>>>> Yes, and that's one of the reasons why I didn't touched the memcg >>>>> by this >>>>> and only affected the classic OOM killer. >>>> oom_badness is for all oom handlers, including memcg. Maybe I have >>>> misread an earlier patch but I do not see anything specific to global >>>> oom handling. >>> As far as I can see the oom_badness() function is only used in >>> oom_kill.c and in procfs to return the oom score. Did I missed >>> something? >> oom_kill.c implements most of the oom killer functionality. Memcg oom >> killing is a part of that. Have a look at select_bad_process. >>
On Thu 09-06-22 17:22:14, Christian König wrote: [...] > Those files were never part of any filesystem in the first place, so by > killing all the process referencing them you can indeed free the memory > locked by them. Yes, this would require the oom killer to understand that all processes referencing that file are killed. Theoretically possible but I am not sure a feasible solution.
Am 09.06.22 um 17:07 schrieb Michal Hocko: > On Thu 09-06-22 16:29:46, Christian König wrote: > [...] >> Is that a show stopper? How should we address this? > This is a hard problem to deal with and I am not sure this simple > solution is really a good fit. Not only because of the memcg side of > things. I have my doubts that sparse files handling is ok as well. Well I didn't claimed that this would be easy, we juts need to start somewhere. Regarding the sparse file handling, how about using file->f_mapping->nrpages as badness for shmem files? That should give us the real number of pages allocated through this shmem file and gracefully handles sparse files. > I do realize this is a long term problem and there is a demand for some > solution at least. I am not sure how to deal with shared resources > myself. The best approximation I can come up with is to limit the scope > of the damage into a memcg context. One idea I was playing with (but > never convinced myself it is really a worth) is to allow a new mode of > the oom victim selection for the global oom event. It would be an opt in > and the victim would be selected from the biggest leaf memcg (or kill > the whole memcg if it has group_oom configured. > > That would address at least some of the accounting issue because charges > are better tracked than per process memory consumption. It is a crude > and ugly hack and it doesn't solve the underlying problem as shared > resources are not guaranteed to be freed when processes die but maybe it > would be just slightly better than the existing scheme which is clearly > lacking behind existing userspace. Well, what is so bad at the approach of giving each process holding a reference to some shared memory it's equal amount of badness even when the processes belong to different memory control groups? If you really think that this would be a hard problem for upstreaming we could as well keep the behavior for memcg as it is for now. We would just need to adjust the paramters to oom_badness() a bit. Regards, Christian.
On Fri 10-06-22 12:58:53, Christian König wrote: > Am 09.06.22 um 17:07 schrieb Michal Hocko: > > On Thu 09-06-22 16:29:46, Christian König wrote: > > [...] > > > Is that a show stopper? How should we address this? > > This is a hard problem to deal with and I am not sure this simple > > solution is really a good fit. Not only because of the memcg side of > > things. I have my doubts that sparse files handling is ok as well. > > Well I didn't claimed that this would be easy, we juts need to start > somewhere. > > Regarding the sparse file handling, how about using file->f_mapping->nrpages > as badness for shmem files? > > That should give us the real number of pages allocated through this shmem > file and gracefully handles sparse files. Yes, this would be a better approximation. > > I do realize this is a long term problem and there is a demand for some > > solution at least. I am not sure how to deal with shared resources > > myself. The best approximation I can come up with is to limit the scope > > of the damage into a memcg context. One idea I was playing with (but > > never convinced myself it is really a worth) is to allow a new mode of > > the oom victim selection for the global oom event. And just for the clarity. I have mentioned global oom event here but the concept could be extended to per-memcg oom killer as well. > > It would be an opt in > > and the victim would be selected from the biggest leaf memcg (or kill > > the whole memcg if it has group_oom configured. > > > > That would address at least some of the accounting issue because charges > > are better tracked than per process memory consumption. It is a crude > > and ugly hack and it doesn't solve the underlying problem as shared > > resources are not guaranteed to be freed when processes die but maybe it > > would be just slightly better than the existing scheme which is clearly > > lacking behind existing userspace. > > Well, what is so bad at the approach of giving each process holding a > reference to some shared memory it's equal amount of badness even when the > processes belong to different memory control groups? I am not claiming this is wrong per se. It is just an approximation and it can surely be wrong in some cases (e.g. in those workloads where the share memory is mostly owned by one process while the shared content is consumed by many). The primary question is whether it actually helps much or what kind of scenarios it can help with and whether we can actually do better for those. Also do not forget that shared file memory is not the only thing to care about. What about the kernel memory used on behalf of processes? Just consider the above mentioned memcg driven model. It doesn't really require to chase specific files and do some arbitrary math to share the responsibility. It has a clear accounting and responsibility model. It shares the same underlying problem that the oom killing is not resource aware and therefore there is no guarantee that memory really gets freed. But it allows sane configurations where shared resources do not cross memcg boundaries at least. With that in mind and oom_cgroup semantic you can get at least some semi-sane guarantees. Is it pefect? No, by any means. But I would expect it to be more predictable. Maybe we can come up with a saner model, but just going with per file stats sounds like a hard to predict and debug approach to me. OOM killing is a very disruptive operation and having random tasks killed just because they have mapped few pages from a shared resource sounds like a terrible thing to debug and explain to users. > If you really think that this would be a hard problem for upstreaming we > could as well keep the behavior for memcg as it is for now. We would just > need to adjust the paramters to oom_badness() a bit. Say we ignore the memcg side of things for now. How does it help long term? Special casing the global oom is not all that hard but any future change would very likely be disruptive with some semantic implications AFAICS.
Am 10.06.22 um 13:44 schrieb Michal Hocko: > On Fri 10-06-22 12:58:53, Christian König wrote: > [SNIP] >>> I do realize this is a long term problem and there is a demand for some >>> solution at least. I am not sure how to deal with shared resources >>> myself. The best approximation I can come up with is to limit the scope >>> of the damage into a memcg context. One idea I was playing with (but >>> never convinced myself it is really a worth) is to allow a new mode of >>> the oom victim selection for the global oom event. > And just for the clarity. I have mentioned global oom event here but the > concept could be extended to per-memcg oom killer as well. Then what exactly do you mean with "limiting the scope of the damage"? Cause that doesn't make sense without memcg. >>> It would be an opt in >>> and the victim would be selected from the biggest leaf memcg (or kill >>> the whole memcg if it has group_oom configured. >>> >>> That would address at least some of the accounting issue because charges >>> are better tracked than per process memory consumption. It is a crude >>> and ugly hack and it doesn't solve the underlying problem as shared >>> resources are not guaranteed to be freed when processes die but maybe it >>> would be just slightly better than the existing scheme which is clearly >>> lacking behind existing userspace. >> Well, what is so bad at the approach of giving each process holding a >> reference to some shared memory it's equal amount of badness even when the >> processes belong to different memory control groups? > I am not claiming this is wrong per se. It is just an approximation and > it can surely be wrong in some cases (e.g. in those workloads where the > share memory is mostly owned by one process while the shared content is > consumed by many). Yeah, completely agree. Basically we can only do an educated guess. Key point is that we should do the most educated guess we can and not just try to randomly kill something until we hit the right target. That's essentially what's happening today. > The primary question is whether it actually helps much or what kind of > scenarios it can help with and whether we can actually do better for > those. Well, it does help massively with a standard Linux desktop and GPU workloads (e.g. games). See what currently happens is that when games allocate for example textures the memory for that is not accounted against that game. Instead it's usually the display server (X or Wayland) which most of the shared resources accounts to because it needs to compose a desktop from it and usually also mmaps it for fallback CPU operations. So what happens when a games over allocates texture resources is that your whole desktop restarts because the compositor is killed. This obviously also kills the game, but it would be much nice if we would be more selective here. For hardware rendering DMA-buf and GPU drivers are used, but for the software fallback shmem files is what is used under the hood as far as I know. And the underlying problem is the same for both. > Also do not forget that shared file memory is not the only thing > to care about. What about the kernel memory used on behalf of processes? Yeah, I'm aware of that as well. But at least inside the GPU drivers we try to keep that in a reasonable ratio. > Just consider the above mentioned memcg driven model. It doesn't really > require to chase specific files and do some arbitrary math to share the > responsibility. It has a clear accounting and responsibility model. Ok, how does that work then? > It shares the same underlying problem that the oom killing is not > resource aware and therefore there is no guarantee that memory really > gets freed. But it allows sane configurations where shared resources do > not cross memcg boundaries at least. With that in mind and oom_cgroup > semantic you can get at least some semi-sane guarantees. Is it > pefect? No, by any means. But I would expect it to be more predictable. > > Maybe we can come up with a saner model, but just going with per file > stats sounds like a hard to predict and debug approach to me. OOM > killing is a very disruptive operation and having random tasks killed > just because they have mapped few pages from a shared resource sounds > like a terrible thing to debug and explain to users. Well to be honest I think it's much saner than what we do today. As I said you currently can get any Linux system down within seconds and that's basically a perfect deny of service attack. >> If you really think that this would be a hard problem for upstreaming we >> could as well keep the behavior for memcg as it is for now. We would just >> need to adjust the paramters to oom_badness() a bit. > Say we ignore the memcg side of things for now. How does it help long > term? Special casing the global oom is not all that hard but any future > change would very likely be disruptive with some semantic implications > AFAICS. What else can we do? I mean the desktop instability we are facing is really massive. Regards, Christian.
On Fri 10-06-22 14:17:27, Christian König wrote: > Am 10.06.22 um 13:44 schrieb Michal Hocko: > > On Fri 10-06-22 12:58:53, Christian König wrote: > > [SNIP] > > > > I do realize this is a long term problem and there is a demand for some > > > > solution at least. I am not sure how to deal with shared resources > > > > myself. The best approximation I can come up with is to limit the scope > > > > of the damage into a memcg context. One idea I was playing with (but > > > > never convinced myself it is really a worth) is to allow a new mode of > > > > the oom victim selection for the global oom event. > > And just for the clarity. I have mentioned global oom event here but the > > concept could be extended to per-memcg oom killer as well. > > Then what exactly do you mean with "limiting the scope of the damage"? Cause > that doesn't make sense without memcg. What I meant to say is to use the scheme of the damage control not only to the global oom situation (on the global shortage of memory) but also to the memcg oom situation (when the hard limit on a hierarchy is reached). [...] > > The primary question is whether it actually helps much or what kind of > > scenarios it can help with and whether we can actually do better for > > those. > > Well, it does help massively with a standard Linux desktop and GPU workloads > (e.g. games). > > See what currently happens is that when games allocate for example textures > the memory for that is not accounted against that game. Instead it's usually > the display server (X or Wayland) which most of the shared resources > accounts to because it needs to compose a desktop from it and usually also > mmaps it for fallback CPU operations. Let me try to understand some more. So the game (or the entity to be responsible for the resource) doesn't really allocate the memory but it relies on somebody else (from memcg perspective living in a different resource domain - i.e. a different memcg) to do that on its behalf. Correct? If that is the case then that is certainly not fitting into the memcg model then. I am not really sure there is any reasonable model where you cannot really tell who is responsible for the resource. > So what happens when a games over allocates texture resources is that your > whole desktop restarts because the compositor is killed. This obviously also > kills the game, but it would be much nice if we would be more selective > here. > > For hardware rendering DMA-buf and GPU drivers are used, but for the > software fallback shmem files is what is used under the hood as far as I > know. And the underlying problem is the same for both. For shmem files the end user of the buffer can preallocate and so own the buffer and be accounted for it. > > > Also do not forget that shared file memory is not the only thing > > to care about. What about the kernel memory used on behalf of processes? > > Yeah, I'm aware of that as well. But at least inside the GPU drivers we try > to keep that in a reasonable ratio. > > > Just consider the above mentioned memcg driven model. It doesn't really > > require to chase specific files and do some arbitrary math to share the > > responsibility. It has a clear accounting and responsibility model. > > Ok, how does that work then? The memory is accounted to whoever faults that memory in or to the allocating context if that is a kernel memory (in most situations).
Am 10.06.22 um 16:16 schrieb Michal Hocko: > [...] >>> The primary question is whether it actually helps much or what kind of >>> scenarios it can help with and whether we can actually do better for >>> those. >> Well, it does help massively with a standard Linux desktop and GPU workloads >> (e.g. games). >> >> See what currently happens is that when games allocate for example textures >> the memory for that is not accounted against that game. Instead it's usually >> the display server (X or Wayland) which most of the shared resources >> accounts to because it needs to compose a desktop from it and usually also >> mmaps it for fallback CPU operations. > Let me try to understand some more. So the game (or the entity to be > responsible for the resource) doesn't really allocate the memory but it > relies on somebody else (from memcg perspective living in a different > resource domain - i.e. a different memcg) to do that on its behalf. > Correct? If that is the case then that is certainly not fitting into the > memcg model then. More or less: yes, that is one possible use case. But we could leave that one out since it is not the primary use case. What happens more is that 99% of the resources are only allocated per process, but around 1% are shared with somebody else. But see two comments below of a better description of the problem I'm facing. > I am not really sure there is any reasonable model where you cannot > really tell who is responsible for the resource. Well it would be fine with me to leave out those 1% of resources shared with different memcgs. What breaks my neck are those 99% which are allocated by a game and could potentially be shared but are most of the time not. >> So what happens when a games over allocates texture resources is that your >> whole desktop restarts because the compositor is killed. This obviously also >> kills the game, but it would be much nice if we would be more selective >> here. >> >> For hardware rendering DMA-buf and GPU drivers are used, but for the >> software fallback shmem files is what is used under the hood as far as I >> know. And the underlying problem is the same for both. > For shmem files the end user of the buffer can preallocate and so own > the buffer and be accounted for it. The problem is just that it can easily happen that one process is allocating the resource and a different one freeing it. So just imaging the following example: Process opens X window, get reference to the handle of the buffer backing this window for drawing, tells X to close the window again and then a bit later closes the buffer handle. In this example the X server would be charged allocating the buffer and the client (which is most likely in a different memcg group) is charged freeing it. I could of course add something to struct page to track which memcg (or process) it was charged against, but extending struct page is most likely a no-go. Alternative I could try to track the "owner" of a buffer (e.g. a shmem file), but then it can happen that one processes creates the object and another one is writing to it and actually allocating the memory. >>> Also do not forget that shared file memory is not the only thing >>> to care about. What about the kernel memory used on behalf of processes? >> Yeah, I'm aware of that as well. But at least inside the GPU drivers we try >> to keep that in a reasonable ratio. >> >>> Just consider the above mentioned memcg driven model. It doesn't really >>> require to chase specific files and do some arbitrary math to share the >>> responsibility. It has a clear accounting and responsibility model. >> Ok, how does that work then? > The memory is accounted to whoever faults that memory in or to the > allocating context if that is a kernel memory (in most situations). That's what I had in mind as well. Problem with this approach is that file descriptors are currently not informed that they are shared between processes. So to make this work we would need something like attach/detach to process in struct file_operations. And as I noted, this happens rather often. For example a game which renders 120 frames per second needs to transfer 120 buffers per second between client and X. So this is not something which could take a lot of time and the file descriptor tracking structures in the Linux kernel are not made for this either. I think for now I will try something like this specific for DRM drivers. That doesn't solve the shmem file problem, but it at least gives me something at hand for the accelerated Linux desktop case. Regards, Christian.
On Sat 11-06-22 10:06:18, Christian König wrote: > Am 10.06.22 um 16:16 schrieb Michal Hocko: [...] > > > So what happens when a games over allocates texture resources is that your > > > whole desktop restarts because the compositor is killed. This obviously also > > > kills the game, but it would be much nice if we would be more selective > > > here. > > > > > > For hardware rendering DMA-buf and GPU drivers are used, but for the > > > software fallback shmem files is what is used under the hood as far as I > > > know. And the underlying problem is the same for both. > > For shmem files the end user of the buffer can preallocate and so own > > the buffer and be accounted for it. > > The problem is just that it can easily happen that one process is allocating > the resource and a different one freeing it. > > So just imaging the following example: Process opens X window, get reference > to the handle of the buffer backing this window for drawing, tells X to > close the window again and then a bit later closes the buffer handle. > > In this example the X server would be charged allocating the buffer and the > client (which is most likely in a different memcg group) is charged freeing > it. Thanks for the clarification. > I could of course add something to struct page to track which memcg (or > process) it was charged against, but extending struct page is most likely a > no-go. Struct page already maintains is memcg. The one which has charged it and it will stay constatnt throughout of the allocation lifetime (cgroup v1 has a concept of the charge migration but this hasn't been adopted in v2). We have a concept of active_memcg which allows to charge against a different memcg than the allocating context. From your example above I do not think this is really usable for the described usecase as the X is not aware where the request comes from? > Alternative I could try to track the "owner" of a buffer (e.g. a shmem > file), but then it can happen that one processes creates the object and > another one is writing to it and actually allocating the memory. If you can enforce that the owner is really responsible for the allocation then all should be fine. That would require MAP_POPULATE like semantic and I suspect this is not really feasible with the existing userspace. It would be certainly hard to enforce for bad players.
On 2022-06-11 10:06, Christian König wrote: > Am 10.06.22 um 16:16 schrieb Michal Hocko: >> [...] >>>> Just consider the above mentioned memcg driven model. It doesn't really >>>> require to chase specific files and do some arbitrary math to share the >>>> responsibility. It has a clear accounting and responsibility model. >>> Ok, how does that work then? >> The memory is accounted to whoever faults that memory in or to the >> allocating context if that is a kernel memory (in most situations). > > That's what I had in mind as well. Problem with this approach is that file descriptors are currently not informed that they are shared between processes. > > So to make this work we would need something like attach/detach to process in struct file_operations. > > And as I noted, this happens rather often. For example a game which renders 120 frames per second needs to transfer 120 buffers per second between client and X. FWIW, in the steady state, the game will cycle between a small (generally 2-5) set of buffers. The game will not cause new buffers to be exported & imported for every frame. In general, I'd expect dma-buf export & import to happen relatively rarely, e.g. when a window is opened or resized.
Am 13.06.22 um 11:08 schrieb Michel Dänzer: > On 2022-06-11 10:06, Christian König wrote: >> Am 10.06.22 um 16:16 schrieb Michal Hocko: >>> [...] >>>>> Just consider the above mentioned memcg driven model. It doesn't really >>>>> require to chase specific files and do some arbitrary math to share the >>>>> responsibility. It has a clear accounting and responsibility model. >>>> Ok, how does that work then? >>> The memory is accounted to whoever faults that memory in or to the >>> allocating context if that is a kernel memory (in most situations). >> That's what I had in mind as well. Problem with this approach is that file descriptors are currently not informed that they are shared between processes. >> >> So to make this work we would need something like attach/detach to process in struct file_operations. >> >> And as I noted, this happens rather often. For example a game which renders 120 frames per second needs to transfer 120 buffers per second between client and X. > FWIW, in the steady state, the game will cycle between a small (generally 2-5) set of buffers. The game will not cause new buffers to be exported & imported for every frame. > > In general, I'd expect dma-buf export & import to happen relatively rarely, e.g. when a window is opened or resized. Yeah, on a normal Linux desktop. Just unfortunately not on Android :) Anyway even when this only happens on game start we can't go over all the processes/fds and check where a DMA-buf is opened to account this against each process. We would need to add callbacks for this to make it work halve way reliable. Christian.
Am 13.06.22 um 09:45 schrieb Michal Hocko: > On Sat 11-06-22 10:06:18, Christian König wrote: >> Am 10.06.22 um 16:16 schrieb Michal Hocko: > [...] >> I could of course add something to struct page to track which memcg (or >> process) it was charged against, but extending struct page is most likely a >> no-go. > Struct page already maintains is memcg. The one which has charged it and > it will stay constatnt throughout of the allocation lifetime (cgroup v1 > has a concept of the charge migration but this hasn't been adopted in > v2). > > We have a concept of active_memcg which allows to charge against a > different memcg than the allocating context. From your example above I > do not think this is really usable for the described usecase as the X is > not aware where the request comes from? Well X/Wayland is aware, but not the underlying kernel drivers. When X/Wayland would want to forward this information to the kernel we would need to extend the existing UAPI quite a bit. And that of course doesn't help us at all with existing desktops. >> Alternative I could try to track the "owner" of a buffer (e.g. a shmem >> file), but then it can happen that one processes creates the object and >> another one is writing to it and actually allocating the memory. > If you can enforce that the owner is really responsible for the > allocation then all should be fine. That would require MAP_POPULATE like > semantic and I suspect this is not really feasible with the existing > userspace. It would be certainly hard to enforce for bad players. I've tried this today and the result was: "BUG: Bad rss-counter state mm:000000008751d9ff type:MM_FILEPAGES val:-571286". The problem is once more that files are not informed when the process clones. So what happened is that somebody called fork() with an mm_struct I've accounted my pages to. The result is just that we messed up the rss_stats and the the "BUG..." above. The key difference between normal allocated pages and the resources here is just that we are not bound to an mm_struct in any way. I could just potentially add a dummy VMA to the mm_struct, but to be honest I think that this would just be an absolutely hack. So I'm running out of ideas how to fix this, except for adding this per file oom badness like I proposed. Regards, Christian.
On Mon 13-06-22 13:50:28, Christian König wrote: > Am 13.06.22 um 09:45 schrieb Michal Hocko: > > On Sat 11-06-22 10:06:18, Christian König wrote: > > > Am 10.06.22 um 16:16 schrieb Michal Hocko: [...] > > > Alternative I could try to track the "owner" of a buffer (e.g. a shmem > > > file), but then it can happen that one processes creates the object and > > > another one is writing to it and actually allocating the memory. > > If you can enforce that the owner is really responsible for the > > allocation then all should be fine. That would require MAP_POPULATE like > > semantic and I suspect this is not really feasible with the existing > > userspace. It would be certainly hard to enforce for bad players. > > I've tried this today and the result was: "BUG: Bad rss-counter state > mm:000000008751d9ff type:MM_FILEPAGES val:-571286". > > The problem is once more that files are not informed when the process > clones. So what happened is that somebody called fork() with an mm_struct > I've accounted my pages to. The result is just that we messed up the > rss_stats and the the "BUG..." above. > > The key difference between normal allocated pages and the resources here is > just that we are not bound to an mm_struct in any way. It is not really clear to me what exactly you have tried.
Am 13.06.22 um 14:11 schrieb Michal Hocko: > [SNIP] >>>> Alternative I could try to track the "owner" of a buffer (e.g. a shmem >>>> file), but then it can happen that one processes creates the object and >>>> another one is writing to it and actually allocating the memory. >>> If you can enforce that the owner is really responsible for the >>> allocation then all should be fine. That would require MAP_POPULATE like >>> semantic and I suspect this is not really feasible with the existing >>> userspace. It would be certainly hard to enforce for bad players. >> I've tried this today and the result was: "BUG: Bad rss-counter state >> mm:000000008751d9ff type:MM_FILEPAGES val:-571286". >> >> The problem is once more that files are not informed when the process >> clones. So what happened is that somebody called fork() with an mm_struct >> I've accounted my pages to. The result is just that we messed up the >> rss_stats and the the "BUG..." above. >> >> The key difference between normal allocated pages and the resources here is >> just that we are not bound to an mm_struct in any way. > It is not really clear to me what exactly you have tried. I've tried to track the "owner" of a driver connection by keeping a reference to the mm_struct who created this connection inside our file private and then use add_mm_counter() to account all the allocations of the driver to this mm_struct. This works to the extend that now the right process is killed in an OOM situation. The problem with this approach is that the driver is not informed about operations like fork() or clone(), so what happens is that after a fork()/clone() we have an unbalanced rss-counter. Let me maybe get back to the initial question: We have resources which are not related to the virtual address space of a process, how should we tell the OOM killer about them? Thanks for all the input so far, Christian.
On Mon 13-06-22 14:55:54, Christian König wrote: > Am 13.06.22 um 14:11 schrieb Michal Hocko: > > [SNIP] > > > > > Alternative I could try to track the "owner" of a buffer (e.g. a shmem > > > > > file), but then it can happen that one processes creates the object and > > > > > another one is writing to it and actually allocating the memory. > > > > If you can enforce that the owner is really responsible for the > > > > allocation then all should be fine. That would require MAP_POPULATE like > > > > semantic and I suspect this is not really feasible with the existing > > > > userspace. It would be certainly hard to enforce for bad players. > > > I've tried this today and the result was: "BUG: Bad rss-counter state > > > mm:000000008751d9ff type:MM_FILEPAGES val:-571286". > > > > > > The problem is once more that files are not informed when the process > > > clones. So what happened is that somebody called fork() with an mm_struct > > > I've accounted my pages to. The result is just that we messed up the > > > rss_stats and the the "BUG..." above. > > > > > > The key difference between normal allocated pages and the resources here is > > > just that we are not bound to an mm_struct in any way. > > It is not really clear to me what exactly you have tried. > > I've tried to track the "owner" of a driver connection by keeping a > reference to the mm_struct who created this connection inside our file > private and then use add_mm_counter() to account all the allocations of the > driver to this mm_struct. > > This works to the extend that now the right process is killed in an OOM > situation. The problem with this approach is that the driver is not informed > about operations like fork() or clone(), so what happens is that after a > fork()/clone() we have an unbalanced rss-counter. Yes, I do not think you can make per-process accounting without a concept of the per-process ownership. > Let me maybe get back to the initial question: We have resources which are > not related to the virtual address space of a process, how should we tell > the OOM killer about them? I would say memcg, but we have discussed this already... I do not think that exposing a resource (in a form of a counter or something like that) is sufficient. The existing oom killer implementation is hevily process centric (with memcg extension for grouping but not changing the overall design in principle). If you want to make it aware of resources which are not directly accounted to processes then a a new implementation is necessary IMHO. You would need to evaluate those resources and kill all the tasks that can hold on that resource. This is also the reason why I am not really fan of the per file badness because it adds a notion of resource that is not process bound in general so it will add all sorts of weird runtime corner cases which are impossible to anticipate [*]. Maybe that will work in some scenarios but definitely not something to be done by default without users opting into that and being aware of consequences. There have been discussions that the existing oom implementation cannot fit all potential usecases so maybe we need to finally decide to use a plugable, BPFable etc architecture allow implementations that fit specific needs. [*] I know it is not directly related but kinda similar. In the past we used to have heuristics to consider work done as a resource . That is kill younger processes preferably to reduce the damage. This has turned out to have a very unpredictable behavior and many complains by users. Situation has improved when the selection was solely based on rss. This has its own cons of course but at least they are predictable.
Am 13.06.22 um 16:11 schrieb Michal Hocko: > [SNIP] >> Let me maybe get back to the initial question: We have resources which are >> not related to the virtual address space of a process, how should we tell >> the OOM killer about them? > I would say memcg, but we have discussed this already... Well memcg is at least closer to the requirements than the classic mm_struct accounting. It won't work for really shared buffers, but if that's the requirement to find some doable solution for the remaining 99% then I can live with that. > I do not think that exposing a resource (in a form of a counter > or something like that) is sufficient. The existing oom killer > implementation is hevily process centric (with memcg extension for > grouping but not changing the overall design in principle). If you > want to make it aware of resources which are not directly accounted to > processes then a a new implementation is necessary IMHO. You would need > to evaluate those resources and kill all the tasks that can hold on that > resource. Well the OOM killer is process centric because processes are what you can kill. Even the classic mm_struct based accounting includes MM_SHMEMPAGES into the badness. So accounting shared resources as badness to make a decision is nothing new here. The difference is that this time the badness doesn't come from the memory management subsystem, but rather from the I/O subsystem. > This is also the reason why I am not really fan of the per file > badness because it adds a notion of resource that is not process bound > in general so it will add all sorts of weird runtime corner cases which > are impossible to anticipate [*]. Maybe that will work in some scenarios > but definitely not something to be done by default without users opting > into that and being aware of consequences. Would a kernel command line option to control the behavior be helpful here? > There have been discussions that the existing oom implementation cannot > fit all potential usecases so maybe we need to finally decide to use a > plugable, BPFable etc architecture allow implementations that fit > specific needs. Yeah, BPF came to my mind as well. But need to talk with out experts on that topic first. When the OOM killer runs allocating more memory is pretty much a no-go and I'm not sure what the requirements of running a BPF to find the badness are. > [*] I know it is not directly related but kinda similar. In the past > we used to have heuristics to consider work done as a resource . That is > kill younger processes preferably to reduce the damage. This has turned > out to have a very unpredictable behavior and many complains by > users. Situation has improved when the selection was solely based on > rss. This has its own cons of course but at least they are predictable. Good to know, thanks. Regards, Christian.
On Wed 15-06-22 14:35:22, Christian König wrote: [...] > Even the classic mm_struct based accounting includes MM_SHMEMPAGES into the > badness. So accounting shared resources as badness to make a decision is > nothing new here. Yeah, it is nothing really new but it also doesn't mean it is an example worth following as this doesn't really work currently. Also please note that MM_SHMEMPAGES is counting at least something process specific as those pages are mapped in to the process (and with enough of wishful thinking unmapping can drop the last reference and free something up actually) . With generic per-file memory this is even more detached from process. > The difference is that this time the badness doesn't come from the memory > management subsystem, but rather from the I/O subsystem. > > > This is also the reason why I am not really fan of the per file > > badness because it adds a notion of resource that is not process bound > > in general so it will add all sorts of weird runtime corner cases which > > are impossible to anticipate [*]. Maybe that will work in some scenarios > > but definitely not something to be done by default without users opting > > into that and being aware of consequences. > > Would a kernel command line option to control the behavior be helpful here? I am not sure what would be the proper way to control that that would be future extensible. Kernel command line is certainly and option but if we want to extend that to module like or eBPF interface then it wouldn't stand a future test very quickly.
Am 15.06.22 um 15:15 schrieb Michal Hocko: > On Wed 15-06-22 14:35:22, Christian König wrote: > [...] >> Even the classic mm_struct based accounting includes MM_SHMEMPAGES into the >> badness. So accounting shared resources as badness to make a decision is >> nothing new here. > Yeah, it is nothing really new but it also doesn't mean it is an example > worth following as this doesn't really work currently. Also please note > that MM_SHMEMPAGES is counting at least something process specific as > those pages are mapped in to the process (and with enough of wishful > thinking unmapping can drop the last reference and free something up > actually) . With generic per-file memory this is even more detached from > process. But this is exactly the use case here. See I do have the 1% which is shared between processes, but 99% of the allocations only one process has a reference to them. So that wishful thinking that we can drop the last reference when we kill this specific process is perfectly justified. It can be that this doesn't fit all use cases for the shmem file, but it certainly does for DRM and DMA-buf. >> The difference is that this time the badness doesn't come from the memory >> management subsystem, but rather from the I/O subsystem. >> >>> This is also the reason why I am not really fan of the per file >>> badness because it adds a notion of resource that is not process bound >>> in general so it will add all sorts of weird runtime corner cases which >>> are impossible to anticipate [*]. Maybe that will work in some scenarios >>> but definitely not something to be done by default without users opting >>> into that and being aware of consequences. >> Would a kernel command line option to control the behavior be helpful here? > I am not sure what would be the proper way to control that that would be > future extensible. Kernel command line is certainly and option but if we > want to extend that to module like or eBPF interface then it wouldn't > stand a future test very quickly. Well kernel command lines are not really meant to be stable, aren't they? Regards, Christian.
diff --git a/mm/shmem.c b/mm/shmem.c index 4b2fea33158e..a4ad92a16968 100644 --- a/mm/shmem.c +++ b/mm/shmem.c @@ -2179,6 +2179,11 @@ unsigned long shmem_get_unmapped_area(struct file *file, return inflated_addr; } +static long shmem_oom_badness(struct file *file) +{ + return i_size_read(file_inode(file)) >> PAGE_SHIFT; +} + #ifdef CONFIG_NUMA static int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *mpol) { @@ -3780,6 +3785,7 @@ EXPORT_SYMBOL(shmem_aops); static const struct file_operations shmem_file_operations = { .mmap = shmem_mmap, .get_unmapped_area = shmem_get_unmapped_area, + .oom_badness = shmem_oom_badness, #ifdef CONFIG_TMPFS .llseek = shmem_file_llseek, .read_iter = shmem_file_read_iter,
This gives the OOM killer an additional hint which processes are referencing shmem files with potentially no other accounting for them. Signed-off-by: Christian König <christian.koenig@amd.com> --- mm/shmem.c | 6 ++++++ 1 file changed, 6 insertions(+)