summaryrefslogtreecommitdiff
AgeCommit message (Collapse)Author
2020-05-27IB/ipoib: Fix double free of skb in case of multicast traffic in CM modeValentine Fatiev
When connected mode is set, and we have connected and datagram traffic in parallel, ipoib might crash with double free of datagram skb. The current mechanism assumes that the order in the completion queue is the same as the order of sent packets for all QPs. Order is kept only for specific QP, in case of mixed UD and CM traffic we have few QPs (one UD and few CM's) in parallel. The problem: ---------------------------------------------------------- Transmit queue: ----------------- UD skb pointer kept in queue itself, CM skb kept in spearate queue and uses transmit queue as a placeholder to count the number of total transmitted packets. 0 1 2 3 4 5 6 7 8 9 10 11 12 13 .........127 ------------------------------------------------------------ NL ud1 UD2 CM1 ud3 cm2 cm3 ud4 cm4 ud5 NL NL NL ........... ------------------------------------------------------------ ^ ^ tail head Completion queue (problematic scenario) - the order not the same as in the transmit queue: 1 2 3 4 5 6 7 8 9 ------------------------------------ ud1 CM1 UD2 ud3 cm2 cm3 ud4 cm4 ud5 ------------------------------------ 1. CM1 'wc' processing - skb freed in cm separate ring. - tx_tail of transmit queue increased although UD2 is not freed. Now driver assumes UD2 index is already freed and it could be used for new transmitted skb. 0 1 2 3 4 5 6 7 8 9 10 11 12 13 .........127 ------------------------------------------------------------ NL NL UD2 CM1 ud3 cm2 cm3 ud4 cm4 ud5 NL NL NL ........... ------------------------------------------------------------ ^ ^ ^ (Bad)tail head (Bad - Could be used for new SKB) In this case (due to heavy load) UD2 skb pointer could be replaced by new transmitted packet UD_NEW, as the driver assumes its free. At this point we will have to process two 'wc' with same index but we have only one pointer to free. During second attempt to free the same skb we will have NULL pointer exception. 2. UD2 'wc' processing - skb freed according the index we got from 'wc', but it was already overwritten by mistake. So actually the skb that was released is the skb of the new transmitted packet and not the original one. 3. UD_NEW 'wc' processing - attempt to free already freed skb. NUll pointer exception. The fix: ----------------------------------------------------------------------- The fix is to stop using the UD ring as a placeholder for CM packets, the cyclic ring variables tx_head and tx_tail will manage the UD tx_ring, a new cyclic variables global_tx_head and global_tx_tail are introduced for managing and counting the overall outstanding sent packets, then the send queue will be stopped and waken based on these variables only. Note that no locking is needed since global_tx_head is updated in the xmit flow and global_tx_tail is updated in the NAPI flow only. A previous attempt tried to use one variable to count the outstanding sent packets, but it did not work since xmit and NAPI flows can run at the same time and the counter will be updated wrongly. Thus, we use the same simple cyclic head and tail scheme that we have today for the UD tx_ring. Fixes: 2c104ea68350 ("IB/ipoib: Get rid of the tx_outstanding variable in all modes") Link: https://lore.kernel.org/r/20200527134705.480068-1-leon@kernel.org Signed-off-by: Valentine Fatiev <valentinef@mellanox.com> Signed-off-by: Alaa Hleihel <alaa@mellanox.com> Signed-off-by: Leon Romanovsky <leonro@mellanox.com> Acked-by: Doug Ledford <dledford@redhat.com> Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
2020-05-27drm/amd/display: Fix potential integer wraparound resulting in a hangAric Cyr
[Why] If VUPDATE_END is before VUPDATE_START the delay calculated can become very large, causing a soft hang. [How] Take the absolute value of the difference between START and END. Signed-off-by: Aric Cyr <aric.cyr@amd.com> Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com> Acked-by: Qingqing Zhuo <qingqing.zhuo@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2020-05-27drm/amd/display: drop cursor position check in atomic testSimon Ser
get_cursor_position already handles the case where the cursor has negative off-screen coordinates by not setting dc_cursor_position.enabled. Signed-off-by: Simon Ser <contact@emersion.fr> Fixes: 626bf90fe03f ("drm/amd/display: add basic atomic check for cursor plane") Cc: Alex Deucher <alexander.deucher@amd.com> Cc: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2020-05-27drm/amd/display: Fix potential integer wraparound resulting in a hangAric Cyr
[Why] If VUPDATE_END is before VUPDATE_START the delay calculated can become very large, causing a soft hang. [How] Take the absolute value of the difference between START and END. Signed-off-by: Aric Cyr <aric.cyr@amd.com> Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com> Acked-by: Qingqing Zhuo <qingqing.zhuo@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Cc: stable@vger.kernel.org
2020-05-27drm/amd/display: drop cursor position check in atomic testSimon Ser
get_cursor_position already handles the case where the cursor has negative off-screen coordinates by not setting dc_cursor_position.enabled. Signed-off-by: Simon Ser <contact@emersion.fr> Fixes: 626bf90fe03f ("drm/amd/display: add basic atomic check for cursor plane") Cc: Alex Deucher <alexander.deucher@amd.com> Cc: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Cc: stable@vger.kernel.org
2020-05-27net: dsa: declare lockless TX feature for slave portsVladimir Oltean
Be there a platform with the following layout: Regular NIC | +----> DSA master for switch port | +----> DSA master for another switch port After changing DSA back to static lockdep class keys in commit 1a33e10e4a95 ("net: partially revert dynamic lockdep key changes"), this kernel splat can be seen: [ 13.361198] ============================================ [ 13.366524] WARNING: possible recursive locking detected [ 13.371851] 5.7.0-rc4-02121-gc32a05ecd7af-dirty #988 Not tainted [ 13.377874] -------------------------------------------- [ 13.383201] swapper/0/0 is trying to acquire lock: [ 13.388004] ffff0000668ff298 (&dsa_slave_netdev_xmit_lock_key){+.-.}-{2:2}, at: __dev_queue_xmit+0x84c/0xbe0 [ 13.397879] [ 13.397879] but task is already holding lock: [ 13.403727] ffff0000661a1698 (&dsa_slave_netdev_xmit_lock_key){+.-.}-{2:2}, at: __dev_queue_xmit+0x84c/0xbe0 [ 13.413593] [ 13.413593] other info that might help us debug this: [ 13.420140] Possible unsafe locking scenario: [ 13.420140] [ 13.426075] CPU0 [ 13.428523] ---- [ 13.430969] lock(&dsa_slave_netdev_xmit_lock_key); [ 13.435946] lock(&dsa_slave_netdev_xmit_lock_key); [ 13.440924] [ 13.440924] *** DEADLOCK *** [ 13.440924] [ 13.446860] May be due to missing lock nesting notation [ 13.446860] [ 13.453668] 6 locks held by swapper/0/0: [ 13.457598] #0: ffff800010003de0 ((&idev->mc_ifc_timer)){+.-.}-{0:0}, at: call_timer_fn+0x0/0x400 [ 13.466593] #1: ffffd4d3fb478700 (rcu_read_lock){....}-{1:2}, at: mld_sendpack+0x0/0x560 [ 13.474803] #2: ffffd4d3fb478728 (rcu_read_lock_bh){....}-{1:2}, at: ip6_finish_output2+0x64/0xb10 [ 13.483886] #3: ffffd4d3fb478728 (rcu_read_lock_bh){....}-{1:2}, at: __dev_queue_xmit+0x6c/0xbe0 [ 13.492793] #4: ffff0000661a1698 (&dsa_slave_netdev_xmit_lock_key){+.-.}-{2:2}, at: __dev_queue_xmit+0x84c/0xbe0 [ 13.503094] #5: ffffd4d3fb478728 (rcu_read_lock_bh){....}-{1:2}, at: __dev_queue_xmit+0x6c/0xbe0 [ 13.512000] [ 13.512000] stack backtrace: [ 13.516369] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.7.0-rc4-02121-gc32a05ecd7af-dirty #988 [ 13.530421] Call trace: [ 13.532871] dump_backtrace+0x0/0x1d8 [ 13.536539] show_stack+0x24/0x30 [ 13.539862] dump_stack+0xe8/0x150 [ 13.543271] __lock_acquire+0x1030/0x1678 [ 13.547290] lock_acquire+0xf8/0x458 [ 13.550873] _raw_spin_lock+0x44/0x58 [ 13.554543] __dev_queue_xmit+0x84c/0xbe0 [ 13.558562] dev_queue_xmit+0x24/0x30 [ 13.562232] dsa_slave_xmit+0xe0/0x128 [ 13.565988] dev_hard_start_xmit+0xf4/0x448 [ 13.570182] __dev_queue_xmit+0x808/0xbe0 [ 13.574200] dev_queue_xmit+0x24/0x30 [ 13.577869] neigh_resolve_output+0x15c/0x220 [ 13.582237] ip6_finish_output2+0x244/0xb10 [ 13.586430] __ip6_finish_output+0x1dc/0x298 [ 13.590709] ip6_output+0x84/0x358 [ 13.594116] mld_sendpack+0x2bc/0x560 [ 13.597786] mld_ifc_timer_expire+0x210/0x390 [ 13.602153] call_timer_fn+0xcc/0x400 [ 13.605822] run_timer_softirq+0x588/0x6e0 [ 13.609927] __do_softirq+0x118/0x590 [ 13.613597] irq_exit+0x13c/0x148 [ 13.616918] __handle_domain_irq+0x6c/0xc0 [ 13.621023] gic_handle_irq+0x6c/0x160 [ 13.624779] el1_irq+0xbc/0x180 [ 13.627927] cpuidle_enter_state+0xb4/0x4d0 [ 13.632120] cpuidle_enter+0x3c/0x50 [ 13.635703] call_cpuidle+0x44/0x78 [ 13.639199] do_idle+0x228/0x2c8 [ 13.642433] cpu_startup_entry+0x2c/0x48 [ 13.646363] rest_init+0x1ac/0x280 [ 13.649773] arch_call_rest_init+0x14/0x1c [ 13.653878] start_kernel+0x490/0x4bc Lockdep keys themselves were added in commit ab92d68fc22f ("net: core: add generic lockdep keys"), and it's very likely that this splat existed since then, but I have no real way to check, since this stacked platform wasn't supported by mainline back then. >From Taehee's own words: This patch was considered that all stackable devices have LLTX flag. But the dsa doesn't have LLTX, so this splat happened. After this patch, dsa shares the same lockdep class key. On the nested dsa interface architecture, which you illustrated, the same lockdep class key will be used in __dev_queue_xmit() because dsa doesn't have LLTX. So that lockdep detects deadlock because the same lockdep class key is used recursively although actually the different locks are used. There are some ways to fix this problem. 1. using NETIF_F_LLTX flag. If possible, using the LLTX flag is a very clear way for it. But I'm so sorry I don't know whether the dsa could have LLTX or not. 2. using dynamic lockdep again. It means that each interface uses a separate lockdep class key. So, lockdep will not detect recursive locking. But this way has a problem that it could consume lockdep class key too many. Currently, lockdep can have 8192 lockdep class keys. - you can see this number with the following command. cat /proc/lockdep_stats lock-classes: 1251 [max: 8192] ... The [max: 8192] means that the maximum number of lockdep class keys. If too many lockdep class keys are registered, lockdep stops to work. So, using a dynamic(separated) lockdep class key should be considered carefully. In addition, updating lockdep class key routine might have to be existing. (lockdep_register_key(), lockdep_set_class(), lockdep_unregister_key()) 3. Using lockdep subclass. A lockdep class key could have 8 subclasses. The different subclass is considered different locks by lockdep infrastructure. But "lock-classes" is not counted by subclasses. So, it could avoid stopping lockdep infrastructure by an overflow of lockdep class keys. This approach should also have an updating lockdep class key routine. (lockdep_set_subclass()) 4. Using nonvalidate lockdep class key. The lockdep infrastructure supports nonvalidate lockdep class key type. It means this lockdep is not validated by lockdep infrastructure. So, the splat will not happen but lockdep couldn't detect real deadlock case because lockdep really doesn't validate it. I think this should be used for really special cases. (lockdep_set_novalidate_class()) Further discussion here: https://patchwork.ozlabs.org/project/netdev/patch/20200503052220.4536-2-xiyou.wangcong@gmail.com/ There appears to be no negative side-effect to declaring lockless TX for the DSA virtual interfaces, which means they handle their own locking. So that's what we do to make the splat go away. Patch tested in a wide variety of cases: unicast, multicast, PTP, etc. Fixes: ab92d68fc22f ("net: core: add generic lockdep keys") Suggested-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Reviewed-by: Florian Fainelli <f.fainelli@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-27copy_xstate_to_kernel(): don't leave parts of destination uninitializedAl Viro
copy the corresponding pieces of init_fpstate into the gaps instead. Cc: stable@kernel.org Tested-by: Alexander Potapenko <glider@google.com> Acked-by: Borislav Petkov <bp@suse.de> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2020-05-27net: dsa: felix: send VLANs on CPU port as egress-taggedVladimir Oltean
As explained in other commits before (b9cd75e66895 and 87b0f983f66f), ocelot switches have a single egress-untagged VLAN per port, and the driver would deny adding a second one while an egress-untagged VLAN already exists. But on the CPU port (where the VLAN configuration is implicit, because there is no net device for the bridge to control), the DSA core attempts to add a VLAN using the same flags as were used for the front-panel port. This would make adding any untagged VLAN fail due to the CPU port rejecting the configuration: bridge vlan add dev swp0 vid 100 pvid untagged [ 1865.854253] mscc_felix 0000:00:00.5: Port already has a native VLAN: 1 [ 1865.860824] mscc_felix 0000:00:00.5: Failed to add VLAN 100 to port 5: -16 (note that port 5 is the CPU port and not the front-panel swp0). So this hardware will send all VLANs as tagged towards the CPU. Fixes: 56051948773e ("net: dsa: ocelot: add driver for Felix switch family") Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Reviewed-by: Florian Fainelli <f.fainelli@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-27bridge: multicast: work around clang bugArnd Bergmann
Clang-10 and clang-11 run into a corner case of the register allocator on 32-bit ARM, leading to excessive stack usage from register spilling: net/bridge/br_multicast.c:2422:6: error: stack frame size of 1472 bytes in function 'br_multicast_get_stats' [-Werror,-Wframe-larger-than=] Work around this by marking one of the internal functions as noinline_for_stack. Link: https://bugs.llvm.org/show_bug.cgi?id=45802#c9 Signed-off-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-27nvme-pci: avoid race between nvme_reap_pending_cqes() and nvme_poll()Dongli Zhang
There may be a race between nvme_reap_pending_cqes() and nvme_poll(), e.g., when doing live reset while polling the nvme device. CPU X CPU Y nvme_poll() nvme_dev_disable() -> nvme_stop_queues() -> nvme_suspend_io_queues() -> nvme_suspend_queue() -> spin_lock(&nvmeq->cq_poll_lock); -> nvme_reap_pending_cqes() -> nvme_process_cq() -> nvme_process_cq() In the above scenario, the nvme_process_cq() for the same queue may be running on both CPU X and CPU Y concurrently. It is much more easier to reproduce the issue when CONFIG_PREEMPT is enabled in kernel. When CONFIG_PREEMPT is disabled, it would take longer time for nvme_stop_queues()-->blk_mq_quiesce_queue() to wait for grace period. This patch protects nvme_process_cq() with nvmeq->cq_poll_lock in nvme_reap_pending_cqes(). Fixes: fa46c6fb5d61 ("nvme/pci: move cqe check after device shutdown") Signed-off-by: Dongli Zhang <dongli.zhang@oracle.com> Reviewed-by: Ming Lei <ming.lei@redhat.com> Reviewed-by: Keith Busch <kbusch@kernel.org> Signed-off-by: Christoph Hellwig <hch@lst.de>
2020-05-27vsock: fix timeout in vsock_accept()Stefano Garzarella
The accept(2) is an "input" socket interface, so we should use SO_RCVTIMEO instead of SO_SNDTIMEO to set the timeout. So this patch replace sock_sndtimeo() with sock_rcvtimeo() to use the right timeout in the vsock_accept(). Fixes: d021c344051a ("VSOCK: Introduce VM Sockets") Signed-off-by: Stefano Garzarella <sgarzare@redhat.com> Reviewed-by: Jorgen Hansen <jhansen@vmware.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-27nfp: flower: fix used time of merge flow statisticsHeinrich Kuhn
Prior to this change the correct value for the used counter is calculated but not stored nor, therefore, propagated to user-space. In use-cases such as OVS use-case at least this results in active flows being removed from the hardware datapath. Which results in both unnecessary flow tear-down and setup, and packet processing on the host. This patch addresses the problem by saving the calculated used value which allows the value to propagate to user-space. Found by inspection. Fixes: aa6ce2ea0c93 ("nfp: flower: support stats update for merge flows") Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com> Signed-off-by: Simon Horman <simon.horman@netronome.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-27net/sched: fix infinite loop in sch_fq_pieDavide Caratti
this command hangs forever: # tc qdisc add dev eth0 root fq_pie flows 65536 watchdog: BUG: soft lockup - CPU#1 stuck for 23s! [tc:1028] [...] CPU: 1 PID: 1028 Comm: tc Not tainted 5.7.0-rc6+ #167 RIP: 0010:fq_pie_init+0x60e/0x8b7 [sch_fq_pie] Code: 4c 89 65 50 48 89 f8 48 c1 e8 03 42 80 3c 30 00 0f 85 2a 02 00 00 48 8d 7d 10 4c 89 65 58 48 89 f8 48 c1 e8 03 42 80 3c 30 00 <0f> 85 a7 01 00 00 48 8d 7d 18 48 c7 45 10 46 c3 23 00 48 89 f8 48 RSP: 0018:ffff888138d67468 EFLAGS: 00000246 ORIG_RAX: ffffffffffffff13 RAX: 1ffff9200018d2b2 RBX: ffff888139c1c400 RCX: ffffffffffffffff RDX: 000000000000c5e8 RSI: ffffc900000e5000 RDI: ffffc90000c69590 RBP: ffffc90000c69580 R08: fffffbfff79a9699 R09: fffffbfff79a9699 R10: 0000000000000700 R11: fffffbfff79a9698 R12: ffffc90000c695d0 R13: 0000000000000000 R14: dffffc0000000000 R15: 000000002347c5e8 FS: 00007f01e1850e40(0000) GS:ffff88814c880000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 000000000067c340 CR3: 000000013864c000 CR4: 0000000000340ee0 Call Trace: qdisc_create+0x3fd/0xeb0 tc_modify_qdisc+0x3be/0x14a0 rtnetlink_rcv_msg+0x5f3/0x920 netlink_rcv_skb+0x121/0x350 netlink_unicast+0x439/0x630 netlink_sendmsg+0x714/0xbf0 sock_sendmsg+0xe2/0x110 ____sys_sendmsg+0x5b4/0x890 ___sys_sendmsg+0xe9/0x160 __sys_sendmsg+0xd3/0x170 do_syscall_64+0x9a/0x370 entry_SYSCALL_64_after_hwframe+0x44/0xa9 we can't accept 65536 as a valid number for 'nflows', because the loop on 'idx' in fq_pie_init() will never end. The extack message is correct, but it doesn't say that 0 is not a valid number for 'flows': while at it, fix this also. Add a tdc selftest to check correct validation of 'flows'. CC: Ivan Vecera <ivecera@redhat.com> Fixes: ec97ecf1ebe4 ("net: sched: add Flow Queue PIE packet scheduler") Signed-off-by: Davide Caratti <dcaratti@redhat.com> Reviewed-by: Ivan Vecera <ivecera@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-27Merge tag 'fsnotify_for_v5.7-rc8' of ↵Linus Torvalds
git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs Pull fanotify FAN_DIR_MODIFY disabling from Jan Kara: "A single patch that disables FAN_DIR_MODIFY support that was merged in this merge window. When discussing further functionality we realized it may be more logical to guard it with a feature flag or to call things slightly differently (or maybe not) so let's not set the API in stone for now." * tag 'fsnotify_for_v5.7-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs: fanotify: turn off support for FAN_DIR_MODIFY
2020-05-27Merge branch 'for-5.7-fixes' of ↵Linus Torvalds
git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup Pull cgroup fixes from Tejun Heo: - Reverted stricter synchronization for cgroup recursive stats which was prepping it for event counter usage which never got merged. The change was causing performation regressions in some cases. - Restore bpf-based device-cgroup operation even when cgroup1 device cgroup is disabled. - An out-param init fix. * 'for-5.7-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup: device_cgroup: Cleanup cgroup eBPF device filter code xattr: fix uninitialized out-param Revert "cgroup: Add memory barriers to plug cgroup_rstat_updated() race window"
2020-05-27RDMA/core: Fix double destruction of uobjectJason Gunthorpe
Fix use after free when user user space request uobject concurrently for the same object, within the RCU grace period. In that case, remove_handle_idr_uobject() is called twice and we will have an extra put on the uobject which cause use after free. Fix it by leaving the uobject write locked after it was removed from the idr. Call to rdma_lookup_put_uobject with UVERBS_LOOKUP_DESTROY instead of UVERBS_LOOKUP_WRITE will do the work. refcount_t: underflow; use-after-free. WARNING: CPU: 0 PID: 1381 at lib/refcount.c:28 refcount_warn_saturate+0xfe/0x1a0 Kernel panic - not syncing: panic_on_warn set ... CPU: 0 PID: 1381 Comm: syz-executor.0 Not tainted 5.5.0-rc3 #8 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.12.1-0-ga5cab58e9a3f-prebuilt.qemu.org 04/01/2014 Call Trace: dump_stack+0x94/0xce panic+0x234/0x56f __warn+0x1cc/0x1e1 report_bug+0x200/0x310 fixup_bug.part.11+0x32/0x80 do_error_trap+0xd3/0x100 do_invalid_op+0x31/0x40 invalid_op+0x1e/0x30 RIP: 0010:refcount_warn_saturate+0xfe/0x1a0 Code: 0f 0b eb 9b e8 23 f6 6d ff 80 3d 6c d4 19 03 00 75 8d e8 15 f6 6d ff 48 c7 c7 c0 02 55 bd c6 05 57 d4 19 03 01 e8 a2 58 49 ff <0f> 0b e9 6e ff ff ff e8 f6 f5 6d ff 80 3d 42 d4 19 03 00 0f 85 5c RSP: 0018:ffffc90002df7b98 EFLAGS: 00010282 RAX: 0000000000000000 RBX: ffff88810f6a193c RCX: ffffffffba649009 RDX: 0000000000000000 RSI: 0000000000000008 RDI: ffff88811b0283cc RBP: 0000000000000003 R08: ffffed10236060e3 R09: ffffed10236060e3 R10: 0000000000000001 R11: ffffed10236060e2 R12: ffff88810f6a193c R13: ffffc90002df7d60 R14: 0000000000000000 R15: ffff888116ae6a08 uverbs_uobject_put+0xfd/0x140 __uobj_perform_destroy+0x3d/0x60 ib_uverbs_close_xrcd+0x148/0x170 ib_uverbs_write+0xaa5/0xdf0 __vfs_write+0x7c/0x100 vfs_write+0x168/0x4a0 ksys_write+0xc8/0x200 do_syscall_64+0x9c/0x390 entry_SYSCALL_64_after_hwframe+0x44/0xa9 RIP: 0033:0x465b49 Code: f7 d8 64 89 02 b8 ff ff ff ff c3 66 0f 1f 44 00 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 bc ff ff ff f7 d8 64 89 01 48 RSP: 002b:00007f759d122c58 EFLAGS: 00000246 ORIG_RAX: 0000000000000001 RAX: ffffffffffffffda RBX: 000000000073bfa8 RCX: 0000000000465b49 RDX: 000000000000000c RSI: 0000000020000080 RDI: 0000000000000003 RBP: 0000000000000003 R08: 0000000000000000 R09: 0000000000000000 R10: 0000000000000000 R11: 0000000000000246 R12: 00007f759d1236bc R13: 00000000004ca27c R14: 000000000070de40 R15: 00000000ffffffff Dumping ftrace buffer: (ftrace buffer empty) Kernel Offset: 0x39400000 from 0xffffffff81000000 (relocation range: 0xffffffff80000000-0xffffffffbfffffff) Fixes: 7452a3c745a2 ("IB/uverbs: Allow RDMA_REMOVE_DESTROY to work concurrently with disassociate") Link: https://lore.kernel.org/r/20200527135534.482279-1-leon@kernel.org Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Leon Romanovsky <leonro@mellanox.com> Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
2020-05-27fanotify: turn off support for FAN_DIR_MODIFYAmir Goldstein
FAN_DIR_MODIFY has been enabled by commit 44d705b0370b ("fanotify: report name info for FAN_DIR_MODIFY event") in 5.7-rc1. Now we are planning further extensions to the fanotify API and during that we realized that FAN_DIR_MODIFY may behave slightly differently to be more consistent with extensions we plan. So until we finalize these extensions, let's not bind our hands with exposing FAN_DIR_MODIFY to userland. Signed-off-by: Amir Goldstein <amir73il@gmail.com> Signed-off-by: Jan Kara <jack@suse.cz>
2020-05-27Merge branch 'exec-linus' of ↵Linus Torvalds
git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace Pull execve fix from Eric Biederman: "While working on my exec cleanups I found a bug in exec that winds up miscomputing the ambient credentials during exec. Andy appears to have to been confused as to why credentials are computed for both the script and the interpreter From the original patch description: [3] Linux very confusingly processes both the script and the interpreter if applicable, for reasons that elude me. The results from thinking about a script's file capabilities and/or setuid bits are mostly discarded. The only value in struct cred that gets changed in cap_bprm_set_creds that I could find that might persist between the script and the interpreter was cap_ambient. Which is fixed with this trivial change" * 'exec-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace: exec: Always set cap_ambient in cap_bprm_set_creds
2020-05-27hwmon: (applesmc) avoid overlong udelay()Arnd Bergmann
Building this driver with "clang -O3" produces a link error after the compiler partially unrolls the loop and 256ms becomes a compile-time constant that triggers the check in udelay(): ld.lld: error: undefined symbol: __bad_udelay >>> referenced by applesmc.c >>> hwmon/applesmc.o:(read_smc) in archive drivers/built-in.a I can see no reason against using a sleeping function here, as no part of the driver runs in atomic context, so instead use usleep_range() with a wide range and use jiffies for the end condition. Signed-off-by: Arnd Bergmann <arnd@arndb.de> Link: https://lore.kernel.org/r/20200527135207.1118624-1-arnd@arndb.de Signed-off-by: Guenter Roeck <linux@roeck-us.net>
2020-05-27PM: hibernate: Restrict writes to the resume deviceDomenico Andreoli
Hibernation via snapshot device requires write permission to the swap block device, the one that more often (but not necessarily) is used to store the hibernation image. With this patch, such permissions are granted iff: 1) snapshot device config option is enabled 2) swap partition is used as resume device In other circumstances the swap device is not writable from userspace. In order to achieve this, every write attempt to a swap device is checked against the device configured as part of the uswsusp API [0] using a pointer to the inode struct in memory. If the swap device being written was not configured for resuming, the write request is denied. NOTE: this implementation works only for swap block devices, where the inode configured by swapon (which sets S_SWAPFILE) is the same used by SNAPSHOT_SET_SWAP_AREA. In case of swap file, SNAPSHOT_SET_SWAP_AREA indeed receives the inode of the block device containing the filesystem where the swap file is located (+ offset in it) which is never passed to swapon and then has not set S_SWAPFILE. As result, the swap file itself (as a file) has never an option to be written from userspace. Instead it remains writable if accessed directly from the containing block device, which is always writeable from root. [0] Documentation/power/userland-swsusp.rst v2: - rename is_hibernate_snapshot_dev() to is_hibernate_resume_dev() - fix description so to correctly refer to the resume device Signed-off-by: Domenico Andreoli <domenico.andreoli@linux.com> Acked-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2020-05-27ACPI: GED: use correct trigger type field in _Exx / _Lxx handlingArd Biesheuvel
Commit ea6f3af4c5e63f69 ("ACPI: GED: add support for _Exx / _Lxx handler methods") added a reference to the 'triggering' field of either the normal or the extended ACPI IRQ resource struct, but inadvertently used the wrong pointer in the latter case. Note that both pointers refer to the same union, and the 'triggering' field appears at the same offset in both struct types, so it currently happens to work by accident. But let's fix it nonetheless Fixes: ea6f3af4c5e63f69 ("ACPI: GED: add support for _Exx / _Lxx handler methods") Signed-off-by: Ard Biesheuvel <ardb@kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2020-05-27xfs: more lockdep whackamole with kmem_alloc*Darrick J. Wong
Dave Airlie reported the following lockdep complaint: > ====================================================== > WARNING: possible circular locking dependency detected > 5.7.0-0.rc5.20200515git1ae7efb38854.1.fc33.x86_64 #1 Not tainted > ------------------------------------------------------ > kswapd0/159 is trying to acquire lock: > ffff9b38d01a4470 (&xfs_nondir_ilock_class){++++}-{3:3}, > at: xfs_ilock+0xde/0x2c0 [xfs] > > but task is already holding lock: > ffffffffbbb8bd00 (fs_reclaim){+.+.}-{0:0}, at: > __fs_reclaim_acquire+0x5/0x30 > > which lock already depends on the new lock. > > > the existing dependency chain (in reverse order) is: > > -> #1 (fs_reclaim){+.+.}-{0:0}: > fs_reclaim_acquire+0x34/0x40 > __kmalloc+0x4f/0x270 > kmem_alloc+0x93/0x1d0 [xfs] > kmem_alloc_large+0x4c/0x130 [xfs] > xfs_attr_copy_value+0x74/0xa0 [xfs] > xfs_attr_get+0x9d/0xc0 [xfs] > xfs_get_acl+0xb6/0x200 [xfs] > get_acl+0x81/0x160 > posix_acl_xattr_get+0x3f/0xd0 > vfs_getxattr+0x148/0x170 > getxattr+0xa7/0x240 > path_getxattr+0x52/0x80 > do_syscall_64+0x5c/0xa0 > entry_SYSCALL_64_after_hwframe+0x49/0xb3 > > -> #0 (&xfs_nondir_ilock_class){++++}-{3:3}: > __lock_acquire+0x1257/0x20d0 > lock_acquire+0xb0/0x310 > down_write_nested+0x49/0x120 > xfs_ilock+0xde/0x2c0 [xfs] > xfs_reclaim_inode+0x3f/0x400 [xfs] > xfs_reclaim_inodes_ag+0x20b/0x410 [xfs] > xfs_reclaim_inodes_nr+0x31/0x40 [xfs] > super_cache_scan+0x190/0x1e0 > do_shrink_slab+0x184/0x420 > shrink_slab+0x182/0x290 > shrink_node+0x174/0x680 > balance_pgdat+0x2d0/0x5f0 > kswapd+0x21f/0x510 > kthread+0x131/0x150 > ret_from_fork+0x3a/0x50 > > other info that might help us debug this: > > Possible unsafe locking scenario: > > CPU0 CPU1 > ---- ---- > lock(fs_reclaim); > lock(&xfs_nondir_ilock_class); > lock(fs_reclaim); > lock(&xfs_nondir_ilock_class); > > *** DEADLOCK *** > > 4 locks held by kswapd0/159: > #0: ffffffffbbb8bd00 (fs_reclaim){+.+.}-{0:0}, at: > __fs_reclaim_acquire+0x5/0x30 > #1: ffffffffbbb7cef8 (shrinker_rwsem){++++}-{3:3}, at: > shrink_slab+0x115/0x290 > #2: ffff9b39f07a50e8 > (&type->s_umount_key#56){++++}-{3:3}, at: super_cache_scan+0x38/0x1e0 > #3: ffff9b39f077f258 > (&pag->pag_ici_reclaim_lock){+.+.}-{3:3}, at: > xfs_reclaim_inodes_ag+0x82/0x410 [xfs] This is a known false positive because inodes cannot simultaneously be getting reclaimed and the target of a getxattr operation, but lockdep doesn't know that. We can (selectively) shut up lockdep until either it gets smarter or we change inode reclaim not to require the ILOCK by applying a stupid GFP_NOLOCKDEP bandaid. Reported-by: Dave Airlie <airlied@gmail.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Tested-by: Dave Airlie <airlied@gmail.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-05-27xfs: force writes to delalloc regions to unwrittenDarrick J. Wong
When writing to a delalloc region in the data fork, commit the new allocations (of the da reservation) as unwritten so that the mappings are only marked written once writeback completes successfully. This fixes the problem of stale data exposure if the system goes down during targeted writeback of a specific region of a file, as tested by generic/042. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-05-27xfs: refactor xfs_iomap_prealloc_sizeDarrick J. Wong
Refactor xfs_iomap_prealloc_size to be the function that dynamically computes the per-file preallocation size by moving the allocsize= case to the caller. Break up the huge comment preceding the function to annotate the relevant parts of the code, and remove the impossible check_writeio case. Suggested-by: Christoph Hellwig <hch@infradead.org> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-05-27xfs: measure all contiguous previous extents for prealloc sizeDarrick J. Wong
When we're estimating a new speculative preallocation length for an extending write, we should walk backwards through the extent list to determine the number of number of blocks that are physically and logically contiguous with the write offset, and use that as an input to the preallocation size computation. This way, preallocation length is truly measured by the effectiveness of the allocator in giving us contiguous allocations without being influenced by the state of a given extent. This fixes both the problem where ZERO_RANGE within an EOF can reduce preallocation, and prevents the unnecessary shrinkage of preallocation when delalloc extents are turned into unwritten extents. This was found as a regression in xfs/014 after changing delalloc writes to create unwritten extents during writeback. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-05-27xfs: don't fail unwritten extent conversion on writeback due to edquotDarrick J. Wong
During writeback, it's possible for the quota block reservation in xfs_iomap_write_unwritten to fail with EDQUOT because we hit the quota limit. This causes writeback errors for data that was already written to disk, when it's not even guaranteed that the bmbt will expand to exceed the quota limit. Irritatingly, this condition is reported to userspace as EIO by fsync, which is confusing. We wrote the data, so allow the reservation. That might put us slightly above the hard limit, but it's better than losing data after a write. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2020-05-27xfs: rearrange xfs_inode_walk_ag parametersDarrick J. Wong
The perag structure already has a pointer to the xfs_mount, so we don't need to pass that separately and can drop it. Having done that, move iter_flags so that the argument order is the same between xfs_inode_walk and xfs_inode_walk_ag. The latter will make things less confusing for a future patch that enables background scanning work to be done in parallel. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-05-27xfs: straighten out all the naming around incore inode tree walksDarrick J. Wong
We're not very consistent about function names for the incore inode iteration function. Turn them all into xfs_inode_walk* variants. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2020-05-27xfs: move xfs_inode_ag_iterator to be closer to the perag walking codeDarrick J. Wong
Move the xfs_inode_ag_iterator function to be nearer xfs_inode_ag_walk so that we don't have to scroll back and forth to figure out how the incore inode walking function works. No functional changes. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-05-27xfs: use bool for done in xfs_inode_ag_walkDarrick J. Wong
This is a boolean variable, so use the bool type. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-05-27xfs: fix inode ag walk predicate function return valuesDarrick J. Wong
There are a number of predicate functions that help the incore inode walking code decide if we really want to apply the iteration function to the inode. These are boolean decisions, so change the return types to boolean to match. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-05-27xfs: refactor eofb matching into a single helperDarrick J. Wong
Refactor the two eofb-matching logics into a single helper so that we don't repeat ourselves. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2020-05-27xfs: remove __xfs_icache_free_eofblocksDarrick J. Wong
This is now a pointless wrapper, so kill it. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-05-27xfs: remove flags argument from xfs_inode_ag_walkDarrick J. Wong
The incore inode walk code passes a flags argument and a pointer from the xfs_inode_ag_iterator caller all the way to the iteration function. We can reduce the function complexity by passing flags through the private pointer. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-05-27xfs: remove xfs_inode_ag_iterator_flagsDarrick J. Wong
Combine xfs_inode_ag_iterator_flags and xfs_inode_ag_iterator_tag into a single wrapper function since there's only one caller of the _flags variant. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-05-27xfs: remove unused xfs_inode_ag_iterator functionDarrick J. Wong
Not used by anyone, so get rid of it. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-05-27xfs: replace open-coded XFS_ICI_NO_TAGDarrick J. Wong
Use XFS_ICI_NO_TAG instead of -1 when appropriate. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-05-27xfs: move eofblocks conversion function to xfs_ioctl.cDarrick J. Wong
Move xfs_fs_eofblocks_from_user into the only file that actually uses it, so that we don't have this function cluttering up the header file. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-05-27xfs: allow individual quota grace period extensionEric Sandeen
The only grace period which can be set in the kernel today is for id 0, i.e. the default grace period for all users. However, setting an individual grace period is useful; for example: Alice has a soft quota of 100 inodes, and a hard quota of 200 inodes Alice uses 150 inodes, and enters a short grace period Alice really needs to use those 150 inodes past the grace period The administrator extends Alice's grace period until next Monday vfs quota users such as ext4 can do this today, with setquota -T To enable this for XFS, we simply move the timelimit assignment out from under the (id == 0) test. Default setting remains under (id == 0). Note that this now is consistent with how we set warnings. (Userspace requires updates to enable this as well; xfs_quota needs to parse new options, and setquota needs to set appropriate field flags.) Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2020-05-27xfs: per-type quota timers and warn limitsEric Sandeen
Move timers and warnings out of xfs_quotainfo and into xfs_def_quota so that we can utilize them on a per-type basis, rather than enforcing them based on the values found in the first enabled quota type. Signed-off-by: Eric Sandeen <sandeen@redhat.com> [zlang: new way to get defquota in xfs_qm_init_timelimits] [zlang: remove redundant defq assign] Signed-off-by: Zorro Lang <zlang@redhat.com> Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2020-05-27xfs: switch xfs_get_defquota to take explicit typeEric Sandeen
xfs_get_defquota() currently takes an xfs_dquot, and from that obtains the type of default quota we should get (user/group/project). But early in init, we don't have access to a fully set up quota, so that's not possible. The next patch needs go set up default quota timers early, so switch xfs_get_defquota to take an explicit type and add a helper function to obtain that type from an xfs_dquot for the existing callers. Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2020-05-27xfs: pass xfs_dquot to xfs_qm_adjust_dqtimersEric Sandeen
Pass xfs_dquot rather than xfs_disk_dquot to xfs_qm_adjust_dqtimers; this makes it symmetric with xfs_qm_adjust_dqlimits and will help the next patch. Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Allison Collins <allison.henderson@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2020-05-27xfs: fix up some whitespace in quota codeEric Sandeen
There is a fair bit of whitespace damage in the quota code, so fix up enough of it that subsequent patches are restricted to functional change to aid review. Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Allison Collins <allison.henderson@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2020-05-27xfs: always return -ENOSPC on project quota reservation failureEric Sandeen
XFS project quota treats project hierarchies as "mini filesysems" and so rather than -EDQUOT, the intent is to return -ENOSPC when a quota reservation fails, but this behavior is not consistent. The only place we make a decision between -EDQUOT and -ENOSPC returns based on quota type is in xfs_trans_dqresv(). This behavior is currently controlled by whether or not the XFS_QMOPT_ENOSPC flag gets passed into the quota reservation. However, its use is not consistent; paths such as xfs_create() and xfs_symlink() don't set the flag, so a reservation failure will return -EDQUOT for project quota reservation failures rather than -ENOSPC for these sorts of operations, even for project quota: # mkdir mnt/project # xfs_quota -x -c "project -s -p mnt/project 42" mnt # xfs_quota -x -c 'limit -p isoft=2 ihard=3 42' mnt # touch mnt/project/file{1,2,3} touch: cannot touch ‘mnt/project/file3’: Disk quota exceeded We can make this consistent by not requiring the flag to be set at the top of the callchain; instead we can simply test whether we are reserving a project quota with XFS_QM_ISPDQ in xfs_trans_dqresv and if so, return -ENOSPC for that failure. This removes the need for the XFS_QMOPT_ENOSPC altogether and simplifies the code a fair bit. Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2020-05-27xfs: group quota should return EDQUOT when prj quota enabledEric Sandeen
Long ago, group & project quota were mutually exclusive, and so when we turned on XFS_QMOPT_ENOSPC ("return ENOSPC if project quota is exceeded") when project quota was enabled, we only needed to disable it again for user quota. When group & project quota got separated, this got missed, and as a result if project quota is enabled and group quota is exceeded, the error code returned is incorrectly returned as ENOSPC not EDQUOT. Fix this by stripping XFS_QMOPT_ENOSPC out of flags for group quota when we try to reserve the space. Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2020-05-27xfs: remove the m_active_trans counterDave Chinner
It's a global atomic counter, and we are hitting it at a rate of half a million transactions a second, so it's bouncing the counter cacheline all over the place on large machines. We don't actually need it anymore - it used to be required because the VFS freeze code could not track/prevent filesystem transactions that were running, but that problem no longer exists. Hence to remove the counter, we simply have to ensure that nothing calls xfs_sync_sb() while we are trying to quiesce the filesytem. That only happens if the log worker is still running when we call xfs_quiesce_attr(). The log worker is cancelled at the end of xfs_quiesce_attr() by calling xfs_log_quiesce(), so just call it early here and then we can remove the counter altogether. Concurrent create, 50 million inodes, identical 16p/16GB virtual machines on different physical hosts. Machine A has twice the CPU cores per socket of machine B: unpatched patched machine A: 3m16s 2m00s machine B: 4m04s 4m05s Create rates: unpatched patched machine A: 282k+/-31k 468k+/-21k machine B: 231k+/-8k 233k+/-11k Concurrent rm of same 50 million inodes: unpatched patched machine A: 6m42s 2m33s machine B: 4m47s 4m47s The transaction rate on the fast machine went from just under 300k/sec to 700k/sec, which indicates just how much of a bottleneck this atomic counter was. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2020-05-27xfs: separate read-only variables in struct xfs_mountDave Chinner
Seeing massive cpu usage from xfs_agino_range() on one machine; instruction level profiles look similar to another machine running the same workload, only one machine is consuming 10x as much CPU as the other and going much slower. The only real difference between the two machines is core count per socket. Both are running identical 16p/16GB virtual machine configurations Machine A: 25.83% [k] xfs_agino_range 12.68% [k] __xfs_dir3_data_check 6.95% [k] xfs_verify_ino 6.78% [k] xfs_dir2_data_entry_tag_p 3.56% [k] xfs_buf_find 2.31% [k] xfs_verify_dir_ino 2.02% [k] xfs_dabuf_map.constprop.0 1.65% [k] xfs_ag_block_count And takes around 13 minutes to remove 50 million inodes. Machine B: 13.90% [k] __pv_queued_spin_lock_slowpath 3.76% [k] do_raw_spin_lock 2.83% [k] xfs_dir3_leaf_check_int 2.75% [k] xfs_agino_range 2.51% [k] __raw_callee_save___pv_queued_spin_unlock 2.18% [k] __xfs_dir3_data_check 2.02% [k] xfs_log_commit_cil And takes around 5m30s to remove 50 million inodes. Suspect is cacheline contention on m_sectbb_log which is used in one of the macros in xfs_agino_range. This is a read-only variable but shares a cacheline with m_active_trans which is a global atomic that gets bounced all around the machine. The workload is trying to run hundreds of thousands of transactions per second and hence cacheline contention will be occurring on this atomic counter. Hence xfs_agino_range() is likely just be an innocent bystander as the cache coherency protocol fights over the cacheline between CPU cores and sockets. On machine A, this rearrangement of the struct xfs_mount results in the profile changing to: 9.77% [kernel] [k] xfs_agino_range 6.27% [kernel] [k] __xfs_dir3_data_check 5.31% [kernel] [k] __pv_queued_spin_lock_slowpath 4.54% [kernel] [k] xfs_buf_find 3.79% [kernel] [k] do_raw_spin_lock 3.39% [kernel] [k] xfs_verify_ino 2.73% [kernel] [k] __raw_callee_save___pv_queued_spin_unlock Vastly less CPU usage in xfs_agino_range(), but still 3x the amount of machine B and still runs substantially slower than it should. Current rm -rf of 50 million files: vanilla patched machine A 13m20s 6m42s machine B 5m30s 5m02s It's an improvement, hence indicating that separation and further optimisation of read-only global filesystem data is worthwhile, but it clearly isn't the underlying issue causing this specific performance degradation. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2020-05-27xfs: reduce free inode accounting overheadDave Chinner
Shaokun Zhang reported that XFS was using substantial CPU time in percpu_count_sum() when running a single threaded benchmark on a high CPU count (128p) machine from xfs_mod_ifree(). The issue is that the filesystem is empty when the benchmark runs, so inode allocation is running with a very low inode free count. With the percpu counter batching, this means comparisons when the counter is less that 128 * 256 = 32768 use the slow path of adding up all the counters across the CPUs, and this is expensive on high CPU count machines. The summing in xfs_mod_ifree() is only used to fire an assert if an underrun occurs. The error is ignored by the higher level code. Hence this is really just debug code and we don't need to run it on production kernels, nor do we need such debug checks to return error values just to trigger an assert. Finally, xfs_mod_icount/xfs_mod_ifree are only called from xfs_trans_unreserve_and_mod_sb(), so get rid of them and just directly call the percpu_counter_add/percpu_counter_compare functions. The compare functions are now run only on debug builds as they are internal to ASSERT() checks and so only compiled in when ASSERTs are active (CONFIG_XFS_DEBUG=y or CONFIG_XFS_WARN=y). Reported-by: Shaokun Zhang <zhangshaokun@hisilicon.com> Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2020-05-27xfs: gut error handling in xfs_trans_unreserve_and_mod_sb()Dave Chinner
xfs: gut error handling in xfs_trans_unreserve_and_mod_sb() From: Dave Chinner <dchinner@redhat.com> The error handling in xfs_trans_unreserve_and_mod_sb() is largely incorrect - rolling back the changes in the transaction if only one counter underruns makes all the other counters incorrect. We still allow the change to proceed and committing the transaction, except now we have multiple incorrect counters instead of a single underflow. Further, we don't actually report the error to the caller, so this is completely silent except on debug kernels that will assert on failure before we even get to the rollback code. Hence this error handling is broken, untested, and largely unnecessary complexity. Just remove it. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2020-05-27x86: Hide the archdata.iommu field behind generic IOMMU_APIKrzysztof Kozlowski
There is a generic, kernel wide configuration symbol for enabling the IOMMU specific bits: CONFIG_IOMMU_API. Implementations (including INTEL_IOMMU and AMD_IOMMU driver) select it so use it here as well. This makes the conditional archdata.iommu field consistent with other platforms and also fixes any compile test builds of other IOMMU drivers, when INTEL_IOMMU or AMD_IOMMU are not selected). For the case when INTEL_IOMMU/AMD_IOMMU and COMPILE_TEST are not selected, this should create functionally equivalent code/choice. With COMPILE_TEST this field could appear if other IOMMU drivers are chosen but neither INTEL_IOMMU nor AMD_IOMMU are not. Reported-by: kbuild test robot <lkp@intel.com> Fixes: e93a1695d7fb ("iommu: Enable compile testing for some of drivers") Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org> Acked-by: Borislav Petkov <bp@suse.de> Link: https://lore.kernel.org/r/20200518120855.27822-2-krzk@kernel.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
generated by cgit (git 2.34.1) at 2025-09-14 06:47:01 +0000