user/sven/linux.git/include/linux/sched, branch v5.9.8

mm, oom_adj: don't loop through tasks in __set_oom_adj when not necessary

2020-10-29T09:11:37Z

[ Upstream commit 67197a4f28d28d0b073ab0427b03cb2ee5382578 ] Currently __set_oom_adj loops through all processes in the system to keep oom_score_adj and oom_score_adj_min in sync between processes sharing their mm. This is done for any task with more that one mm_users, which includes processes with multiple threads (sharing mm and signals). However for such processes the loop is unnecessary because their signal structure is shared as well. Android updates oom_score_adj whenever a tasks changes its role (background/foreground/...) or binds to/unbinds from a service, making it more/less important. Such operation can happen frequently. We noticed that updates to oom_score_adj became more expensive and after further investigation found out that the patch mentioned in "Fixes" introduced a regression. Using Pixel 4 with a typical Android workload, write time to oom_score_adj increased from ~3.57us to ~362us. Moreover this regression linearly depends on the number of multi-threaded processes running on the system. Mark the mm with a new MMF_MULTIPROCESS flag bit when task is created with (CLONE_VM && !CLONE_THREAD && !CLONE_VFORK). Change __set_oom_adj to use MMF_MULTIPROCESS instead of mm_users to decide whether oom_score_adj update should be synchronized between multiple processes. To prevent races between clone() and __set_oom_adj(), when oom_score_adj of the process being cloned might be modified from userspace, we use oom_adj_mutex. Its scope is changed to global. The combination of (CLONE_VM && !CLONE_THREAD) is rarely used except for the case of vfork(). To prevent performance regressions of vfork(), we skip taking oom_adj_mutex and setting MMF_MULTIPROCESS when CLONE_VFORK is specified. Clearing the MMF_MULTIPROCESS flag (when the last process sharing the mm exits) is left out of this patch to keep it simple and because it is believed that this threading model is rare. Should there ever be a need for optimizing that case as well, it can be done by hooking into the exit path, likely following the mm_update_next_owner pattern. With the combination of (CLONE_VM && !CLONE_THREAD && !CLONE_VFORK) being quite rare, the regression is gone after the change is applied. [surenb@google.com: v3] Link: https://lkml.kernel.org/r/20200902012558.2335613-1-surenb@google.com Fixes: 44a70adec910 ("mm, oom_adj: make sure processes sharing mm have same view of oom_score_adj") Reported-by: Tim Murray Suggested-by: Michal Hocko Signed-off-by: Suren Baghdasaryan Signed-off-by: Andrew Morton Acked-by: Christian Brauner Acked-by: Michal Hocko Acked-by: Oleg Nesterov Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: Eugene Syromiatnikov Cc: Christian Kellner Cc: Adrian Reber Cc: Shakeel Butt Cc: Aleksa Sarai Cc: Alexey Dobriyan Cc: "Eric W. Biederman" Cc: Alexey Gladkov Cc: Michel Lespinasse Cc: Daniel Jordan Cc: Andrei Vagin Cc: Bernd Edlinger Cc: John Johansen Cc: Yafang Shao Link: https://lkml.kernel.org/r/20200824153036.3201505-1-surenb@google.com Debugged-by: Minchan Kim Signed-off-by: Linus Torvalds Signed-off-by: Sasha Levin

watch_queue: Limit the number of watches a user can hold

2020-08-17T16:39:18Z

Impose a limit on the number of watches that a user can hold so that they can't use this mechanism to fill up all the available memory. This is done by putting a counter in user_struct that's incremented when a watch is allocated and decreased when it is released. If the number exceeds the RLIMIT_NOFILE limit, the watch is rejected with EAGAIN. This can be tested by the following means: (1) Create a watch queue and attach it to fd 5 in the program given - in this case, bash: keyctl watch_session /tmp/nlog /tmp/gclog 5 bash (2) In the shell, set the maximum number of files to, say, 99: ulimit -n 99 (3) Add 200 keyrings: for ((i=0; i<200; i++)); do keyctl newring a$i @s || break; done (4) Try to watch all of the keyrings: for ((i=0; i<200; i++)); do echo $i; keyctl watch_add 5 %:a$i || break; done This should fail when the number of watches belonging to the user hits 99. (5) Remove all the keyrings and all of those watches should go away: for ((i=0; i<200; i++)); do keyctl unlink %:a$i; done (6) Kill off the watch queue by exiting the shell spawned by watch_session. Fixes: c73be61cede5 ("pipe: Add general notification queue support") Reported-by: Linus Torvalds Signed-off-by: David Howells Reviewed-by: Jarkko Sakkinen Signed-off-by: Linus Torvalds

kernel: add a kernel_wait helper

2020-08-12T17:57:59Z

Add a helper that waits for a pid and stores the status in the passed in kernel pointer. Use it to fix the usage of kernel_wait4 in call_usermodehelper_exec_sync that only happens to work due to the implicit set_fs(KERNEL_DS) for kernel threads. Signed-off-by: Christoph Hellwig Signed-off-by: Andrew Morton Reviewed-by: Andrew Morton Acked-by: "Eric W. Biederman" Cc: Luis Chamberlain Link: http://lkml.kernel.org/r/20200721130449.5008-1-hch@lst.de Signed-off-by: Linus Torvalds

include/linux/sched/mm.h: optimize current_gfp_context()

2020-08-12T17:57:57Z

The current_gfp_context() converts a number of PF_MEMALLOC_* per-process flags into the corresponding GFP_* flags for memory allocation. In that function, current->flags is accessed 3 times. That may lead to duplicated access of the same memory location. This is not usually a problem with minimal debug config options on as the compiler can optimize away the duplicated memory accesses. With most of the debug config options on, however, that may not be the case. For example, the x86-64 object size of the __need_fs_reclaim() in a debug kernel that calls current_gfp_context() was 309 bytes. With this patch applied, the object size is reduced to 202 bytes. This is a saving of 107 bytes and will probably be slightly faster too. Use READ_ONCE() to access current->flags to prevent the compiler from possibly accessing current->flags multiple times. Signed-off-by: Waiman Long Signed-off-by: Andrew Morton Cc: Peter Zijlstra Cc: Ingo Molnar Cc: Mathieu Desnoyers Cc: Michel Lespinasse Link: http://lkml.kernel.org/r/20200618212936.9776-1-longman@redhat.com Signed-off-by: Linus Torvalds

mm/page_alloc: fix memalloc_nocma_{save/restore} APIs

2020-08-07T18:33:29Z

Currently, memalloc_nocma_{save/restore} API that prevents CMA area in page allocation is implemented by using current_gfp_context(). However, there are two problems of this implementation. First, this doesn't work for allocation fastpath. In the fastpath, original gfp_mask is used since current_gfp_context() is introduced in order to control reclaim and it is on slowpath. So, CMA area can be allocated through the allocation fastpath even if memalloc_nocma_{save/restore} APIs are used. Currently, there is just one user for these APIs and it has a fallback method to prevent actual problem. Second, clearing __GFP_MOVABLE in current_gfp_context() has a side effect to exclude the memory on the ZONE_MOVABLE for allocation target. To fix these problems, this patch changes the implementation to exclude CMA area in page allocation. Main point of this change is using the alloc_flags. alloc_flags is mainly used to control allocation so it fits for excluding CMA area in allocation. Fixes: d7fefcc8de91 (mm/cma: add PF flag to force non cma alloc) Signed-off-by: Joonsoo Kim Signed-off-by: Andrew Morton Reviewed-by: Vlastimil Babka Cc: Christoph Hellwig Cc: Roman Gushchin Cc: Mike Kravetz Cc: Naoya Horiguchi Cc: Michal Hocko Cc: "Aneesh Kumar K . V" Link: http://lkml.kernel.org/r/1595468942-29687-1-git-send-email-iamjoonsoo.kim@lge.com Signed-off-by: Linus Torvalds

Merge tag 'fork-v5.9' of git://git.kernel.org/pub/scm/linux/kernel/git/brauner/linux

2020-08-04T21:47:45Z

Pull fork cleanups from Christian Brauner: "This is cleanup series from when we reworked a chunk of the process creation paths in the kernel and switched to struct {kernel_}clone_args. High-level this does two main things: - Remove the double export of both do_fork() and _do_fork() where do_fork() used the incosistent legacy clone calling convention. Now we only export _do_fork() which is based on struct kernel_clone_args. - Remove the copy_thread_tls()/copy_thread() split making the architecture specific HAVE_COYP_THREAD_TLS config option obsolete. This switches all remaining architectures to select HAVE_COPY_THREAD_TLS and thus to the copy_thread_tls() calling convention. The current split makes the process creation codepaths more convoluted than they need to be. Each architecture has their own copy_thread() function unless it selects HAVE_COPY_THREAD_TLS then it has a copy_thread_tls() function. The split is not needed anymore nowadays, all architectures support CLONE_SETTLS but quite a few of them never bothered to select HAVE_COPY_THREAD_TLS and instead simply continued to use copy_thread() and use the old calling convention. Removing this split cleans up the process creation codepaths and paves the way for implementing clone3() on such architectures since it requires the copy_thread_tls() calling convention. After having made each architectures support copy_thread_tls() this series simply renames that function back to copy_thread(). It also switches all architectures that call do_fork() directly over to _do_fork() and the struct kernel_clone_args calling convention. This is a corollary of switching the architectures that did not yet support it over to copy_thread_tls() since do_fork() is conditional on not supporting copy_thread_tls() (Mostly because it lacks a separate argument for tls which is trivial to fix but there's no need for this function to exist.). The do_fork() removal is in itself already useful as it allows to to remove the export of both do_fork() and _do_fork() we currently have in favor of only _do_fork(). This has already been discussed back when we added clone3(). The legacy clone() calling convention is - as is probably well-known - somewhat odd: # # ABI hall of shame # config CLONE_BACKWARDS config CLONE_BACKWARDS2 config CLONE_BACKWARDS3 that is aggravated by the fact that some architectures such as sparc follow the CLONE_BACKWARDSx calling convention but don't really select the corresponding config option since they call do_fork() directly. So do_fork() enforces a somewhat arbitrary calling convention in the first place that doesn't really help the individual architectures that deviate from it. They can thus simply be switched to _do_fork() enforcing a single calling convention. (I really hope that any new architectures will __not__ try to implement their own calling conventions...) Most architectures already have made a similar switch (m68k comes to mind). Overall this removes more code than it adds even with a good portion of added comments. It simplifies a chunk of arch specific assembly either by moving the code into C or by simply rewriting the assembly. Architectures that have been touched in non-trivial ways have all been actually boot and stress tested: sparc and ia64 have been tested with Debian 9 images. They are the two architectures which have been touched the most. All non-trivial changes to architectures have seen acks from the relevant maintainers. nios2 with a custom built buildroot image. h8300 I couldn't get something bootable to test on but the changes have been fairly automatic and I'm sure we'll hear people yell if I broke something there. All other architectures that have been touched in trivial ways have been compile tested for each single patch of the series via git rebase -x "make ..." v5.8-rc2. arm{64} and x86{_64} have been boot tested even though they have just been trivially touched (removal of the HAVE_COPY_THREAD_TLS macro from their Kconfig) because well they are basically "core architectures" and since it is trivial to get your hands on a useable image" * tag 'fork-v5.9' of git://git.kernel.org/pub/scm/linux/kernel/git/brauner/linux: arch: rename copy_thread_tls() back to copy_thread() arch: remove HAVE_COPY_THREAD_TLS unicore: switch to copy_thread_tls() sh: switch to copy_thread_tls() nds32: switch to copy_thread_tls() microblaze: switch to copy_thread_tls() hexagon: switch to copy_thread_tls() c6x: switch to copy_thread_tls() alpha: switch to copy_thread_tls() fork: remove do_fork() h8300: select HAVE_COPY_THREAD_TLS, switch to kernel_clone_args nios2: enable HAVE_COPY_THREAD_TLS, switch to kernel_clone_args ia64: enable HAVE_COPY_THREAD_TLS, switch to kernel_clone_args sparc: unconditionally enable HAVE_COPY_THREAD_TLS sparc: share process creation helpers between sparc and sparc64 sparc64: enable HAVE_COPY_THREAD_TLS fork: fold legacy_clone_args_valid() into _do_fork()

Merge branch 'exec-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace

2020-08-04T21:27:25Z

Pull execve updates from Eric Biederman: "During the development of v5.7 I ran into bugs and quality of implementation issues related to exec that could not be easily fixed because of the way exec is implemented. So I have been diggin into exec and cleaning up what I can. This cycle I have been looking at different ideas and different implementations to see what is possible to improve exec, and cleaning the way exec interfaces with in kernel users. Only cleaning up the interfaces of exec with rest of the kernel has managed to stabalize and make it through review in time for v5.9-rc1 resulting in 2 sets of changes this cycle. - Implement kernel_execve - Make the user mode driver code a better citizen With kernel_execve the code size got a little larger as the copying of parameters from userspace and copying of parameters from userspace is now separate. The good news is kernel threads no longer need to play games with set_fs to use exec. Which when combined with the rest of Christophs set_fs changes should security bugs with set_fs much more difficult" * 'exec-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace: (23 commits) exec: Implement kernel_execve exec: Factor bprm_stack_limits out of prepare_arg_pages exec: Factor bprm_execve out of do_execve_common exec: Move bprm_mm_init into alloc_bprm exec: Move initialization of bprm->filename into alloc_bprm exec: Factor out alloc_bprm exec: Remove unnecessary spaces from binfmts.h umd: Stop using split_argv umd: Remove exit_umh bpfilter: Take advantage of the facilities of struct pid exit: Factor thread_group_exited out of pidfd_poll umd: Track user space drivers with struct pid bpfilter: Move bpfilter_umh back into init data exec: Remove do_execve_file umh: Stop calling do_execve_file umd: Transform fork_usermode_blob into fork_usermode_driver umd: Rename umd_info.cmdline umd_info.driver_name umd: For clarity rename umh_info umd_info umh: Separate the user mode driver and the user mode helper support umh: Remove call_usermodehelper_setup_file. ...

Merge tag 'sched-core-2020-08-03' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

2020-08-03T21:58:38Z

Pull scheduler updates from Ingo Molnar: - Improve uclamp performance by using a static key for the fast path - Add the "sched_util_clamp_min_rt_default" sysctl, to optimize for better power efficiency of RT tasks on battery powered devices. (The default is to maximize performance & reduce RT latencies.) - Improve utime and stime tracking accuracy, which had a fixed boundary of error, which created larger and larger relative errors as the values become larger. This is now replaced with more precise arithmetics, using the new mul_u64_u64_div_u64() helper in math64.h. - Improve the deadline scheduler, such as making it capacity aware - Improve frequency-invariant scheduling - Misc cleanups in energy/power aware scheduling - Add sched_update_nr_running tracepoint to track changes to nr_running - Documentation additions and updates - Misc cleanups and smaller fixes * tag 'sched-core-2020-08-03' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (54 commits) sched/doc: Factorize bits between sched-energy.rst & sched-capacity.rst sched/doc: Document capacity aware scheduling sched: Document arch_scale_*_capacity() arm, arm64: Fix selection of CONFIG_SCHED_THERMAL_PRESSURE Documentation/sysctl: Document uclamp sysctl knobs sched/uclamp: Add a new sysctl to control RT default boost value sched/uclamp: Fix a deadlock when enabling uclamp static key sched: Remove duplicated tick_nohz_full_enabled() check sched: Fix a typo in a comment sched/uclamp: Remove unnecessary mutex_init() arm, arm64: Select CONFIG_SCHED_THERMAL_PRESSURE sched: Cleanup SCHED_THERMAL_PRESSURE kconfig entry arch_topology, sched/core: Cleanup thermal pressure definition trace/events/sched.h: fix duplicated word linux/sched/mm.h: drop duplicated words in comments smp: Fix a potential usage of stale nr_cpus sched/fair: update_pick_idlest() Select group with lowest group_util when idle_cpus are equal sched: nohz: stop passing around unused "ticks" parameter. sched: Better document ttwu() sched: Add a tracepoint to track rq->nr_running ...

sched: Document arch_scale_*_capacity()

2020-08-01T07:19:43Z

Rather that hide their purpose in some dark, damp corner of Documentation/, add some documentation to the default implementations. Signed-off-by: Valentin Schneider Signed-off-by: Ingo Molnar Link: https://lore.kernel.org/r/20200731192016.7484-2-valentin.schneider@arm.com

sched/uclamp: Add a new sysctl to control RT default boost value

2020-07-29T11:51:47Z

RT tasks by default run at the highest capacity/performance level. When uclamp is selected this default behavior is retained by enforcing the requested uclamp.min (p->uclamp_req[UCLAMP_MIN]) of the RT tasks to be uclamp_none(UCLAMP_MAX), which is SCHED_CAPACITY_SCALE; the maximum value. This is also referred to as 'the default boost value of RT tasks'. See commit 1a00d999971c ("sched/uclamp: Set default clamps for RT tasks"). On battery powered devices, it is desired to control this default (currently hardcoded) behavior at runtime to reduce energy consumed by RT tasks. For example, a mobile device manufacturer where big.LITTLE architecture is dominant, the performance of the little cores varies across SoCs, and on high end ones the big cores could be too power hungry. Given the diversity of SoCs, the new knob allows manufactures to tune the best performance/power for RT tasks for the particular hardware they run on. They could opt to further tune the value when the user selects a different power saving mode or when the device is actively charging. The runtime aspect of it further helps in creating a single kernel image that can be run on multiple devices that require different tuning. Keep in mind that a lot of RT tasks in the system are created by the kernel. On Android for instance I can see over 50 RT tasks, only a handful of which created by the Android framework. To control the default behavior globally by system admins and device integrator, introduce the new sysctl_sched_uclamp_util_min_rt_default to change the default boost value of the RT tasks. I anticipate this to be mostly in the form of modifying the init script of a particular device. To avoid polluting the fast path with unnecessary code, the approach taken is to synchronously do the update by traversing all the existing tasks in the system. This could race with a concurrent fork(), which is dealt with by introducing sched_post_fork() function which will ensure the racy fork will get the right update applied. Tested on Juno-r2 in combination with the RT capacity awareness [1]. By default an RT task will go to the highest capacity CPU and run at the maximum frequency, which is particularly energy inefficient on high end mobile devices because the biggest core[s] are 'huge' and power hungry. With this patch the RT task can be controlled to run anywhere by default, and doesn't cause the frequency to be maximum all the time. Yet any task that really needs to be boosted can easily escape this default behavior by modifying its requested uclamp.min value (p->uclamp_req[UCLAMP_MIN]) via sched_setattr() syscall. [1] 804d402fb6f6: ("sched/rt: Make RT capacity-aware") Signed-off-by: Qais Yousef Signed-off-by: Peter Zijlstra (Intel) Link: https://lkml.kernel.org/r/20200716110347.19553-2-qais.yousef@arm.com