kernel.git/mm/workingset.c, branch linux-4.20.y

Merge branch 'xarray' of git://git.infradead.org/users/willy/linux-dax

2018-10-28T18:35:40Z

Pull XArray conversion from Matthew Wilcox: "The XArray provides an improved interface to the radix tree data structure, providing locking as part of the API, specifying GFP flags at allocation time, eliminating preloading, less re-walking the tree, more efficient iterations and not exposing RCU-protected pointers to its users. This patch set 1. Introduces the XArray implementation 2. Converts the pagecache to use it 3. Converts memremap to use it The page cache is the most complex and important user of the radix tree, so converting it was most important. Converting the memremap code removes the only other user of the multiorder code, which allows us to remove the radix tree code that supported it. I have 40+ followup patches to convert many other users of the radix tree over to the XArray, but I'd like to get this part in first. The other conversions haven't been in linux-next and aren't suitable for applying yet, but you can see them in the xarray-conv branch if you're interested" * 'xarray' of git://git.infradead.org/users/willy/linux-dax: (90 commits) radix tree: Remove multiorder support radix tree test: Convert multiorder tests to XArray radix tree tests: Convert item_delete_rcu to XArray radix tree tests: Convert item_kill_tree to XArray radix tree tests: Move item_insert_order radix tree test suite: Remove multiorder benchmarking radix tree test suite: Remove __item_insert memremap: Convert to XArray xarray: Add range store functionality xarray: Move multiorder_check to in-kernel tests xarray: Move multiorder_shrink to kernel tests xarray: Move multiorder account test in-kernel radix tree test suite: Convert iteration test to XArray radix tree test suite: Convert tag_tagged_items to XArray radix tree: Remove radix_tree_clear_tags radix tree: Remove radix_tree_maybe_preload_order radix tree: Remove split/join code radix tree: Remove radix_tree_update_node_t page cache: Finish XArray conversion dax: Convert page fault handlers to XArray ...

mm: zero-seek shrinkers

2018-10-26T23:26:33Z

The page cache and most shrinkable slab caches hold data that has been read from disk, but there are some caches that only cache CPU work, such as the dentry and inode caches of procfs and sysfs, as well as the subset of radix tree nodes that track non-resident page cache. Currently, all these are shrunk at the same rate: using DEFAULT_SEEKS for the shrinker's seeks setting tells the reclaim algorithm that for every two page cache pages scanned it should scan one slab object. This is a bogus setting. A virtual inode that required no IO to create is not twice as valuable as a page cache page; shadow cache entries with eviction distances beyond the size of memory aren't either. In most cases, the behavior in practice is still fine. Such virtual caches don't tend to grow and assert themselves aggressively, and usually get picked up before they cause problems. But there are scenarios where that's not true. Our database workloads suffer from two of those. For one, their file workingset is several times bigger than available memory, which has the kernel aggressively create shadow page cache entries for the non-resident parts of it. The workingset code does tell the VM that most of these are expendable, but the VM ends up balancing them 2:1 to cache pages as per the seeks setting. This is a huge waste of memory. These workloads also deal with tens of thousands of open files and use /proc for introspection, which ends up growing the proc_inode_cache to absurdly large sizes - again at the cost of valuable cache space, which isn't a reasonable trade-off, given that proc inodes can be re-created without involving the disk. This patch implements a "zero-seek" setting for shrinkers that results in a target ratio of 0:1 between their objects and IO-backed caches. This allows such virtual caches to grow when memory is available (they do cache/avoid CPU work after all), but effectively disables them as soon as IO-backed objects are under pressure. It then switches the shrinkers for procfs and sysfs metadata, as well as excess page cache shadow nodes, to the new zero-seek setting. Link: http://lkml.kernel.org/r/20181009184732.762-5-hannes@cmpxchg.org Signed-off-by: Johannes Weiner Reported-by: Domas Mituzas Reviewed-by: Andrew Morton Reviewed-by: Rik van Riel Acked-by: Peter Zijlstra (Intel) Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

mm: workingset: add vmstat counter for shadow nodes

2018-10-26T23:26:33Z

Make it easier to catch bugs in the shadow node shrinker by adding a counter for the shadow nodes in circulation. [akpm@linux-foundation.org: assert that irqs are disabled, for __inc_lruvec_page_state()] [akpm@linux-foundation.org: s/WARN_ON_ONCE/VM_WARN_ON_ONCE/, per Johannes] Link: http://lkml.kernel.org/r/20181009184732.762-4-hannes@cmpxchg.org Signed-off-by: Johannes Weiner Reviewed-by: Andrew Morton Acked-by: Peter Zijlstra (Intel) Cc: Rik van Riel Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

mm: workingset: use cheaper __inc_lruvec_state in irqsafe node reclaim

2018-10-26T23:26:32Z

No need to use the preemption-safe lruvec state function inside the reclaim region that has irqs disabled. Link: http://lkml.kernel.org/r/20181009184732.762-3-hannes@cmpxchg.org Signed-off-by: Johannes Weiner Reviewed-by: Andrew Morton Reviewed-by: Rik van Riel Acked-by: Peter Zijlstra (Intel) Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

mm: workingset: tell cache transitions from workingset thrashing

2018-10-26T23:26:32Z

Refaults happen during transitions between workingsets as well as in-place thrashing. Knowing the difference between the two has a range of applications, including measuring the impact of memory shortage on the system performance, as well as the ability to smarter balance pressure between the filesystem cache and the swap-backed workingset. During workingset transitions, inactive cache refaults and pushes out established active cache. When that active cache isn't stale, however, and also ends up refaulting, that's bonafide thrashing. Introduce a new page flag that tells on eviction whether the page has been active or not in its lifetime. This bit is then stored in the shadow entry, to classify refaults as transitioning or thrashing. How many page->flags does this leave us with on 32-bit? 20 bits are always page flags 21 if you have an MMU 23 with the zone bits for DMA, Normal, HighMem, Movable 29 with the sparsemem section bits 30 if PAE is enabled 31 with this patch. So on 32-bit PAE, that leaves 1 bit for distinguishing two NUMA nodes. If that's not enough, the system can switch to discontigmem and re-gain the 6 or 7 sparsemem section bits. Link: http://lkml.kernel.org/r/20180828172258.3185-3-hannes@cmpxchg.org Signed-off-by: Johannes Weiner Acked-by: Peter Zijlstra (Intel) Tested-by: Daniel Drake Tested-by: Suren Baghdasaryan Cc: Christopher Lameter Cc: Ingo Molnar Cc: Johannes Weiner Cc: Mike Galbraith Cc: Peter Enderborg Cc: Randy Dunlap Cc: Shakeel Butt Cc: Tejun Heo Cc: Vinayak Menon Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

mm: workingset: don't drop refault information prematurely

2018-10-26T23:26:32Z

Patch series "psi: pressure stall information for CPU, memory, and IO", v4. Overview PSI reports the overall wallclock time in which the tasks in a system (or cgroup) wait for (contended) hardware resources. This helps users understand the resource pressure their workloads are under, which allows them to rootcause and fix throughput and latency problems caused by overcommitting, underprovisioning, suboptimal job placement in a grid; as well as anticipate major disruptions like OOM. Real-world applications We're using the data collected by PSI (and its previous incarnation, memdelay) quite extensively at Facebook, and with several success stories. One usecase is avoiding OOM hangs/livelocks. The reason these happen is because the OOM killer is triggered by reclaim not being able to free pages, but with fast flash devices there is *always* some clean and uptodate cache to reclaim; the OOM killer never kicks in, even as tasks spend 90% of the time thrashing the cache pages of their own executables. There is no situation where this ever makes sense in practice. We wrote a <100 line POC python script to monitor memory pressure and kill stuff way before such pathological thrashing leads to full system losses that would require forcible hard resets. We've since extended and deployed this code into other places to guarantee latency and throughput SLAs, since they're usually violated way before the kernel OOM killer would ever kick in. It is available here: https://github.com/facebookincubator/oomd Eventually we probably want to trigger the in-kernel OOM killer based on extreme sustained pressure as well, so that Linux can avoid memory livelocks - which technically aren't deadlocks, but to the user indistinguishable from them - out of the box. We'd continue using OOMD as the first line of defense to ensure workload health and implement complex kill policies that are beyond the scope of the kernel. We also use PSI memory pressure for loadshedding. Our batch job infrastructure used to use heuristics based on various VM stats to anticipate OOM situations, with lackluster success. We switched it to PSI and managed to anticipate and avoid OOM kills and lockups fairly reliably. The reduction of OOM outages in the worker pool raised the pool's aggregate productivity, and we were able to switch that service to smaller machines. Lastly, we use cgroups to isolate a machine's main workload from maintenance crap like package upgrades, logging, configuration, as well as to prevent multiple workloads on a machine from stepping on each others' toes. We were not able to configure this properly without the pressure metrics; we would see latency or bandwidth drops, but it would often be hard to impossible to rootcause it post-mortem. We now log and graph pressure for the containers in our fleet and can trivially link latency spikes and throughput drops to shortages of specific resources after the fact, and fix the job config/scheduling. PSI has also received testing, feedback, and feature requests from Android and EndlessOS for the purpose of low-latency OOM killing, to intervene in pressure situations before the UI starts hanging. How do you use this feature? A kernel with CONFIG_PSI=y will create a /proc/pressure directory with 3 files: cpu, memory, and io. If using cgroup2, cgroups will also have cpu.pressure, memory.pressure and io.pressure files, which simply aggregate task stalls at the cgroup level instead of system-wide. The cpu file contains one line: some avg10=2.04 avg60=0.75 avg300=0.40 total=157656722 The averages give the percentage of walltime in which one or more tasks are delayed on the runqueue while another task has the CPU. They're recent averages over 10s, 1m, 5m windows, so you can tell short term trends from long term ones, similarly to the load average. The total= value gives the absolute stall time in microseconds. This allows detecting latency spikes that might be too short to sway the running averages. It also allows custom time averaging in case the 10s/1m/5m windows aren't adequate for the usecase (or are too coarse with future hardware). What to make of this "some" metric? If CPU utilization is at 100% and CPU pressure is 0, it means the system is perfectly utilized, with one runnable thread per CPU and nobody waiting. At two or more runnable tasks per CPU, the system is 100% overcommitted and the pressure average will indicate as much. From a utilization perspective this is a great state of course: no CPU cycles are being wasted, even when 50% of the threads were to go idle (as most workloads do vary). From the perspective of the individual job it's not great, however, and they would do better with more resources. Depending on what your priority and options are, raised "some" numbers may or may not require action. The memory file contains two lines: some avg10=70.24 avg60=68.52 avg300=69.91 total=3559632828 full avg10=57.59 avg60=58.06 avg300=60.38 total=3300487258 The some line is the same as for cpu, the time in which at least one task is stalled on the resource. In the case of memory, this includes waiting on swap-in, page cache refaults and page reclaim. The full line, however, indicates time in which *nobody* is using the CPU productively due to pressure: all non-idle tasks are waiting for memory in one form or another. Significant time spent in there is a good trigger for killing things, moving jobs to other machines, or dropping incoming requests, since neither the jobs nor the machine overall are making too much headway. The io file is similar to memory. Because the block layer doesn't have a concept of hardware contention right now (how much longer is my IO request taking due to other tasks?), it reports CPU potential lost on all IO delays, not just the potential lost due to competition. FAQ Q: How is PSI's CPU component different from the load average? A: There are several quirks in the load average that make it hard to impossible to tell how overcommitted the CPU really is. 1. The load average is reported as a raw number of active tasks. You need to know how many CPUs there are in the system, how many CPUs the workload is allowed to use, then think about what the proportion between load and the number of CPUs mean for the tasks trying to run. PSI reports the percentage of wallclock time in which tasks are waiting for a CPU to run on. It doesn't matter how many CPUs are present or usable. The number always tells the quality of life of tasks in the system or in a particular cgroup. 2. The shortest averaging window is 1m, which is extremely coarse, and it's sampled in 5s intervals. A *lot* can happen on a CPU in 5 seconds. This *may* be able to identify persistent long-term trends and very clear and obvious overloads, but it's unusable for latency spikes and more subtle overutilization. PSI's shortest window is 10s. It also exports the cumulative stall times (in microseconds) of synchronously recorded events. 3. On Linux, the load average for historical reasons includes all TASK_UNINTERRUPTIBLE tasks. This gives a broader sense of how busy the system is, but on the flipside it doesn't distinguish whether tasks are likely to contend over the CPU or IO - which obviously requires very different interventions from a sys admin or a job scheduler. PSI reports independent metrics for CPU and IO. You can tell which resource is making the tasks wait, but in conjunction still see how overloaded the system is overall. Q: What's the cost / performance impact of this feature? A: PSI's primary cost is in the scheduler, in particular task wakeups and sleeps. I benchmarked this code using Facebook's two most scheduling sensitive workloads: memcache and webserver. They handle a ton of small requests - lots of wakeups and sleeps with little actual work in between - so they tend to be canaries for scheduler regressions. In the tests, the boxes were handling live traffic over the course of several hours. Half the machines, the control, ran with CONFIG_PSI=n. For memcache I used eight machines total. They're 2-socket, 14 core, 56 thread boxes. The test runs for half the test period, flips the test and control kernels on the hardware to rule out HW factors, DC location etc., then runs the other half of the test. For the webservers, I used 32 machines total. They're single socket, 16 core, 32 thread machines. During the memcache test, CPU load was nopsi=78.05% psi=78.98% in the first half and nopsi=77.52% psi=78.25%, so PSI added between 0.7 and 0.9 percentage points to the CPU load, a difference of about 1%. UPDATE: I re-ran this test with the v3 version of this patch set and the CPU utilization was equivalent between test and control. UPDATE: v4 is on par with v3. As far as end-to-end request latency from the client perspective goes, we don't sample those finely enough to capture the requests going to those particular machines during the test, but we know the p50 turnaround time in this workload is 54us, and perf bench sched pipe on those machines show nopsi=5.232666 us/op and psi=5.587347 us/op, so this doesn't add much here either. The profile for the pipe benchmark shows: 0.87% sched-pipe [kernel.vmlinux] [k] psi_group_change 0.83% perf.real [kernel.vmlinux] [k] psi_group_change 0.82% perf.real [kernel.vmlinux] [k] psi_task_change 0.58% sched-pipe [kernel.vmlinux] [k] psi_task_change The webserver load is running inside 4 nested cgroup levels. The CPU load with both nopsi and psi kernels was indistinguishable at 81%. For comparison, we had to disable the cgroup cpu controller on the webservers because it added 4 percentage points to the CPU% during this same exact test. Versions of this accounting code now run on 80% of our fleet. None of our workloads have reported regressions during the rollout. Daniel Drake said: : I just retested the latest version at : http://git.cmpxchg.org/cgit.cgi/linux-psi.git (Linux 4.18) and the results : are great. : : Test setup: : Endless OS : GeminiLake N4200 low end laptop : 2GB RAM : swap (and zram swap) disabled : : Baseline test: open a handful of large-ish apps and several website : tabs in Google Chrome. : : Results: after a couple of minutes, system is excessively thrashing, mouse : cursor can barely be moved, UI is not responding to mouse clicks, so it's : impractical to recover from this situation as an ordinary user : : Add my simple killer: : https://gist.github.com/dsd/a8988bf0b81a6163475988120fe8d9cd : : Results: when the thrashing causes the UI to become sluggish, the killer : steps in and kills something (usually a chrome tab), and the system : remains usable. I repeatedly opened more apps and more websites over a 15 : minute period but I wasn't able to get the system to a point of UI : unresponsiveness. Suren said: : Backported to 4.9 and retested on ARMv8 8 code system running Android. : Signals behave as expected reacting to memory pressure, no jumps in : "total" counters that would indicate an overflow/underflow issues. Nicely : done! This patch (of 9): If we keep just enough refault information to match the *current* page cache during reclaim time, we could lose a lot of events when there is only a temporary spike in non-cache memory consumption that pushes out all the cache. Once cache comes back, we won't see those refaults. They might not be actionable for LRU aging, but we want to know about them for measuring memory pressure. [hannes@cmpxchg.org: switch to NUMA-aware lru and slab counters] Link: http://lkml.kernel.org/r/20181009184732.762-2-hannes@cmpxchg.org Link: http://lkml.kernel.org/r/20180828172258.3185-2-hannes@cmpxchg.org Signed-off-by: Johannes Weiner Acked-by: Peter Zijlstra (Intel) Reviewed-by: Rik van Riel Tested-by: Daniel Drake Tested-by: Suren Baghdasaryan Cc: Ingo Molnar Cc: Tejun Heo Cc: Vinayak Menon Cc: Christopher Lameter Cc: Peter Enderborg Cc: Shakeel Butt Cc: Mike Galbraith Cc: Randy Dunlap Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

mm: Convert workingset to XArray

2018-10-21T14:46:36Z

We construct an XA_STATE and use it to delete the node with xas_store() rather than adding a special function for this unique use case. Includes a test that simulates this usage for the test suite. Signed-off-by: Matthew Wilcox

xarray: Define struct xa_node

2018-10-21T14:45:56Z

This is a direct replacement for struct radix_tree_node. A couple of struct members have changed name, so convert those. Use a #define so that radix tree users continue to work without change. Signed-off-by: Matthew Wilcox Reviewed-by: Josef Bacik

xarray: Replace exceptional entries

2018-09-30T02:47:49Z

Introduce xarray value entries and tagged pointers to replace radix tree exceptional entries. This is a slight change in encoding to allow the use of an extra bit (we can now store BITS_PER_LONG - 1 bits in a value entry). It is also a change in emphasis; exceptional entries are intimidating and different. As the comment explains, you can choose to store values or pointers in the xarray and they are both first-class citizens. Signed-off-by: Matthew Wilcox Reviewed-by: Josef Bacik

mm/list_lru: introduce list_lru_shrink_walk_irq()

2018-08-17T23:20:32Z

Provide list_lru_shrink_walk_irq() and let it behave like list_lru_walk_one() except that it locks the spinlock with spin_lock_irq(). This is used by scan_shadow_nodes() because its lock nests within the i_pages lock which is acquired with IRQ. This change allows to use proper locking promitives instead hand crafted lock_irq_disable() plus spin_lock(). There is no EXPORT_SYMBOL provided because the current user is in-kernel only. Add list_lru_shrink_walk_irq() which acquires the spinlock with the proper locking primitives. Link: http://lkml.kernel.org/r/20180716111921.5365-5-bigeasy@linutronix.de Signed-off-by: Sebastian Andrzej Siewior Reviewed-by: Vladimir Davydov Cc: Thomas Gleixner Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds