capsnaps will take inode references via ihold when queueing to flush.
When force unmounting, the client will just close the sessions and
may never get a flush reply, causing a leak and inode ref leak.
Fix this by removing the capsnaps for an inode when removing the caps.
URL: https://tracker.ceph.com/issues/52295
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
The current code will update the mtime and then try to get caps to
handle the write. If we end up having to request caps from the MDS, then
the mtime in the cap grant will clobber the updated mtime and it'll be
lost.
This is most noticable when two clients are alternately writing to the
same file. Fw caps are continually being granted and revoked, and the
mtime ends up stuck because the updated mtimes are always being
overwritten with the old one.
Fix this by changing the order of operations in ceph_write_iter to get
the caps before updating the times. Also, make sure we check the pool
full conditions before even getting any caps or uninlining.
URL: https://tracker.ceph.com/issues/46574
Reported-by: Jozef Kováč <kovac@firma.zoznam.sk>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Luis Henriques <lhenriques@suse.de>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
In the case where the export MDS has crashed just after the EImportStart
journal is flushed, a standby MDS takes over for it and when replaying
the EImportStart journal the MDS will wait the client to reconnect. That
may never happen because the client may not have registered or opened
the sessions yet.
When receiving a new map, ensure we reconnect to valid export targets as
well if their sessions don't exist yet.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Print a bit more information when we can't find the realm during
ceph_add_cap. Show both the inode number and the old realm inode
number.
Suggested-by: Sage Weil <sage@redhat.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Ilya Dryomov <idryomov@gmail.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Consolidate some fiddly code for changing an inode's snap_realm
into a new helper function, and change the callers to use it.
While we're in here, nothing uses the i_snap_realm_counter field, so
remove that from the inode.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Luis Henriques <lhenriques@suse.de>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
The ceph_mds_client and ceph_mds_session structures are kzalloc'ed so
there's no need to explicitly initialize either of their fields to 0.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Ilya Dryomov <idryomov@gmail.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
The first thing metric_delayed_work does is check mdsc->stopping,
and then return immediately if it's set. That's good since we would
have already torn down the metric structures at this point, otherwise,
but there is no locking around mdsc->stopping.
It's possible that the ceph_metric_destroy call could race with the
delayed_work, in which case we could end up with the delayed_work
accessing destroyed percpu variables.
At this point in the mdsc teardown, the "stopping" flag has already been
set, so there's no benefit to flushing the work. Move the work
cancellation in ceph_metric_destroy ahead of the percpu variable
destruction, and eliminate the flush_delayed_work call in
ceph_mdsc_destroy.
Fixes: 18f473b384 ("ceph: periodically send perf metrics to MDSes")
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Xiubo Li <xiubli@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
For the client requests who will have unsafe and safe replies from
MDS daemons, in the MDS side the MDS daemons won't flush the mdlog
(journal log) immediatelly, because they think it's unnecessary.
That's true for most cases but not all, likes the fsync request.
The fsync will wait until all the unsafe replied requests to be
safely replied.
Normally if there have multiple threads or clients are running, the
whole mdlog in MDS daemons could be flushed in time if any request
will trigger the mdlog submit thread. So usually we won't experience
the normal operations will stuck for a long time. But in case there
has only one client with only thread is running, the stuck phenomenon
maybe obvious and the worst case it must wait at most 5 seconds to
wait the mdlog to be flushed by the MDS's tick thread periodically.
This patch will trigger to flush the mdlog in the relevant and auth
MDSes to which the in-flight requests are sent just before waiting
the unsafe requests to finish.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
If we hit a decoding error late in the frame, then we might exit the
function without putting the pool_ns string. Ensure that we always put
that reference on the way out of the function.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Ilya Dryomov <idryomov@gmail.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Pull VFIO updates from Alex Williamson:
- Fix dma-valid return WAITED implementation (Anthony Yznaga)
- SPDX license cleanups (Cai Huoqing)
- Split vfio-pci-core from vfio-pci and enhance PCI driver matching to
support future vendor provided vfio-pci variants (Yishai Hadas, Max
Gurtovoy, Jason Gunthorpe)
- Replace duplicated reflck with core support for managing first open,
last close, and device sets (Jason Gunthorpe, Max Gurtovoy, Yishai
Hadas)
- Fix non-modular mdev support and don't nag about request callback
support (Christoph Hellwig)
- Add semaphore to protect instruction intercept handler and replace
open-coded locks in vfio-ap driver (Tony Krowiak)
- Convert vfio-ap to vfio_register_group_dev() API (Jason Gunthorpe)
* tag 'vfio-v5.15-rc1' of git://github.com/awilliam/linux-vfio: (37 commits)
vfio/pci: Introduce vfio_pci_core.ko
vfio: Use kconfig if XX/endif blocks instead of repeating 'depends on'
vfio: Use select for eventfd
PCI / VFIO: Add 'override_only' support for VFIO PCI sub system
PCI: Add 'override_only' field to struct pci_device_id
vfio/pci: Move module parameters to vfio_pci.c
vfio/pci: Move igd initialization to vfio_pci.c
vfio/pci: Split the pci_driver code out of vfio_pci_core.c
vfio/pci: Include vfio header in vfio_pci_core.h
vfio/pci: Rename ops functions to fit core namings
vfio/pci: Rename vfio_pci_device to vfio_pci_core_device
vfio/pci: Rename vfio_pci_private.h to vfio_pci_core.h
vfio/pci: Rename vfio_pci.c to vfio_pci_core.c
vfio/ap_ops: Convert to use vfio_register_group_dev()
s390/vfio-ap: replace open coded locks for VFIO_GROUP_NOTIFY_SET_KVM notification
s390/vfio-ap: r/w lock for PQAP interception handler function pointer
vfio/type1: Fix vfio_find_dma_valid return
vfio-pci/zdev: Remove repeated verbose license text
vfio: platform: reset: Convert to SPDX identifier
vfio: Remove struct vfio_device_ops open/release
...
Switch the driver to support the .get_state() method.
Signed-off-by: Jitao Shi <jitao.shi@mediatek.com>
[thierry.reding@gmail.com: add missing linux/bitfield.h include]
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
Current calculation for period and high_width may have 64-bit overflow.
state->period and rate are u64. rate * state->period will overflow.
clk_div = div_u64(rate * state->period, NSEC_PER_SEC)
period = div64_u64(rate * state->period, div);
high_width = div64_u64(rate * state->duty_cycle, div);
This patch is to resolve it by using mul_u64_u64_div_u64().
Signed-off-by: Jitao Shi <jitao.shi@mediatek.com>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
The isa_bus_to_virt is only needed under X86 and in fact the code
that sets the ibft_phys_addr is only compiled under X86.
As such lets just ifdef the code.
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Vijayendra Suman <vijayendra.suman@oracle.com>
CC: Maurizio Lombardi <mlombard@redhat.com>
CC: Mike Rapoport <rppt@kernel.org>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
---
v2: Remove the ibft_phys_addr as it is defined in iscsi_ibft.h
Pull xen updates from Juergen Gross:
- some small cleanups
- a fix for a bug when running as Xen PV guest which could result in
not all memory being transferred in case of a migration of the guest
- a small series for getting rid of code for supporting very old Xen
hypervisor versions nobody should be using since many years now
- a series for hardening the Xen block frontend driver
- a fix for Xen PV boot code issuing warning messages due to a stray
preempt_disable() on the non-boot processors
* tag 'for-linus-5.15-rc1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
xen: remove stray preempt_disable() from PV AP startup code
xen/pcifront: Removed unnecessary __ref annotation
x86: xen: platform-pci-unplug: use pr_err() and pr_warn() instead of raw printk()
drivers/xen/xenbus/xenbus_client.c: fix bugon.cocci warnings
xen/blkfront: don't trust the backend response data blindly
xen/blkfront: don't take local copy of a request from the ring page
xen/blkfront: read response from backend only once
xen: assume XENFEAT_gnttab_map_avail_bits being set for pv guests
xen: assume XENFEAT_mmu_pt_update_preserve_ad being set for pv guests
xen: check required Xen features
xen: fix setting of max_pfn in shared_info
Pull parisc architecture updates from Helge Deller:
- Fix a kernel crash when a signal is delivered to bad userspace stack
- Fix fall-through warnings in math-emu code
- Increase size of gcc stack frame check
- Switch coding from 'pci_' to 'dma_' API
- Make struct parisc_driver::remove() return void
- Some parisc related Makefile changes
- Minor cleanups, e.g. change to octal permissions, fix macro
collisions, fix PMD_ORDER collision, replace spaces with tabs
* tag 'for-5.15/parisc' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
parisc: math-emu: Fix fall-through warnings
parisc: fix crash with signals and alloca
parisc: Fix compile failure when building 64-bit kernel natively
parisc: ccio-dma.c: Added tab instead of spaces
parisc/parport_gsc: switch from 'pci_' to 'dma_' API
parisc: move core-y in arch/parisc/Makefile to arch/parisc/Kbuild
parisc: switch from 'pci_' to 'dma_' API
parisc: Make struct parisc_driver::remove() return void
parisc: remove unused arch/parisc/boot/install.sh and its phony target
parisc: Rename PMD_ORDER to PMD_TABLE_ORDER
parisc: math-emu: Avoid "fmt" macro collision
parisc: Increase size of gcc stack frame check
parisc: Replace symbolic permissions with octal permissions
Pull Xtensa updates from Max Filippov:
- fix kconfig unmet dependency warning for HAVE_FUTEX_CMPXCHG
- add fairness to handling IRQs of the same priority
- fix pointer usage before NULL check in ISS console driver
- build system cleanups
* tag 'xtensa-20210902' of git://github.com/jcmvbkbc/linux-xtensa:
xtensa: move core-y in arch/xtensa/Makefile to arch/xtensa/Kbuild
xtensa: build platform directories unconditionally
xtensa: do not build variants directory
xtensa: remove unneeded exports
xtensa: ISS: don't use string pointer before NULL check
xtensa: add fairness to IRQ handling
xtensa: fix kconfig unmet dependency warning for HAVE_FUTEX_CMPXCHG
No need to do the full VPID based task lookup and validation of the top
waiter when the user space futex was acquired on it's behalf during the
requeue_pi operation. The task is known already and it cannot go away
before requeue_pi_wake_futex() has been invoked.
Split out the actual attach code from attach_pi_state_owner() and use that
instead of the full blown variant.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20210902094414.676104881@linutronix.de
The recent rework of the requeue PI code introduced a possibility for
going back to user space in inconsistent state:
CPU 0 CPU 1
requeue_futex()
if (lock_pifutex_user()) {
dequeue_waiter();
wake_waiter(task);
sched_in(task);
return_from_futex_syscall();
---> Inconsistent state because PI state is not established
It becomes worse if the woken up task immediately exits:
sys_exit();
attach_pistate(vpid); <--- FAIL
Attach the pi state before dequeuing and waking the waiter. If the waiter
gets a spurious wakeup before the dequeue operation it will wait in
futex_requeue_pi_wakeup_sync() and therefore cannot return and exit.
Fixes: 07d91ef510 ("futex: Prevent requeue_pi() lock nesting issue on RT")
Reported-by: syzbot+4d1bd0725ef09168e1a0@syzkaller.appspotmail.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20210902094414.558914045@linutronix.de
The check on the rt_waiter and top_waiter->pi_state is assigning an error
return code to ret but this later gets re-assigned, hence the check is
ineffective.
Return -EINVAL rather than assigning it to ret which was the original
intent.
Fixes: dc7109aaa2 ("futex: Validate waiter correctly in futex_proxy_trylock_atomic()")
Addresses-Coverity: ("Unused value")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: André Almeida <andrealmeid@collabora.com>
Link: https://lore.kernel.org/r/20210818131840.34262-1-colin.king@canonical.com
The clks "main" and "mm" are prepared in .probe() (and unprepared in
.remove()). This results in the clocks being on during suspend which
results in unnecessarily increased power consumption.
Remove the clock operations from .probe() and .remove(). Add the
clk_prepare_enable() in .enable() and the clk_disable_unprepare() in
.disable().
Signed-off-by: Jitao Shi <jitao.shi@mediatek.com>
[thierry.reding@gmail.com: squashed in fixup patch]
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
Pull memory model updates from Ingo Molnar:
"LKMM updates:
- Update documentation and code example
KCSAN updates:
- Introduce CONFIG_KCSAN_STRICT (which RCU uses)
- Optimize use of get_ctx() by kcsan_found_watchpoint()
- Rework atomic.h into permissive.h
- Add the ability to ignore writes that change only one bit of a
given data-racy variable.
- Improve comments"
* tag 'locking-debug-2021-09-01' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
tools/memory-model: Document data_race(READ_ONCE())
tools/memory-model: Heuristics using data_race() must handle all values
tools/memory-model: Add example for heuristic lockless reads
tools/memory-model: Make read_foo_diagnostic() more clearly diagnostic
kcsan: Make strict mode imply interruptible watchers
kcsan: permissive: Ignore data-racy 1-bit value changes
kcsan: Print if strict or non-strict during init
kcsan: Rework atomic.h into permissive.h
kcsan: Reduce get_ctx() uses in kcsan_found_watchpoint()
kcsan: Introduce CONFIG_KCSAN_STRICT
kcsan: Remove CONFIG_KCSAN_DEBUG
kcsan: Improve some Kconfig comments
- Add max-virtual-functions to endpoint binding (Kishon Vijay Abraham I)
- Add pci_epf_add_vepf() API to add virtual function to endpoint (Kishon
Vijay Abraham I)
- Add pci_epf_vepf_link() to link virtual function to endpoint physical
function (Kishon Vijay Abraham I)
- Add virtual function number to pci_epc_ops endpoint ops interfaces
(Kishon Vijay Abraham I)
- Simplify register base address computation for endpoint BAR configuration
(Kishon Vijay Abraham I)
- Add support to configure virtual functions in cadence endpoint driver
(Kishon Vijay Abraham I)
- Add SR-IOV configuration to endpoint test driver (Kishon Vijay Abraham I)
- Document configfs usage to create virtual functions for endpoints (Kishon
Vijay Abraham I)
* remotes/lorenzo/pci/endpoint:
Documentation: PCI: endpoint/pci-endpoint-cfs: Guide to use SR-IOV
misc: pci_endpoint_test: Populate sriov_configure ops to configure SR-IOV device
PCI: cadence: Add support to configure virtual functions
PCI: cadence: Simplify code to get register base address for configuring BAR
PCI: endpoint: Add virtual function number in pci_epc ops
PCI: endpoint: Add support to link a physical function to a virtual function
PCI: endpoint: Add support to add virtual function in endpoint core
dt-bindings: PCI: pci-ep: Add binding to specify virtual function
- Fix runtime PM imbalance in rcar_pcie_ep_probe() (Dinghao Liu)
* remotes/lorenzo/pci/rcar:
PCI: rcar: Add L1 link state fix into data abort hook
PCI: rcar: Fix runtime PM imbalance in rcar_pcie_ep_probe()
- Split DT bindings for PCIe controllers with independent MSI domains into
separate nodes for MT2712/MT7622 (Chuanjia Liu)
- Locate shared registers from "mediatek,generic-pciecfg" property
(Chuanjia Liu)
- Get IRQ from "pcie_irq" if "interrupt-names" property is present to fix
an MSI issue (Chuanjia Liu)
- Get PCI domain from "linux,pci-domain" property if present (Chuanjia Liu)
* remotes/lorenzo/pci/mediatek:
PCI: mediatek: Use PCI domain to handle ports detection
PCI: mediatek: Add new method to get irq number
PCI: mediatek: Add new method to get shared pcie-cfg base address
dt-bindings: PCI: mediatek: Update the Device tree bindings
- Add domain_nr in struct pci_host_bridge (Boqun Feng)
- Use host bridge MSI domain for root buses if present (Boqun Feng)
- Allow ARM64 virtual host bridge with no ACPI companion (e.g., Hyper-V)
(Boqun Feng)
- Make Hyper-V enumeration more generic (Arnd Bergmann)
- Set Hyper-V domain_nr at probe-time (Boqun Feng)
- Set up Hyper-V MSI domain at bridge probe-time (Boqun Feng)
- Enable Hyper-V bridge probing on ARM64 (Boqun Feng)
* remotes/lorenzo/pci/hyper-v:
PCI: hv: Turn on the host bridge probing on ARM64
PCI: hv: Set up MSI domain at bridge probing time
PCI: hv: Set ->domain_nr of pci_host_bridge at probing time
PCI: hv: Generify PCI probing
arm64: PCI: Support root bridge preparation for Hyper-V
arm64: PCI: Restructure pcibios_root_bridge_prepare()
PCI: Support populating MSI domains of root buses via bridges
PCI: Introduce domain_nr in pci_host_bridge
- Convert bool in structs to bitfield (Kishon Vijay Abraham I)
- Work around J7200 non-PCIe SERDES lane electrical issue that prevents
PCIe link training (Nadeem Athani)
- Add J7200 PCIe support to j721e (Kishon Vijay Abraham I)
- Add AM64 PCIe support to j721e (Kishon Vijay Abraham I)
- Add J7200 and AM64 device IDs to endpoint test (Kishon Vijay Abraham I)
* remotes/lorenzo/pci/cadence:
misc: pci_endpoint_test: Add deviceID for AM64 and J7200
PCI: j721e: Add PCIe support for AM64
PCI: j721e: Add PCIe support for J7200
PCI: cadence: Add quirk flag to set minimum delay in LTSSM Detect.Quiet state
PCI: cadence: Use bitfield for *quirk_retrain_flag* instead of bool
- Fix PIO config access status checking (Evan Wang)
- Increase config access polling delay to 1.5s (Pali Rohár)
- Add PCIe Root Capabilities to bridge emulation (Pali Rohár)
- Report Config Request Retry Status when Software Visibility enabled (Pali
Rohár)
- Add back configuration of PCIe resources from 'ranges' DT property and
pay attention to DT size and CPU/PCI offset to fix issues with I/O port
space (Pali Rohár)
- Serialize masking and unmasking legacy INTx interrupts (Pali Rohár)
* remotes/lorenzo/pci/aardvark:
PCI: aardvark: Fix masking and unmasking legacy INTx interrupts
PCI: aardvark: Configure PCIe resources from 'ranges' DT property
PCI: aardvark: Fix reporting CRS value
PCI: pci-bridge-emul: Add PCIe Root Capabilities Register
PCI: aardvark: Increase polling delay to 1.5s while waiting for PIO response
PCI: aardvark: Fix checking for PIO status
