- 18 Aug, 2015 33 commits
-
-
Tejun Heo authored
cgroup stats are local to each cgroup and doesn't propagate to ancestors by default. When recursive stats are necessary, the sum is calculated over all the descendants. This initially was for backward compatibility to support both group-local and recursive stats but this mode of operation makes general sense as stat update is much hotter thafn reporting those stats. This however ends up losing recursive stats when a child is removed. To work around this, cfq-iosched adds its stats to its parent cfq_group->dead_stats which is summed up together when calculating recursive stats. It's planned that the core stats will be moved to blkcg_gq, so we want to move the mechanism for keeping track of the stats of dead children from cfq to blkcg core. This patch adds blkg_[rw]stat->aux_cnt which are atomic64_t's keeping track of auxiliary counts which are excluded when reading local counts but included for recursive. blkg_[rw]stat_merge() which were used by cfq to implement dead_stats are replaced by blkg_[rw]stat_add_aux(), and cfq now forwards stats of a dead cgroup to the aux counts of parent->stats instead of separate ->dead_stats. This will also help making blkg_[rw]stats per-cpu. Signed-off-by:
Tejun Heo <tj@kernel.org> Cc: Vivek Goyal <vgoyal@redhat.com> Signed-off-by:
Jens Axboe <axboe@fb.com>
-
Tejun Heo authored
blkg (blkcg_gq) currently is created by blkcg policies invoking blkg_lookup_create() which ends up repeating about the same code in different policies. Theoretically, this can avoid the overhead of looking and/or creating blkg's if blkcg is enabled but no policy is in use; however, the cost of blkg lookup / creation is very low especially if only the root blkcg is in use which is highly likely if no blkcg policy is in active use - it boils down to a single very predictable conditional and surrounding RCU protection. This patch consolidates blkg creation to a new function blkcg_bio_issue_check() which is called during bio issue from generic_make_request_checks(). blkcg_bio_issue_check() is now the only function which tries to create missing blkg's. The subsequent policy and request_list operations just perform blkg_lookup() and if missing falls back to the root. * blk_get_rl() no longer tries to create blkg. It uses blkg_lookup() instead of blkg_lookup_create(). * blk_throtl_bio() is now called from blkcg_bio_issue_check() with rcu read locked and blkg already looked up. Both throtl_lookup_tg() and throtl_lookup_create_tg() are dropped. * cfq is similarly updated. cfq_lookup_create_cfqg() is replaced with cfq_lookup_cfqg()which uses blkg_lookup(). This consolidates blkg handling and avoids unnecessary blkg creation retries under memory pressure. In addition, this provides a common bio entry point into blkcg where things like common accounting can be performed. v2: Build fixes for !CONFIG_CFQ_GROUP_IOSCHED and !CONFIG_BLK_DEV_THROTTLING. Signed-off-by: -
Tejun Heo authored
If a queue is bypassing, all blkcg policies should become noops but blk-throttle wasn't. It only became noop if the queue was dying. While this wouldn't lead to an oops as falling back to the root blkg is safe in this case, this can be a bit surprising - a bypassing queue could still be applying throttle limits. Fix it by removing blk_queue_dying() test in throtl_lookup_create_tg() and testing blk_queue_bypass() in blk_throtl_bio() and bypassing before doing anything else. Signed-off-by:
-
Tejun Heo authored
Currently, both throttle and cfq policies implement their own root blkg (blkcg_gq) lookup fast path. This patch moves root blkg optimization from throtl_lookup_tg() to __blkg_lookup(). cfq-iosched currently doesn't use blkg_lookup() but will be converted and drop the optimization too. Signed-off-by:
-
Tejun Heo authored
blkg_lookup() checks whether the target queue is bypassing and, if not, calls __blkg_lookup() which first checks the lookup hint and then performs radix tree walk. The operations upto hint checking are trivial and there are many users of this function. This patch inlines blkg_lookup() and the fast path part of __blkg_lookup(). The radix tree lookup and hint update are now in blkg_lookup_slowpath(). This will help consolidating blkg handling by easing moving root blkcg short-circuit to inlined lookup fast path. Signed-off-by:
-
Tejun Heo authored
Each active policy has a cpd (blkcg_policy_data) on each blkcg. The cpd's were allocated by blkcg core and each policy could request to allocate extra space at the end by setting blkcg_policy->cpd_size larger than the size of cpd. This is a bit unusual but blkg (blkcg_gq) policy data used to be handled this way too so it made sense to be consistent; however, blkg policy data switched to alloc/free callbacks. This patch makes similar changes to cpd handling. blkcg_policy->cpd_alloc/free_fn() are added to replace ->cpd_size. As cpd allocation is now done from policy side, it can simply allocate a larger area which embeds cpd at the beginning. As ->cpd_alloc_fn() may be able to perform all necessary initializations, this patch makes ->cpd_init_fn() optional. Signed-off-by:
-
Tejun Heo authored
* Rename blkcg->pd[] to blkcg->cpd[] so that cpd is consistently used for blkcg_policy_data. * Make blkcg_policy->cpd_init_fn() take blkcg_policy_data instead of blkcg. This makes it consistent with blkg_policy_data methods and to-be-added cpd alloc/free methods. * blkcg_policy_data->blkcg and cpd_to_blkcg() added so that cpd_init_fn() can determine the associated blkcg from blkcg_policy_data. v2: blkcg_policy_data->blkcg initializations were missing. Added. Signed-off-by:
-
Tejun Heo authored
The newly added ->pd_alloc_fn() and ->pd_free_fn() deal with pd (blkg_policy_data) while the older ones use blkg (blkcg_gq). As using blkg doesn't make sense for ->pd_alloc_fn() and after allocation pd can always be mapped to blkg and given that these are policy-specific methods, it makes sense to converge on pd. This patch makes all methods deal with pd instead of blkg. Most conversions are trivial. In blk-cgroup.c, a couple method invocation sites now test whether pd exists instead of policy state for consistency. This shouldn't cause any behavioral differences. Signed-off-by:
-
Tejun Heo authored
With the recent addition of alloc and free methods, things became messier. This patch reorganizes them according to the followings. * ->pd_alloc_fn() Responsible for allocation and static initializations - the ones which can be done independent of where the pd might be attached. * ->pd_init_fn() Initializations which require the knowledge of where the pd is attached. * ->pd_free_fn() The counter part of pd_alloc_fn(). Static de-init and freeing. This leaves ->pd_exit_fn() without any users. Removed. While at it, collapse an one liner function throtl_pd_exit(), which has only one user, into its user. Signed-off-by:
-
Tejun Heo authored
Because percpu allocator couldn't do non-blocking allocations, blk-throttle was forced to implement an ad-hoc asynchronous allocation mechanism for its percpu stats for cases where blkg's (blkcg_gq's) are allocated from an IO path without sleepable context. Now that percpu allocator can handle gfp_mask and blkg_policy_data alloc / free are handled by policy methods, the ad-hoc asynchronous allocation mechanism can be replaced with direct allocation from tg_stats_alloc_fn(). Rit it out. This ensures that an active throtl_grp always has valid non-NULL ->stats_cpu. Remove checks on it. Signed-off-by:
-
Tejun Heo authored
A blkg (blkcg_gq) represents the relationship between a cgroup and request_queue. Each active policy has a pd (blkg_policy_data) on each blkg. The pd's were allocated by blkcg core and each policy could request to allocate extra space at the end by setting blkcg_policy->pd_size larger than the size of pd. This is a bit unusual but was done this way mostly to simplify error handling and all the existing use cases could be handled this way; however, this is becoming too restrictive now that percpu memory can be allocated without blocking. This introduces two new mandatory blkcg_policy methods - pd_alloc_fn() and pd_free_fn() - which are used to allocate and release pd for a given policy. As pd allocation is now done from policy side, it can simply allocate a larger area which embeds pd at the beginning. This change makes ->pd_size pointless. Removed. Signed-off-by:
-
Tejun Heo authored
blkg_create() allows NULL ->pd_init_fn() but blkcg_activate_policy() doesn't. As both in-kernel policies implement ->pd_init_fn, it currently doesn't break anything. Update blkcg_activate_policy() so that its behavior is consistent with blkg_create(). Signed-off-by:
-
Tejun Heo authored
When a policy gets activated, it needs to allocate and install its policy data on all existing blkg's (blkcg_gq's). Because blkg iteration is protected by a spinlock, it currently counts the total number of blkg's in the system, allocates the matching number of policy data on a list and installs them during a single iteration. This can be simplified by using speculative GFP_NOWAIT allocations while iterating and falling back to a preallocated policy data on failure. If the preallocated one has already been consumed, it releases the lock, preallocate with GFP_KERNEL and then restarts the iteration. This can be a bit more expensive than before but policy activation is a very cold path and shouldn't matter. Signed-off-by:
-
Tejun Heo authored
blkcg_css_alloc() bypasses policy data allocation and blkcg_css_free() bypasses policy data and blkcg freeing for blkcg_root. There's no reason to to treat policy data any differently for blkcg_root. If the root css gets allocated after policies are registered, policy registration path will add policy data; otherwise, the alloc path will. The free path isn't never invoked for root csses. This patch removes the unnecessary special handling of blkcg_root from css_alloc/free paths. Signed-off-by:
-
Tejun Heo authored
When blkcg_init_queue() fails midway after creating a new blkg, it performs kfree() directly; however, this doesn't free the policy data areas. Make it use blkg_free() instead. In turn, blkg_free() is updated to handle root request_list special case. While this fixes a possible memory leak, it's on an unlikely failure path of an already cold path and the size leaked per occurrence is miniscule too. I don't think it needs to be tagged for -stable. Signed-off-by:
-
Tejun Heo authored
Since ec13b1d6 ("blkcg: always create the blkcg_gq for the root blkcg"), a request_list always has its blkg associated. Drop unnecessary rl->blkg NULL test from blk_put_rl(). Signed-off-by:
-
Tejun Heo authored
Up until now, all async IOs were queued to async queues which are shared across the whole request_queue, which means that blkcg resource control is completely void on async IOs including all writeback IOs. It was done this way because writeback didn't support writeback and there was no way of telling which writeback IO belonged to which cgroup; however, writeback recently became cgroup aware and writeback bio's are sent down properly tagged with the blkcg's to charge them against. This patch makes async cfq_queues per-cfq_cgroup instead of per-cfq_data so that each async IO is charged to the blkcg that it was tagged for instead of unconditionally attributing it to root. * cfq_data->async_cfqq and ->async_idle_cfqq are moved to cfq_group and alloc / destroy paths are updated accordingly. * cfq_link_cfqq_cfqg() no longer overrides @cfqg to root for async queues. * check_blkcg_changed() now also invalidates async queues as they no longer stay the same across cgroups. After this patch, cfq's proportional IO control through blkio.weight works correctly when cgroup writeback is in use. Signed-off-by:
Jeff Moyer <jmoyer@redhat.com> Cc: Vivek Goyal <vgoyal@redhat.com> Cc: Arianna Avanzini <avanzini.arianna@gmail.com> Signed-off-by:
-
Tejun Heo authored
cfq_find_alloc_queue() checks whether a queue actually needs to be allocated, which is unnecessary as its sole caller, cfq_get_queue(), only calls it if so. Also, the oom queue fallback logic is scattered between cfq_get_queue() and cfq_find_alloc_queue(). There really isn't much going on in the latter and things can be made simpler by folding it into cfq_get_queue(). This patch collapses cfq_find_alloc_queue() into cfq_get_queue(). The change is fairly straight-forward with one exception - async_cfqq is now initialized to NULL and the "!is_sync" test in the last if conditional is replaced with "async_cfqq" test. This is because gcc (5.1.1) gets confused for some reason and warns that async_cfqq may be used uninitialized otherwise. Oh well, the code isn't necessarily worse this way. This patch doesn't cause any functional difference. v2: Updated to reflect GFP_ATOMIC -> GPF_NOWAIT. Signed-off-by:
-
Tejun Heo authored
This is necessary for making async cfq_cgroups per-cfq_group instead of per-cfq_data. While this change makes cfq_get_queue() perform RCU locking and look up cfq_group even when it reuses async queue, the extra overhead is extremely unlikely to be noticeable given that this is already sitting behind cic->cfqq[] cache and the overall cost of cfq operation. Signed-off-by:
-
Tejun Heo authored
Even when allocations fail, cfq_find_alloc_queue() always returns a valid cfq_queue by falling back to the oom cfq_queue. As such, there isn't much point in taking @gfp_mask and trying "harder" if __GFP_WAIT is set. GFP_NOWAIT allocations don't fail often and even when they do the degraded behavior is acceptable and temporary. After all, the only reason get_request(), which ultimately determines the gfp_mask, cares about __GFP_WAIT is to guarantee request allocation, assuming IO forward progress, for callers which are willing to wait. There's no reason for cfq_find_alloc_queue() to behave differently on __GFP_WAIT when it already has a fallback mechanism. Remove @gfp_mask from cfq_find_alloc_queue() and propagate the changes to its callers. This simplifies the function quite a bit and will help making async queues per-cfq_group. v2: Updated to reflect GFP_ATOMIC -> GPF_NOWAIT. Signed-off-by:
-
Tejun Heo authored
blkcg performs several allocations to track IOs per cgroup and enforce resource control. Most of these allocations are performed lazily on demand in the IO path and thus can't involve reclaim path. Currently, these allocations use GFP_ATOMIC; however, blkcg can gracefully deal with occassional failures of these allocations by punting IOs to the root cgroup and there's no reason to reach into the emergency reserve. This patch replaces GFP_ATOMIC with GFP_NOWAIT for the following allocations. * bdi_writeback_congested and blkcg_gq allocations in blkg_create(). * radix tree node allocations for blkcg->blkg_tree. * cfq_queue allocation on ioprio changes. Signed-off-by:
Vivek Goyal <vgoyal@redhat.com> Signed-off-by:
-
Tejun Heo authored
* Some were accessing cic->cfqq[] directly. Always use cic_to_cfqq() and cic_set_cfqq(). * check_ioprio_changed() doesn't need to verify cfq_get_queue()'s return for NULL. It's always non-NULL. Simplify accordingly. This patch doesn't cause any functional changes. Signed-off-by:
-
Tejun Heo authored
If the cfq_queue cached in cfq_io_cq is the oom one, cfq_set_request() replaces it by invoking cfq_get_queue() again without putting the oom queue leaking the reference it was holding. While oom queues are not released through reference counting, they're still reference counted and this can theoretically lead to the reference count overflowing and incorrectly invoke the usual release path on it. Fix it by making cfq_set_request() put the ref it was holding. Signed-off-by:
-
Tejun Heo authored
Async cfqq's (cfq_queue's) are shared across cfq_data. When cfq_get_queue() obtains a new queue from cfq_find_alloc_queue(), it stashes the pointer in cfq_data and reuses it from then on; however, the function doesn't consider that cfq_find_alloc_queue() may return the oom_cfqq under memory pressure and installs the returned queue unconditionally. If the oom_cfqq is installed as an async cfqq, cfq_set_request() will continue calling cfq_get_queue() hoping to replace it with a proper queue; however, cfq_get_queue() will keep returning the cached queue for the slot - the oom_cfqq. Fix it by skipping caching if the queue is the oom one. Signed-off-by:
-
Tejun Heo authored
cfq_get_queue()'s control flow looks like the following. async_cfqq = NULL; cfqq = NULL; if (!is_sync) { ... async_cfqq = ...; cfqq = *async_cfqq; } if (!cfqq) cfqq = ...; if (!is_sync && !(*async_cfqq)) ...; The only thing the local variable init, the second if, and the async_cfqq test in the third if achieves is to skip cfqq creation and installation if *async_cfqq was already non-NULL. This is needlessly complicated with different tests examining the same condition. Simplify it to the following. if (!is_sync) { ... async_cfqq = ...; cfqq = *async_cfqq; if (cfqq) goto out; } cfqq = ...; if (!is_sync) ...; out: Signed-off-by: -
Tejun Heo authored
The following tracepoints are updated to report the cgroup used during cgroup writeback. * writeback_write_inode[_start] * writeback_queue * writeback_exec * writeback_start * writeback_written * writeback_wait * writeback_nowork * writeback_wake_background * wbc_writepage * writeback_queue_io * bdi_dirty_ratelimit * balance_dirty_pages * writeback_sb_inodes_requeue * writeback_single_inode[_start] Note that writeback_bdi_register is separated out from writeback_class as reporting cgroup doesn't make sense to it. Tracepoints which take bdi are updated to take bdi_writeback instead. Signed-off-by:
Jan Kara <jack@suse.cz> Reviewed-by:
-
Tejun Heo authored
Add a function to determine the path length of a kernfs node. This for now will be used by writeback tracepoint updates. Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by:
-
Tejun Heo authored
Signed-off-by:
-
Tejun Heo authored
wb_writeback_work->single_wait/done are used for the wait mechanism for synchronous wb_work (wb_writeback_work) items which are issued when bdi_split_work_to_wbs() fails to allocate memory for asynchronous wb_work items; however, there's no reason to use a separate wait mechanism for this. bdi_split_work_to_wbs() can simply use on-stack fallback wb_work item and separate wb_completion to wait for it. This patch removes wb_work->single_wait/done and the related code and make bdi_split_work_to_wbs() use on-stack fallback wb_work and wb_completion instead. Signed-off-by:
-
Tejun Heo authored
wb's (bdi_writeback's) are currently keyed by memcg ID; however, in an earlier implementation, wb's were keyed by blkcg ID. bdi_for_each_wb() walks bdi->cgwb_tree in the ascending ID order and allows iterations to start from an arbitrary ID which is used to interrupt and resume iterations. Unfortunately, while changing wb to be keyed by memcg ID instead of blkcg, bdi_for_each_wb() was missed and is still assuming that wb's are keyed by blkcg ID. This doesn't affect iterations which don't get interrupted but bdi_split_work_to_wbs() makes use of iteration resuming on allocation failures and thus may incorrectly skip or repeat wb's. Fix it by changing bdi_for_each_wb() to take memcg IDs instead of blkcg IDs and updating bdi_split_work_to_wbs() accordingly. Signed-off-by:
-
Jens Axboe authored
Merge branch 'for-4.3-unified-base' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup into for-4.3/blkcg
-
Tejun Heo authored
This allows cgroup subsystems to use a different name on the unified hierarchy. cgroup_subsys->name is used on the unified hierarchy, ->legacy_name elsewhere. If ->legacy_name is not explicitly set, it's automatically set to ->name and the userland visible behavior remains unchanged. v2: Make parse_cgroupfs_options() only consider ->legacy_name as mount options are used only on legacy hierarchies. Suggested by Li Zefan. Signed-off-by:Li Zefan <lizefan@huawei.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: cgroups@vger.kernel.org
-
Tejun Heo authored
It doesn't make sense to print subsystems on mount option or /proc/PID/cgroup for the default hierarchy. * cgroup.controllers file at the root of the default hierarchy lists the currently attached controllers. * The default hierarchy is catch-all for unmounted subsystems. * The default hierarchy doesn't accept any mount options. Suppress subsystem printing on mount options and /proc/PID/cgroup for the default hierarchy. Signed-off-by:
-
- 11 Aug, 2015 1 commit
-
-
Tejun Heo authored
It's pretty unusual to have an int as a private data field and it makes it impossible to carray a pointer value through it. Let's make it an unsigned long. AFAICS, this shouldn't break anything. Signed-off-by:
-
- 05 Aug, 2015 1 commit
-
-
Tejun Heo authored
While cgroup subsystems can't be modules, blkcg supports dynamically loadable policies which interact with cgroup core. Export cgrp_dfl_root so that cgroup_on_dfl() can be used in those modules. Signed-off-by:
-
- 04 Aug, 2015 1 commit
-
-
Tejun Heo authored
Traditionally, each cgroup controller implemented whatever interface it wanted leading to interfaces which are widely inconsistent. Examining the requirements of the controllers readily yield that there are only a few control schemes shared among all. Two major controllers already had to implement new interface for the unified hierarchy due to significant structural changes. Let's take the chance to establish common conventions throughout all controllers. This patch defines CGROUP_WEIGHT_MIN/DFL/MAX to be used on all weight based control knobs and documents the conventions that controllers should follow on the unified hierarchy. Except for io.weight knob, all existing unified hierarchy knobs are already compliant. A follow-up patch will update io.weight. v2: Added descriptions of min, low and high knobs. Signed-off-by:
Johannes Weiner <hannes@cmpxchg.org> Cc: Li Zefan <lizefan@huawei.com> Cc: Peter Zijlstra <peterz@infradead.org>
-
- 27 Jul, 2015 1 commit
-
-
Jens Axboe authored
Merge branch 'stable/for-jens-4.2' of git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen into for-linus Konrad writes: "There are three bugs that have been found in the xen-blkfront (and backend). Two of them have the stable tree CC-ed. They have been found where an guest is migrating to a host that is missing 'feature-persistent' support (from one that has it enabled). We end up hitting an BUG() in the driver code."
-
- 26 Jul, 2015 3 commits
-
-
Linus Torvalds authored
-
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tipLinus Torvalds authored
Pull perf fix from Thomas Gleixner: "A single fix for the intel cqm perf facility to prevent IPIs from interrupt context" * 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: perf/x86/intel/cqm: Return cached counter value from IRQ context
-
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tipLinus Torvalds authored
Pull x86 fixes from Thomas Gleixner: "This update contains: - the manual revert of the SYSCALL32 changes which caused a regression - a fix for the MPX vma handling - three fixes for the ioremap 'is ram' checks. - PAT warning fixes - a trivial fix for the size calculation of TLB tracepoints - handle old EFI structures gracefully This also contains a PAT fix from Jan plus a revert thereof. Toshi explained why the code is correct" * 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: x86/mm/pat: Revert 'Adjust default caching mode translation tables' x86/asm/entry/32: Revert 'Do not use R9 in SYSCALL32' commit x86/mm: Fix newly introduced printk format warnings mm: Fix bugs in region_is_ram() x86/mm: Remove region_is_ram() call from ioremap x86/mm: Move warning from __ioremap_check_ram() to the call site x86/mm/pat, drivers/media/ivtv: Move the PAT warning and replace WARN() with pr_warn() x86/mm/pat, drivers/infiniband/ipath: Replace WARN() with pr_warn() x86/mm/pat: Adjust default caching mode translation tables x86/fpu: Disable dependent CPU features on "noxsave" x86/mpx: Do not set ->vm_ops on MPX VMAs x86/mm: Add parenthesis for TLB tracepoint size calculation efi: Handle memory error structures produced based on old versions of standard
-
