| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Since the RDMA Read I/O state is now contained in the recv_ctxt,
svc_rdma_read_special() can use that recv_ctxt to derive the
read_info rather than the other way around. This removes another
usage of the ri_readctxt field, enabling its removal in a
subsequent patch.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Since the RDMA Read I/O state is now contained in the recv_ctxt,
svc_rdma_read_call_chunk() can use that recv_ctxt to derive the
read_info rather than the other way around. This removes another
usage of the ri_readctxt field, enabling its removal in a
subsequent patch.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Since the RDMA Read I/O state is now contained in the recv_ctxt,
svc_rdma_read_multiple_chunks() can use that recv_ctxt to derive the
read_info rather than the other way around. This removes another
usage of the ri_readctxt field, enabling its removal in a
subsequent patch.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Since the RDMA Read I/O state is now contained in the recv_ctxt,
svc_rdma_copy_inline_range() can use that recv_ctxt to derive the
read_info rather than the other way around. This removes another
usage of the ri_readctxt field, enabling its removal in a
subsequent patch.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Since the RDMA Read I/O state is now contained in the recv_ctxt,
svc_rdma_build_read_data_item() can use that recv_ctxt to derive
that information rather than the other way around. This removes
another usage of the ri_readctxt field, enabling its removal in a
subsequent patch.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Since the RDMA Read I/O state is now contained in the recv_ctxt,
svc_rdma_build_read_chunk_range() can use that recv_ctxt to derive
that information rather than the other way around. This removes
another usage of the ri_readctxt field, enabling its removal in a
subsequent patch.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Since the RDMA Read I/O state is now contained in the recv_ctxt,
svc_rdma_build_read_chunk() can use that recv_ctxt to derive that
information rather than the other way around. This removes another
usage of the ri_readctxt field, enabling its removal in a
subsequent patch.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Since the RDMA Read I/O state is now contained in the recv_ctxt,
svc_rdma_build_read_segment() can use the recv_ctxt to derive that
information rather than the other way around. This removes one usage
of the ri_readctxt field, enabling its removal in a subsequent
patch.
At the same time, the use of ri_rqst can similarly be replaced with
a passed-in function parameter.
Start with build_read_segment() because it is a common utility
function at the bottom of the Read chunk path.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Further clean up: move the starting byte offset field into
svc_rdma_recv_ctxt.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Further clean up: move the page index field into svc_rdma_recv_ctxt.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Since the request's svc_rdma_recv_ctxt will stay around for the
duration of the RDMA Read operation, the contents of struct
svc_rdma_read_info can reside in the request's svc_rdma_recv_ctxt
rather than being allocated separately. This will eventually save a
call to kmalloc() in a hot path.
Start this clean-up by moving the Read chunk's svc_rdma_chunk_ctxt.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Prepare for nestling these into the send and recv ctxts so they
no longer have to be allocated dynamically.
Signed-off-by: Chuck Lever <[email protected]>
|
|
In every instance, the pointer address in that field is now
available by other means.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Enable the eventual removal of the svc_rdma_chunk_ctxt::cc_rdma
field.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Enable the eventual removal of the svc_rdma_chunk_ctxt::cc_rdma
field.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Enable the eventual removal of the svc_rdma_chunk_ctxt::cc_rdma
field.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Enable the eventual removal of the svc_rdma_chunk_ctxt::cc_rdma
field.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Enable the removal of the svc_rdma_chunk_ctxt::cc_rdma field in a
subsequent patch.
Signed-off-by: Chuck Lever <[email protected]>
|
|
SG_CHUNK_SIZE is 128, making struct svc_rdma_rw_ctxt + the first
SGL array more than 4200 bytes in length, pushing the memory
allocation well into order 1.
Even so, the RDMA rw core doesn't seem to use more than max_send_sge
entries in that array (typically 32 or less), so that is all wasted
space.
Signed-off-by: Chuck Lever <[email protected]>
|
|
A send/recv_ctxt already records transport-related information
in the cq.id, thus there is no need to record the IP addresses of
the transport endpoints.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Update the DMA error flow tracepoints to report the completion ID of
the failing context. This ties the wait/failure to a particular
operation or request, which is more useful than knowing only the
failing transport.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Update the Send Queue's error flow tracepoints to report the
completion ID of the waiting or failing context. This ties the
wait/failure to a particular operation or request, which is a little
more useful than knowing only the transport that is about to close.
Signed-off-by: Chuck Lever <[email protected]>
|
|
De-duplicate some code, making it easier to add new tracepoints that
report only a completion ID.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Two svcrdma-related transport locks can become quite contended.
Collate their use and make them easy to find in /proc/lock_stat for
better observability.
Signed-off-by: Chuck Lever <[email protected]>
|
|
There's no need to protect llist_entry() with a spin lock.
Signed-off-by: Chuck Lever <[email protected]>
|
|
DMA unmapping can take quite some time, so it should not be handled
in a single-threaded completion handler. Defer releasing write_info
structs to the recently-added workqueue.
With this patch, DMA unmapping can be handled in parallel, and it
does not cause head-of-queue blocking of Write completions.
Signed-off-by: Chuck Lever <[email protected]>
|
|
DMA unmapping can take quite some time, so it should not be handled
in a single-threaded completion handler. Defer releasing send_ctxts
to the recently-added workqueue.
With this patch, DMA unmapping can be handled in parallel, and it
does not cause head-of-queue blocking of Send completions.
Signed-off-by: Chuck Lever <[email protected]>
|
|
To handle work in the background, set up an UNBOUND workqueue for
svcrdma. Subsequent patches will make use of it.
Signed-off-by: Chuck Lever <[email protected]>
|
|
The original reason for allocating svc_rdma_recv_ctxt objects during
Receive completion was to ensure the objects were allocated on the
NUMA node closest to the underlying IB device.
Since commit c5d68d25bd6b ("svcrdma: Clean up allocation of
svc_rdma_recv_ctxt"), however, the device's favored node is
explicitly passed to the memory allocator.
To enable switching Receive completion to soft IRQ context, move
memory allocation out of completion handling, since it can be
costly, and it can sleep.
A limited number of objects is now allocated at "accept" time.
Signed-off-by: Chuck Lever <[email protected]>
|
|
The svc_rdma_recv_ctxt free list uses a lockless list to avoid the
need for a spin lock in the fast path. llist_del_first(), which is
used by svc_rdma_recv_ctxt_get(), requires serialization, however,
when there are multiple list producers that are unserialized.
I mistakenly thought there was only one caller of
svc_rdma_recv_ctxt_get() (svc_rdma_refresh_recvs()), thus explicit
serialization would not be necessary. But there is another caller:
svc_rdma_bc_sendto(), and these two are not serialized against each
other. I haven't seen ill effects that I could directly ascribe to
a lack of serialization. It's just an observation based on code
audit.
When DMA-mapping before sending a Reply, the passed-in struct
svc_rdma_recv_ctxt is used only for its write and reply PCLs. These
are currently always empty in the backchannel case. So, instead of
passing a full svc_rdma_recv_ctxt object to
svc_rdma_map_reply_msg(), let's pass in just the Write and Reply
PCLs.
This change makes it unnecessary for the backchannel to acquire a
dummy svc_rdma_recv_ctxt object when sending an RPC Call. The need
for svc_rdma_recv_ctxt free list serialization is now completely
avoided.
Signed-off-by: Chuck Lever <[email protected]>
|
|
Callback operations enum is defined in client and server, move it to
common header file.
Signed-off-by: ChenXiaoSong <[email protected]>
Acked-by: Anna Schumaker <[email protected]>
Reviewed-by: Jeff Layton <[email protected]>
Signed-off-by: Chuck Lever <[email protected]>
|
|
We check "state" for NULL on the previous line so it can't be NULL here.
No need to check again.
Reported-by: kernel test robot <[email protected]>
Closes: https://lore.kernel.org/r/[email protected]/
Signed-off-by: Dan Carpenter <[email protected]>
Reviewed-by: Jeff Layton <[email protected]>
Signed-off-by: Chuck Lever <[email protected]>
|
|
This flag is no longer used.
Reviewed-by: Jeff Layton <[email protected]>
Signed-off-by: Chuck Lever <[email protected]>
|
|
Avoid the use of an atomic bitop, and prepare for adding a run-time
switch for using splice reads.
Reviewed-by: Jeff Layton <[email protected]>
Signed-off-by: Chuck Lever <[email protected]>
|
|
RQ_SPLICE_OK is a bit of a layering violation. Also, a subsequent
patch is going to provide a mechanism for always disabling splice
reads.
Splicing is an issue only for NFS READs, so refactor nfsd_read() to
check the auth type directly instead of relying on an rq_flag
setting.
The new helper will be added into the NFSv4 read path in a
subsequent patch.
Reviewed-by: Jeff Layton <[email protected]>
Signed-off-by: Chuck Lever <[email protected]>
|
|
NFSD will use this new API to determine whether nfsd_splice_read is
safe to use. This avoids the need to add a dependency to NFSD for
CONFIG_SUNRPC_GSS.
Reviewed-by: Jeff Layton <[email protected]>
Signed-off-by: Chuck Lever <[email protected]>
|
|
Al Viro notes that normal system calls hold f_pos_lock when calling
->iterate_shared and ->llseek; however nfsd_readdir() does not take
that mutex when calling these methods.
It should be safe however because the struct file acquired by
nfsd_readdir() is not visible to other threads.
Reviewed-by: Jeff Layton <[email protected]>
Signed-off-by: Chuck Lever <[email protected]>
|
|
This trace point was for debugging the DRC's garbage collection. In
the field it's just noise.
Reviewed-by: Jeff Layton <[email protected]>
Signed-off-by: Chuck Lever <[email protected]>
|
|
workqueue: nfsd_file_delayed_close [nfsd] hogged CPU for >13333us 8
times, consider switching to WQ_UNBOUND
There's no harm in closing a cached file descriptor on another core.
Signed-off-by: Chuck Lever <[email protected]>
|
|
The usage of read_seqbegin_or_lock() in nfsd_copy_write_verifier()
is wrong. "seq" is always even and thus "or_lock" has no effect,
this code can never take ->writeverf_lock for writing.
I guess this is fine, nfsd_copy_write_verifier() just copies 8 bytes
and nfsd_reset_write_verifier() is supposed to be very rare operation
so we do not need the adaptive locking in this case.
Yet the code looks wrong and sub-optimal, it can use read_seqbegin()
without changing the behaviour.
[ cel: Note also that it eliminates this Sparse warning:
fs/nfsd/nfssvc.c:360:6: warning: context imbalance in 'nfsd_copy_write_verifier' -
different lock contexts for basic block
]
Signed-off-by: Oleg Nesterov <[email protected]>
Reviewed-by: Jeff Layton <[email protected]>
Reviewed-by: NeilBrown <[email protected]>
Signed-off-by: Chuck Lever <[email protected]>
|
|
We've had a number of attempts at different NFSv4 client tracking
methods over the years, but now nfsdcld has emerged as the clear winner
since the others (recoverydir and the usermodehelper upcall) are
problematic.
As a case in point, the recoverydir backend uses MD5 hashes to encode
long form clientid strings, which means that nfsd repeatedly gets dinged
on FIPS audits, since MD5 isn't considered secure. Its use of MD5 is not
cryptographically significant, so there is no danger there, but allowing
us to compile that out allows us to sidestep the issue entirely.
As a prelude to eventually removing support for these client tracking
methods, add a new Kconfig option that enables them. Mark it deprecated
and make it default to N.
Acked-by: NeilBrown <[email protected]>
Signed-off-by: Jeff Layton <[email protected]>
Signed-off-by: Chuck Lever <[email protected]>
|
|
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux
Pull i2c fixes from Wolfram Sang:
"Improve the detection when to run atomic transfer handlers for kernels
with preemption disabled. This removes some false positive splats a
number of users were seeing if their driver didn't have support for
atomic transfers.
Also, fix a typo in the docs while we are here"
* tag 'i2c-for-6.7-final' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
i2c: core: Fix atomic xfer check for non-preempt config
Documentation/i2c: fix spelling error in i2c-address-translators
|
|
Since commit aa49c90894d0 ("i2c: core: Run atomic i2c xfer when
!preemptible"), the whole reboot/power off sequence on non-preempt kernels
is using atomic i2c xfer, as !preemptible() always results to 1.
During device_shutdown(), the i2c might be used a lot and not all busses
have implemented an atomic xfer handler. This results in a lot of
avoidable noise, like:
[ 12.687169] No atomic I2C transfer handler for 'i2c-0'
[ 12.692313] WARNING: CPU: 6 PID: 275 at drivers/i2c/i2c-core.h:40 i2c_smbus_xfer+0x100/0x118
...
Fix this by allowing non-atomic xfer when the interrupts are enabled, as
it was before.
Link: https://lore.kernel.org/r/20231222230106.73f030a5@yea
Link: https://lore.kernel.org/r/[email protected]
Link: https://lore.kernel.org/linux-i2c/[email protected]/
Fixes: aa49c90894d0 ("i2c: core: Run atomic i2c xfer when !preemptible")
Cc: [email protected] # v5.2+
Signed-off-by: Benjamin Bara <[email protected]>
Tested-by: Michael Walle <[email protected]>
Tested-by: Tor Vic <[email protected]>
[wsa: removed a comment which needs more work, code is ok]
Signed-off-by: Wolfram Sang <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
Pull misc mm fixes from Andrew Morton:
"12 hotfixes.
Two are cc:stable and the remainder either address post-6.7 issues or
aren't considered necessary for earlier kernel versions"
* tag 'mm-hotfixes-stable-2024-01-05-11-35' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm:
mm: shrinker: use kvzalloc_node() from expand_one_shrinker_info()
mailmap: add entries for Mathieu Othacehe
MAINTAINERS: change vmware.com addresses to broadcom.com
arch/mm/fault: fix major fault accounting when retrying under per-VMA lock
mm/mglru: skip special VMAs in lru_gen_look_around()
MAINTAINERS: hand over hwpoison maintainership to Miaohe Lin
MAINTAINERS: remove hugetlb maintainer Mike Kravetz
mm: fix unmap_mapping_range high bits shift bug
mm: memcg: fix split queue list crash when large folio migration
mm: fix arithmetic for max_prop_frac when setting max_ratio
mm: fix arithmetic for bdi min_ratio
mm: align larger anonymous mappings on THP boundaries
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/cel/linux
Pull nfsd fix from Chuck Lever:
- Fix another regression in the NFSD administrative API
* tag 'nfsd-6.7-3' of git://git.kernel.org/pub/scm/linux/kernel/git/cel/linux:
nfsd: drop the nfsd_put helper
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394
Pull firewire fix from Takashi Sakamoto:
"A single patch to suppress unexpected system reboot in AMD Ryzen
machines with PCIe card consisting of Asmedia ASM1083/1085 and
VT6306/6307/6308.
When the 1394 OHCI driver for the card accesses a specific register
in PCI memory space, the system reboot often occurs.
The issue affects all versions of Linux kernel as long as the 1394
OHCI driver is included. The mechanism of unexpected system reboot is
not clear, so the driver is changed to avoid the access itself when
detecting the combination of hardware"
* tag 'firewire-fixes-6.7-final' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394:
firewire: ohci: suppress unexpected system reboot in AMD Ryzen machines and ASM108x/VT630x PCIe cards
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc
Pull MMC fixes from Ulf Hansson:
"MMC core:
- Fix releasing the host by canceling the delayed work
- Fix pause retune on all RPMB partitions
MMC host:
- meson-mx-sdhc: Fix HW hang during card initialization
- sdhci-sprd: Fix eMMC init failure after HW reset"
* tag 'mmc-v6.7-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc:
mmc: sdhci-sprd: Fix eMMC init failure after hw reset
mmc: core: Cancel delayed work before releasing host
mmc: rpmb: fixes pause retune on all RPMB partitions.
mmc: meson-mx-sdhc: Fix initialization frozen issue
|
|
Pull more drm fixes from Dave Airlie:
"The amdgpu ones are fairly normal, the one that is a bit large is a
fix for a newly introduced IP in 6.7 so unlikely to cause regressions.
The nouveau ones are mostly memory leaks and debugging cleanups from
the GSP (new nvidia firmware) enablement. There are some GSP changes
to the message passing code and a subsequent fix for eDP panel turn
on, that means my laptop can turn on the panel in GSP mode. These are
fairly low chance of disrupting things since GSP is new in 6.7. The
final not all in GSP fix is a deadlock seen with i915/nouveau when GSP
is used where the the fence and irq paths have locking inversions,
I've pushed some irq enablement out to a workqueue, and this has seen
some fairly decent testing.
amdgpu:
- DP MST fix
- SMU 13.0.6 fixes
- fix displays on macbooks using vega12
- fix VSC and colorimetry on DP/eDP
nouveau:
- fix deadlock between fence signalling and irq paths
- fix GSP memory leaks
- fix GSP leftover debug
- hide some GSP callback messages
- fix GSP display disable path
- fix GSP ACPI interaction
- handle errors in ctrl messages
- use errors info to fix DP link training"
* tag 'drm-fixes-2024-01-05' of git://anongit.freedesktop.org/drm/drm:
drm/nouveau/dp: Honor GSP link training retry timeouts
nouveau: push event block/allowing out of the fence context
nouveau/gsp: always free the alloc messages on r535
nouveau/gsp: don't free ctrl messages on errors
nouveau/gsp: convert gsp errors to generic errors
drm/nouveau/gsp: Fix ACPI MXDM/MXDS method invocations
nouveau/gsp: free userd allocation.
nouveau/gsp: free acpi object after use
nouveau: fix disp disabling with GSP
nouveau/gsp: drop some acpi related debug
nouveau/gsp: add three notifier callbacks that we see in normal operation (v2)
drm/amd/pm: Use gpu_metrics_v1_5 for SMUv13.0.6
drm/amd/pm: Add gpu_metrics_v1_5
drm/amd/pm: Add mem_busy_percent for GCv9.4.3 apu
drm/amd/display: Fix sending VSC (+ colorimetry) packets for DP/eDP displays without PSR
drm/amdgpu: skip gpu_info fw loading on navi12
drm/amd/display: add nv12 bounding box
drm/amd/pm: Update metric table for jpeg/vcn data
drm/amd/pm: Use separate metric table for APU
drm/amd/display: pbn_div need be updated for hotplug event
|
|
syzbot is reporting uninit-value at shrinker_alloc(), for commit
307bececcd12 ("mm: shrinker: add a secondary array for
shrinker_info::{map, nr_deferred}") which assumed that the ->unit was
allocated with __GFP_ZERO forgot to replace kvmalloc_node() in
expand_one_shrinker_info() with kvzalloc_node().
Link: https://lkml.kernel.org/r/[email protected]
Reported-by: syzbot <[email protected]>
Closes: https://syzkaller.appspot.com/bug?extid=1e0ed05798af62917464
Fixes: 307bececcd12 ("mm: shrinker: add a secondary array for shrinker_info::{map, nr_deferred}")
Signed-off-by: Tetsuo Handa <[email protected]>
Acked-by: Qi Zheng <[email protected]>
Cc: Muchun Song <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|