user/sven/linux.git/lib/sbitmap.c, branch v6.2-rc8

Merge tag 'for-6.2/block-2022-12-08' of git://git.kernel.dk/linux

2022-12-13T18:43:59Z

Pull block updates from Jens Axboe: - NVMe pull requests via Christoph: - Support some passthrough commands without CAP_SYS_ADMIN (Kanchan Joshi) - Refactor PCIe probing and reset (Christoph Hellwig) - Various fabrics authentication fixes and improvements (Sagi Grimberg) - Avoid fallback to sequential scan due to transient issues (Uday Shankar) - Implement support for the DEAC bit in Write Zeroes (Christoph Hellwig) - Allow overriding the IEEE OUI and firmware revision in configfs for nvmet (Aleksandr Miloserdov) - Force reconnect when number of queue changes in nvmet (Daniel Wagner) - Minor fixes and improvements (Uros Bizjak, Joel Granados, Sagi Grimberg, Christoph Hellwig, Christophe JAILLET) - Fix and cleanup nvme-fc req allocation (Chaitanya Kulkarni) - Use the common tagset helpers in nvme-pci driver (Christoph Hellwig) - Cleanup the nvme-pci removal path (Christoph Hellwig) - Use kstrtobool() instead of strtobool (Christophe JAILLET) - Allow unprivileged passthrough of Identify Controller (Joel Granados) - Support io stats on the mpath device (Sagi Grimberg) - Minor nvmet cleanup (Sagi Grimberg) - MD pull requests via Song: - Code cleanups (Christoph) - Various fixes - Floppy pull request from Denis: - Fix a memory leak in the init error path (Yuan) - Series fixing some batch wakeup issues with sbitmap (Gabriel) - Removal of the pktcdvd driver that was deprecated more than 5 years ago, and subsequent removal of the devnode callback in struct block_device_operations as no users are now left (Greg) - Fix for partition read on an exclusively opened bdev (Jan) - Series of elevator API cleanups (Jinlong, Christoph) - Series of fixes and cleanups for blk-iocost (Kemeng) - Series of fixes and cleanups for blk-throttle (Kemeng) - Series adding concurrent support for sync queues in BFQ (Yu) - Series bringing drbd a bit closer to the out-of-tree maintained version (Christian, Joel, Lars, Philipp) - Misc drbd fixes (Wang) - blk-wbt fixes and tweaks for enable/disable (Yu) - Fixes for mq-deadline for zoned devices (Damien) - Add support for read-only and offline zones for null_blk (Shin'ichiro) - Series fixing the delayed holder tracking, as used by DM (Yu, Christoph) - Series enabling bio alloc caching for IRQ based IO (Pavel) - Series enabling userspace peer-to-peer DMA (Logan) - BFQ waker fixes (Khazhismel) - Series fixing elevator refcount issues (Christoph, Jinlong) - Series cleaning up references around queue destruction (Christoph) - Series doing quiesce by tagset, enabling cleanups in drivers (Christoph, Chao) - Series untangling the queue kobject and queue references (Christoph) - Misc fixes and cleanups (Bart, David, Dawei, Jinlong, Kemeng, Ye, Yang, Waiman, Shin'ichiro, Randy, Pankaj, Christoph) * tag 'for-6.2/block-2022-12-08' of git://git.kernel.dk/linux: (247 commits) blktrace: Fix output non-blktrace event when blk_classic option enabled block: sed-opal: Don't include sed-opal: allow using IOC_OPAL_SAVE for locking too blk-cgroup: Fix typo in comment block: remove bio_set_op_attrs nvmet: don't open-code NVME_NS_ATTR_RO enumeration nvme-pci: use the tagset alloc/free helpers nvme: add the Apple shared tag workaround to nvme_alloc_io_tag_set nvme: only set reserved_tags in nvme_alloc_io_tag_set for fabrics controllers nvme: consolidate setting the tagset flags nvme: pass nr_maps explicitly to nvme_alloc_io_tag_set block: bio_copy_data_iter nvme-pci: split out a nvme_pci_ctrl_is_dead helper nvme-pci: return early on ctrl state mismatch in nvme_reset_work nvme-pci: rename nvme_disable_io_queues nvme-pci: cleanup nvme_suspend_queue nvme-pci: remove nvme_pci_disable nvme-pci: remove nvme_disable_admin_queue nvme: merge nvme_shutdown_ctrl into nvme_disable_ctrl nvme: use nvme_wait_ready in nvme_shutdown_ctrl ...

treewide: use get_random_u32_below() instead of deprecated function

2022-11-18T01:15:15Z

This is a simple mechanical transformation done by: @@ expression E; @@ - prandom_u32_max + get_random_u32_below (E) Reviewed-by: Kees Cook Reviewed-by: Greg Kroah-Hartman Acked-by: Darrick J. Wong # for xfs Reviewed-by: SeongJae Park # for damon Reviewed-by: Jason Gunthorpe # for infiniband Reviewed-by: Russell King (Oracle) # for arm Acked-by: Ulf Hansson # for mmc Signed-off-by: Jason A. Donenfeld

sbitmap: Try each queue to wake up at least one waiter

2022-11-16T18:33:03Z

Jan reported the new algorithm as merged might be problematic if the queue being awaken becomes empty between the waitqueue_active inside sbq_wake_ptr check and the wake up. If that happens, wake_up_nr will not wake up any waiter and we loose too many wake ups. In order to guarantee progress, we need to wake up at least one waiter here, if there are any. This now requires trying to wake up from every queue. Instead of walking through all the queues with sbq_wake_ptr, this call moves the wake up inside that function. In a previous version of the patch, I found that updating wake_index several times when walking through queues had a measurable overhead. This ensures we only update it once, at the end. Fixes: 4f8126bb2308 ("sbitmap: Use single per-bitmap counting to wake up queued tags") Reported-by: Jan Kara Signed-off-by: Gabriel Krisman Bertazi Reviewed-by: Jan Kara Link: https://lore.kernel.org/r/20221115224553.23594-4-krisman@suse.de Signed-off-by: Jens Axboe

sbitmap: Advance the queue index before waking up a queue

2022-11-16T18:33:03Z

When a queue is awaken, the wake_index written by sbq_wake_ptr currently keeps pointing to the same queue. On the next wake up, it will thus retry the same queue, which is unfair to other queues, and can lead to starvation. This patch, moves the index update to happen before the queue is returned, such that it will now try a different queue first on the next wake up, improving fairness. Fixes: 4f8126bb2308 ("sbitmap: Use single per-bitmap counting to wake up queued tags") Reported-by: Jan Kara Reviewed-by: Jan Kara Signed-off-by: Gabriel Krisman Bertazi Link: https://lore.kernel.org/r/20221115224553.23594-2-krisman@suse.de Signed-off-by: Jens Axboe

sbitmap: Use single per-bitmap counting to wake up queued tags

2022-11-11T15:38:29Z

sbitmap suffers from code complexity, as demonstrated by recent fixes, and eventual lost wake ups on nested I/O completion. The later happens, from what I understand, due to the non-atomic nature of the updates to wait_cnt, which needs to be subtracted and eventually reset when equal to zero. This two step process can eventually miss an update when a nested completion happens to interrupt the CPU in between the wait_cnt updates. This is very hard to fix, as shown by the recent changes to this code. The code complexity arises mostly from the corner cases to avoid missed wakes in this scenario. In addition, the handling of wake_batch recalculation plus the synchronization with sbq_queue_wake_up is non-trivial. This patchset implements the idea originally proposed by Jan [1], which removes the need for the two-step updates of wait_cnt. This is done by tracking the number of completions and wakeups in always increasing, per-bitmap counters. Instead of having to reset the wait_cnt when it reaches zero, we simply keep counting, and attempt to wake up N threads in a single wait queue whenever there is enough space for a batch. Waking up less than batch_wake shouldn't be a problem, because we haven't changed the conditions for wake up, and the existing batch calculation guarantees at least enough remaining completions to wake up a batch for each queue at any time. Performance-wise, one should expect very similar performance to the original algorithm for the case where there is no queueing. In both the old algorithm and this implementation, the first thing is to check ws_active, which bails out if there is no queueing to be managed. In the new code, we took care to avoid accounting completions and wakeups when there is no queueing, to not pay the cost of atomic operations unnecessarily, since it doesn't skew the numbers. For more interesting cases, where there is queueing, we need to take into account the cross-communication of the atomic operations. I've been benchmarking by running parallel fio jobs against a single hctx nullb in different hardware queue depth scenarios, and verifying both IOPS and queueing. Each experiment was repeated 5 times on a 20-CPU box, with 20 parallel jobs. fio was issuing fixed-size randwrites with qd=64 against nullb, varying only the hardware queue length per test. queue size 2 4 8 16 32 64 6.1-rc2 1681.1K (1.6K) 2633.0K (12.7K) 6940.8K (16.3K) 8172.3K (617.5K) 8391.7K (367.1K) 8606.1K (351.2K) patched 1721.8K (15.1K) 3016.7K (3.8K) 7543.0K (89.4K) 8132.5K (303.4K) 8324.2K (230.6K) 8401.8K (284.7K) The following is a similar experiment, ran against a nullb with a single bitmap shared by 20 hctx spread across 2 NUMA nodes. This has 40 parallel fio jobs operating on the same device queue size 2 4 8 16 32 64 6.1-rc2 1081.0K (2.3K) 957.2K (1.5K) 1699.1K (5.7K) 6178.2K (124.6K) 12227.9K (37.7K) 13286.6K (92.9K) patched 1081.8K (2.8K) 1316.5K (5.4K) 2364.4K (1.8K) 6151.4K (20.0K) 11893.6K (17.5K) 12385.6K (18.4K) It has also survived blktests and a 12h-stress run against nullb. I also ran the code against nvme and a scsi SSD, and I didn't observe performance regression in those. If there are other tests you think I should run, please let me know and I will follow up with results. [1] https://lore.kernel.org/all/aef9de29-e9f5-259a-f8be-12d1b734e72@google.com/ Cc: Hugh Dickins Cc: Keith Busch Cc: Liu Song Suggested-by: Jan Kara Signed-off-by: Gabriel Krisman Bertazi Link: https://lore.kernel.org/r/20221105231055.25953-1-krisman@suse.de Signed-off-by: Jens Axboe

treewide: use prandom_u32_max() when possible, part 2

2022-10-11T23:42:58Z

Rather than incurring a division or requesting too many random bytes for the given range, use the prandom_u32_max() function, which only takes the minimum required bytes from the RNG and avoids divisions. This was done by hand, covering things that coccinelle could not do on its own. Reviewed-by: Greg Kroah-Hartman Reviewed-by: Kees Cook Reviewed-by: Yury Norov Reviewed-by: Jan Kara # for ext2, ext4, and sbitmap Acked-by: Jakub Kicinski Signed-off-by: Jason A. Donenfeld

treewide: use prandom_u32_max() when possible, part 1

2022-10-11T23:42:55Z

Rather than incurring a division or requesting too many random bytes for the given range, use the prandom_u32_max() function, which only takes the minimum required bytes from the RNG and avoids divisions. This was done mechanically with this coccinelle script: @basic@ expression E; type T; identifier get_random_u32 =~ "get_random_int|prandom_u32|get_random_u32"; typedef u64; @@ ( - ((T)get_random_u32() % (E)) + prandom_u32_max(E) | - ((T)get_random_u32() & ((E) - 1)) + prandom_u32_max(E * XXX_MAKE_SURE_E_IS_POW2) | - ((u64)(E) * get_random_u32() >> 32) + prandom_u32_max(E) | - ((T)get_random_u32() & ~PAGE_MASK) + prandom_u32_max(PAGE_SIZE) ) @multi_line@ identifier get_random_u32 =~ "get_random_int|prandom_u32|get_random_u32"; identifier RAND; expression E; @@ - RAND = get_random_u32(); ... when != RAND - RAND %= (E); + RAND = prandom_u32_max(E); // Find a potential literal @literal_mask@ expression LITERAL; type T; identifier get_random_u32 =~ "get_random_int|prandom_u32|get_random_u32"; position p; @@ ((T)get_random_u32()@p & (LITERAL)) // Add one to the literal. @script:python add_one@ literal << literal_mask.LITERAL; RESULT; @@ value = None if literal.startswith('0x'): value = int(literal, 16) elif literal[0] in '123456789': value = int(literal, 10) if value is None: print("I don't know how to handle %s" % (literal)) cocci.include_match(False) elif value == 2**32 - 1 or value == 2**31 - 1 or value == 2**24 - 1 or value == 2**16 - 1 or value == 2**8 - 1: print("Skipping 0x%x for cleanup elsewhere" % (value)) cocci.include_match(False) elif value & (value + 1) != 0: print("Skipping 0x%x because it's not a power of two minus one" % (value)) cocci.include_match(False) elif literal.startswith('0x'): coccinelle.RESULT = cocci.make_expr("0x%x" % (value + 1)) else: coccinelle.RESULT = cocci.make_expr("%d" % (value + 1)) // Replace the literal mask with the calculated result. @plus_one@ expression literal_mask.LITERAL; position literal_mask.p; expression add_one.RESULT; identifier FUNC; @@ - (FUNC()@p & (LITERAL)) + prandom_u32_max(RESULT) @collapse_ret@ type T; identifier VAR; expression E; @@ { - T VAR; - VAR = (E); - return VAR; + return E; } @drop_var@ type T; identifier VAR; @@ { - T VAR; ... when != VAR } Reviewed-by: Greg Kroah-Hartman Reviewed-by: Kees Cook Reviewed-by: Yury Norov Reviewed-by: KP Singh Reviewed-by: Jan Kara # for ext4 and sbitmap Reviewed-by: Christoph Böhmwalder # for drbd Acked-by: Jakub Kicinski Acked-by: Heiko Carstens # for s390 Acked-by: Ulf Hansson # for mmc Acked-by: Darrick J. Wong # for xfs Signed-off-by: Jason A. Donenfeld

sbitmap: fix lockup while swapping

2022-09-29T23:58:17Z

Commit 4acb83417cad ("sbitmap: fix batched wait_cnt accounting") is a big improvement: without it, I had to revert to before commit 040b83fcecfb ("sbitmap: fix possible io hung due to lost wakeup") to avoid the high system time and freezes which that had introduced. Now okay on the NVME laptop, but 4acb83417cad is a disaster for heavy swapping (kernel builds in low memory) on another: soon locking up in sbitmap_queue_wake_up() (into which __sbq_wake_up() is inlined), cycling around with waitqueue_active() but wait_cnt 0 . Here is a backtrace, showing the common pattern of outer sbitmap_queue_wake_up() interrupted before setting wait_cnt 0 back to wake_batch (in some cases other CPUs are idle, in other cases they're spinning for a lock in dd_bio_merge()): sbitmap_queue_wake_up < sbitmap_queue_clear < blk_mq_put_tag < __blk_mq_free_request < blk_mq_free_request < __blk_mq_end_request < scsi_end_request < scsi_io_completion < scsi_finish_command < scsi_complete < blk_complete_reqs < blk_done_softirq < __do_softirq < __irq_exit_rcu < irq_exit_rcu < common_interrupt < asm_common_interrupt < _raw_spin_unlock_irqrestore < __wake_up_common_lock < __wake_up < sbitmap_queue_wake_up < sbitmap_queue_clear < blk_mq_put_tag < __blk_mq_free_request < blk_mq_free_request < dd_bio_merge < blk_mq_sched_bio_merge < blk_mq_attempt_bio_merge < blk_mq_submit_bio < __submit_bio < submit_bio_noacct_nocheck < submit_bio_noacct < submit_bio < __swap_writepage < swap_writepage < pageout < shrink_folio_list < evict_folios < lru_gen_shrink_lruvec < shrink_lruvec < shrink_node < do_try_to_free_pages < try_to_free_pages < __alloc_pages_slowpath < __alloc_pages < folio_alloc < vma_alloc_folio < do_anonymous_page < __handle_mm_fault < handle_mm_fault < do_user_addr_fault < exc_page_fault < asm_exc_page_fault See how the process-context sbitmap_queue_wake_up() has been interrupted, after bringing wait_cnt down to 0 (and in this example, after doing its wakeups), before advancing wake_index and refilling wake_cnt: an interrupt-context sbitmap_queue_wake_up() of the same sbq gets stuck. I have almost no grasp of all the possible sbitmap races, and their consequences: but __sbq_wake_up() can do nothing useful while wait_cnt 0, so it is better if sbq_wake_ptr() skips on to the next ws in that case: which fixes the lockup and shows no adverse consequence for me. The check for wait_cnt being 0 is obviously racy, and ultimately can lead to lost wakeups: for example, when there is only a single waitqueue with waiters. However, lost wakeups are unlikely to matter in these cases, and a proper fix requires redesign (and benchmarking) of the batched wakeup code: so let's plug the hole with this bandaid for now. Signed-off-by: Hugh Dickins Reviewed-by: Jan Kara Reviewed-by: Keith Busch Link: https://lore.kernel.org/r/9c2038a7-cdc5-5ee-854c-fbc6168bf16@google.com Signed-off-by: Jens Axboe

sbitmap: fix batched wait_cnt accounting

2022-09-12T06:10:34Z

Batched completions can clear multiple bits, but we're only decrementing the wait_cnt by one each time. This can cause waiters to never be woken, stalling IO. Use the batched count instead. Link: https://bugzilla.kernel.org/show_bug.cgi?id=215679 Signed-off-by: Keith Busch Link: https://lore.kernel.org/r/20220909184022.1709476-1-kbusch@fb.com Signed-off-by: Jens Axboe

sbitmap: Use atomic_long_try_cmpxchg in __sbitmap_queue_get_batch

2022-09-08T15:22:42Z

Use atomic_long_try_cmpxchg instead of atomic_long_cmpxchg (*ptr, old, new) == old in __sbitmap_queue_get_batch. x86 CMPXCHG instruction returns success in ZF flag, so this change saves a compare after cmpxchg (and related move instruction in front of cmpxchg). Also, atomic_long_cmpxchg implicitly assigns old *ptr value to "old" when cmpxchg fails, enabling further code simplifications, e.g. an extra memory read can be avoided in the loop. No functional change intended. Cc: Jens Axboe Signed-off-by: Uros Bizjak Link: https://lore.kernel.org/r/20220908151200.9993-1-ubizjak@gmail.com Signed-off-by: Jens Axboe