Currently rtrs when create_qp use a coarse numbers (bigger in general),
which leads to hardware create more resources which only waste memory
with no benefits.
- SERVICE con,
For max_send_wr/max_recv_wr, it's 2 times SERVICE_CON_QUEUE_DEPTH + 2
- IO con
For max_send_wr/max_recv_wr, it's sess->queue_depth * 3 + 1
Fixes: 6a98d71dae ("RDMA/rtrs: client: main functionality")
Link: https://lore.kernel.org/r/20201217141915.56989-6-jinpu.wang@cloud.ionos.com
Signed-off-by: Jack Wang <jinpu.wang@cloud.ionos.com>
Reviewed-by: Md Haris Iqbal <haris.iqbal@cloud.ionos.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Fix a race in rxe_mcast.c that occurs when two QPs try at the same time to
attach a multicast address. Both QPs lookup the mgid address in a pool of
multicast groups and if they do not find it create a new group elem.
Fix this by locking the lookup/alloc/add key sequence and using the
unlocked APIs added in this patch set.
Link: https://lore.kernel.org/r/20201216231550.27224-8-rpearson@hpe.com
Signed-off-by: Bob Pearson <rpearson@hpe.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
The existing pool APIs use the rw_lock pool_lock to protect critical
sections that change the pool state. This does not correctly implement a
typical sequence like the following
elem = <lookup key in pool>
if found use elem else
elem = <alloc new elem in pool>
<add key to elem>
Which is racy if multiple threads are attempting to perform this at the
same time. We want the second thread to use the elem created by the first
thread not create two equivalent elems.
This patch adds new APIs that are the same as existing APIs but do not
take the pool_lock. A caller can then take the lock and perform a sequence
of pool operations and then release the lock.
Link: https://lore.kernel.org/r/20201216231550.27224-7-rpearson@hpe.com
Signed-off-by: Bob Pearson <rpearson@hpe.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Replace 'void *' parameters with 'struct rxe_pool_entry *' and use a macro
to allow:
rxe_add_index,
rxe_drop_index,
rxe_add_key,
rxe_drop_key and
rxe_add_to_pool
APIs to be type safe against changing the position of pelem in the
objects.
Link: https://lore.kernel.org/r/20201216231550.27224-6-rpearson@hpe.com
Signed-off-by: Bob Pearson <rpearson@hpe.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
The allocate, lookup index, lookup key and cleanup routines in rxe_pool.c
currently are not type safe against relocating the pelem field in the
objects. Planned changes to move allocation of objects into rdma-core make
addressing this a requirement.
Use the elem_offset field in rxe_type_info make these APIs safe against
moving the pelem field.
Link: https://lore.kernel.org/r/20201216231550.27224-5-rpearson@hpe.com
Signed-off-by: Bob Pearson <rpearson@hpe.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
The rxe verbs objects each include an rdma-core object 'ib_xxx'
and a rxe_pool_entry 'pelem' in addition to rxe specific data.
Originally these all had pelem first and ib_xxx second. Currently
about half have ib_xxx first and half have pelem first. Saving
the offset of the pelem field in rxe_type info will enable making
the rxe_pool APIs type safe as the pelem field continues to vary.
Link: https://lore.kernel.org/r/20201216231550.27224-4-rpearson@hpe.com
Signed-off-by: Bob Pearson <rpearson@hpe.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
During connection setup, the application may choose to zero-size inbound
and outbound READ queues, as well as the Receive queue. This patch fixes
handling of zero-sized queues, but not prevents it.
Kamal Heib says in an initial error report:
When running the blktests over siw the following shift-out-of-bounds is
reported, this is happening because the passed IRD or ORD from the ulp
could be zero which will lead to unexpected behavior when calling
roundup_pow_of_two(), fix that by blocking zero values of ORD or IRD.
UBSAN: shift-out-of-bounds in ./include/linux/log2.h:57:13
shift exponent 64 is too large for 64-bit type 'long unsigned int'
CPU: 20 PID: 3957 Comm: kworker/u64:13 Tainted: G S 5.10.0-rc6 #2
Hardware name: Dell Inc. PowerEdge R630/02C2CP, BIOS 2.1.5 04/11/2016
Workqueue: iw_cm_wq cm_work_handler [iw_cm]
Call Trace:
dump_stack+0x99/0xcb
ubsan_epilogue+0x5/0x40
__ubsan_handle_shift_out_of_bounds.cold.11+0xb4/0xf3
? down_write+0x183/0x3d0
siw_qp_modify.cold.8+0x2d/0x32 [siw]
? __local_bh_enable_ip+0xa5/0xf0
siw_accept+0x906/0x1b60 [siw]
? xa_load+0x147/0x1f0
? siw_connect+0x17a0/0x17a0 [siw]
? lock_downgrade+0x700/0x700
? siw_get_base_qp+0x1c2/0x340 [siw]
? _raw_spin_unlock_irqrestore+0x39/0x40
iw_cm_accept+0x1f4/0x430 [iw_cm]
rdma_accept+0x3fa/0xb10 [rdma_cm]
? check_flush_dependency+0x410/0x410
? cma_rep_recv+0x570/0x570 [rdma_cm]
nvmet_rdma_queue_connect+0x1a62/0x2680 [nvmet_rdma]
? nvmet_rdma_alloc_cmds+0xce0/0xce0 [nvmet_rdma]
? lock_release+0x56e/0xcc0
? lock_downgrade+0x700/0x700
? lock_downgrade+0x700/0x700
? __xa_alloc_cyclic+0xef/0x350
? __xa_alloc+0x2d0/0x2d0
? rdma_restrack_add+0xbe/0x2c0 [ib_core]
? __ww_mutex_die+0x190/0x190
cma_cm_event_handler+0xf2/0x500 [rdma_cm]
iw_conn_req_handler+0x910/0xcb0 [rdma_cm]
? _raw_spin_unlock_irqrestore+0x39/0x40
? trace_hardirqs_on+0x1c/0x150
? cma_ib_handler+0x8a0/0x8a0 [rdma_cm]
? __kasan_kmalloc.constprop.7+0xc1/0xd0
cm_work_handler+0x121c/0x17a0 [iw_cm]
? iw_cm_reject+0x190/0x190 [iw_cm]
? trace_hardirqs_on+0x1c/0x150
process_one_work+0x8fb/0x16c0
? pwq_dec_nr_in_flight+0x320/0x320
worker_thread+0x87/0xb40
? __kthread_parkme+0xd1/0x1a0
? process_one_work+0x16c0/0x16c0
kthread+0x35f/0x430
? kthread_mod_delayed_work+0x180/0x180
ret_from_fork+0x22/0x30
Fixes: a531975279 ("rdma/siw: main include file")
Fixes: f29dd55b02 ("rdma/siw: queue pair methods")
Fixes: 8b6a361b8c ("rdma/siw: receive path")
Fixes: b9be6f18cf ("rdma/siw: transmit path")
Fixes: 303ae1cdfd ("rdma/siw: application interface")
Link: https://lore.kernel.org/r/20210108125845.1803-1-bmt@zurich.ibm.com
Reported-by: Kamal Heib <kamalheib1@gmail.com>
Reported-by: Yi Zhang <yi.zhang@redhat.com>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Bernard Metzler <bmt@zurich.ibm.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
The parameter of kfree function is NULL, so kfree code is useless, delete
it. Therefore, goto expression is no longer needed, so simplify
it. cma_dev_group is always pre-zero'd before reaching make_cma_ports, so
the NULL set to cma_dev_group->ports is unneeded too.
Link: https://lore.kernel.org/r/20201216080219.18184-1-zhengyongjun3@huawei.com
Signed-off-by: Zheng Yongjun <zhengyongjun3@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
This change fixes the checkpatch warning described in
commit cbacb5ab0a ("docs: printk-formats: Stop encouraging use of
unnecessary %h[xudi] and %hh[xudi]")
Standard integer promotion is already done and %hx and %hhx is useless so
do not encourage the use of %hh[xudi] or %h[xudi].
Link: https://lore.kernel.org/r/20201223193041.122850-1-trix@redhat.com
Signed-off-by: Tom Rix <trix@redhat.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Pull s390 cleanups from Vasily Gorbik:
"Update defconfigs and sort config select list"
* tag 's390-5.11-3' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
s390/Kconfig: sort config S390 select list once again
s390: update defconfigs
Pull power management fixes from Rafael Wysocki:
"These fix a crash in intel_pstate during resume from suspend-to-RAM
that may occur after recent changes and two resource leaks in error
paths in the operating performance points (OPP) framework, add a new
C-states table to intel_idle and update the cpuidle MAINTAINERS entry
to cover the governors too.
Specifics:
- Fix recently introduced crash in the intel_pstate driver that
occurs if scale-invariance is disabled during resume from
suspend-to-RAM due to inconsistent changes of APERF or MPERF MSR
values made by the platform firmware (Rafael Wysocki).
- Fix a memory leak and add a missing clk_put() in error paths in the
OPP framework (Quanyang Wang, Viresh Kumar).
- Add new C-states table for SnowRidge processors to the intel_idle
driver (Artem Bityutskiy).
- Update the MAINTAINERS entry for cpuidle to make it clear that the
governors are covered by it too (Lukas Bulwahn)"
* tag 'pm-5.11-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
intel_idle: add SnowRidge C-state table
cpufreq: intel_pstate: Fix fast-switch fallback path
opp: Call the missing clk_put() on error
opp: fix memory leak in _allocate_opp_table
MAINTAINERS: include governors into CPU IDLE TIME MANAGEMENT FRAMEWORK
Pull SCSI fixes from James Bottomley:
"This is a load of driver fixes (12 ufs, 1 mpt3sas, 1 cxgbi).
The big core two fixes are for power management ("block: Do not accept
any requests while suspended" and "block: Fix a race in the runtime
power management code") which finally sorts out the resume problems
we've occasionally been having.
To make the resume fix, there are seven necessary precursors which
effectively renames REQ_PREEMPT to REQ_PM, so every "special" request
in block is automatically a power management exempt one.
All of the non-PM preempt cases are removed except for the one in the
SCSI Parallel Interface (spi) domain validation which is a genuine
case where we have to run requests at high priority to validate the
bus so this becomes an autopm get/put protected request"
* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi: (22 commits)
scsi: cxgb4i: Fix TLS dependency
scsi: ufs: Un-inline ufshcd_vops_device_reset function
scsi: ufs: Re-enable WriteBooster after device reset
scsi: ufs-mediatek: Use correct path to fix compile error
scsi: mpt3sas: Signedness bug in _base_get_diag_triggers()
scsi: block: Do not accept any requests while suspended
scsi: block: Remove RQF_PREEMPT and BLK_MQ_REQ_PREEMPT
scsi: core: Only process PM requests if rpm_status != RPM_ACTIVE
scsi: scsi_transport_spi: Set RQF_PM for domain validation commands
scsi: ide: Mark power management requests with RQF_PM instead of RQF_PREEMPT
scsi: ide: Do not set the RQF_PREEMPT flag for sense requests
scsi: block: Introduce BLK_MQ_REQ_PM
scsi: block: Fix a race in the runtime power management code
scsi: ufs-pci: Enable UFSHCD_CAP_RPM_AUTOSUSPEND for Intel controllers
scsi: ufs-pci: Fix recovery from hibernate exit errors for Intel controllers
scsi: ufs-pci: Ensure UFS device is in PowerDown mode for suspend-to-disk ->poweroff()
scsi: ufs-pci: Fix restore from S4 for Intel controllers
scsi: ufs-mediatek: Keep VCC always-on for specific devices
scsi: ufs: Allow regulators being always-on
scsi: ufs: Clear UAC for RPMB after ufshcd resets
...
Pull block fixes from Jens Axboe:
"Two minor block fixes from this last week that should go into 5.11:
- Add missing NOWAIT debugfs definition (Andres)
- Fix kerneldoc warning introduced this merge window (Randy)"
* tag 'block-5.11-2021-01-01' of git://git.kernel.dk/linux-block:
block: add debugfs stanza for QUEUE_FLAG_NOWAIT
fs: block_dev.c: fix kernel-doc warnings from struct block_device changes
Pull io_uring fixes from Jens Axboe:
"A few fixes that should go into 5.11, all marked for stable as well:
- Fix issue around identity COW'ing and users that share a ring
across processes
- Fix a hang associated with unregistering fixed files (Pavel)
- Move the 'process is exiting' cancelation a bit earlier, so
task_works aren't affected by it (Pavel)"
* tag 'io_uring-5.11-2021-01-01' of git://git.kernel.dk/linux-block:
kernel/io_uring: cancel io_uring before task works
io_uring: fix io_sqe_files_unregister() hangs
io_uring: add a helper for setting a ref node
io_uring: don't assume mm is constant across submits
Commit 436e980e2e ("kbuild: don't hardcode depmod path") stopped
hard-coding the path of depmod, but in the process caused trouble for
distributions that had that /sbin location, but didn't have it in the
PATH (generally because /sbin is limited to the super-user path).
Work around it for now by just adding /sbin to the end of PATH in the
depmod.sh script.
Reported-and-tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
For cancelling io_uring requests it needs either to be able to run
currently enqueued task_works or having it shut down by that moment.
Otherwise io_uring_cancel_files() may be waiting for requests that won't
ever complete.
Go with the first way and do cancellations before setting PF_EXITING and
so before putting the task_work infrastructure into a transition state
where task_work_run() would better not be called.
Cc: stable@vger.kernel.org # 5.5+
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
io_sqe_files_unregister() uninterruptibly waits for enqueued ref nodes,
however requests keeping them may never complete, e.g. because of some
userspace dependency. Make sure it's interruptible otherwise it would
hang forever.
Cc: stable@vger.kernel.org # 5.6+
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Pull ceph fixes from Ilya Dryomov:
"A fix for an edge case in MClientRequest encoding and a couple of
trivial fixups for the new msgr2 support"
* tag 'ceph-for-5.11-rc2' of git://github.com/ceph/ceph-client:
libceph: add __maybe_unused to DEFINE_MSGR2_FEATURE
libceph: align session_key and con_secret to 16 bytes
libceph: fix auth_signature buffer allocation in secure mode
ceph: reencode gid_list when reconnecting
Add C-state table for the SnowRidge SoC which is found on Intel Jacobsville
platforms.
The following has been changed.
1. C1E latency changed from 10us to 15us. It was measured using the
open source "wult" tool (the "nic" method, 15us is the 99.99th
percentile).
2. C1E power break even changed from 20us to 25us, which may result
in less C1E residency in some workloads.
3. C6 latency changed from 50us to 130us. Measured the same way as C1E.
The C6 C-state is supported only by some SnowRidge revisions, so add a C-state
table commentary about this.
On SnowRidge, C6 support is enumerated via the usual mechanism: "mwait" leaf of
the "cpuid" instruction. The 'intel_idle' driver does check this leaf, so even
though C6 is present in the table, the driver will only use it if the CPU does
support it.
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
When sugov_update_single_perf() falls back to the "frequency"
path due to the missing scale-invariance, it will call
cpufreq_driver_fast_switch() via sugov_fast_switch()
and the driver's ->fast_switch() callback will be invoked,
so it must not be NULL.
However, after commit a365ab6b9d ("cpufreq: intel_pstate: Implement
the ->adjust_perf() callback") intel_pstate sets ->fast_switch() to
NULL when it is going to use intel_cpufreq_adjust_perf(), which is a
mistake, because on x86 the scale-invariance may be turned off
dynamically, so modify it to retain the original ->adjust_perf()
callback pointer.
Fixes: a365ab6b9d ("cpufreq: intel_pstate: Implement the ->adjust_perf() callback")
Reported-by: Kenneth R. Crudup <kenny@panix.com>
Tested-by: Kenneth R. Crudup <kenny@panix.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Pull operating performance points (OPP) framework fixes for 5.11-rc2
from Viresh Kumar:
"This contains two patches to fix freeing of resources in error paths."
* 'opp/linux-next' of git://git.kernel.org/pub/scm/linux/kernel/git/vireshk/pm:
opp: Call the missing clk_put() on error
opp: fix memory leak in _allocate_opp_table
Fix new kernel-doc warnings in fs/block_dev.c:
../fs/block_dev.c:1066: warning: Excess function parameter 'whole' description in 'bd_abort_claiming'
../fs/block_dev.c:1837: warning: Function parameter or member 'dev' not described in 'lookup_bdev'
Fixes: 4e7b5671c6 ("block: remove i_bdev")
Fixes: 37c3fc9abb ("block: simplify the block device claiming interface")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Christoph Hellwig <hch@lst.de>
Cc: linux-fsdevel@vger.kernel.org
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Merge misc fixes from Andrew Morton:
"16 patches
Subsystems affected by this patch series: mm (selftests, hugetlb,
pagecache, mremap, kasan, and slub), kbuild, checkpatch, misc, and
lib"
* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
mm: slub: call account_slab_page() after slab page initialization
zlib: move EXPORT_SYMBOL() and MODULE_LICENSE() out of dfltcc_syms.c
lib/zlib: fix inflating zlib streams on s390
lib/genalloc: fix the overflow when size is too big
kdev_t: always inline major/minor helper functions
sizes.h: add SZ_8G/SZ_16G/SZ_32G macros
local64.h: make <asm/local64.h> mandatory
kasan: fix null pointer dereference in kasan_record_aux_stack
mm: generalise COW SMC TLB flushing race comment
mm/mremap.c: fix extent calculation
mm: memmap defer init doesn't work as expected
mm: add prototype for __add_to_page_cache_locked()
checkpatch: prefer strscpy to strlcpy
Revert "kbuild: avoid static_assert for genksyms"
mm/hugetlb: fix deadlock in hugetlb_cow error path
selftests/vm: fix building protection keys test
It's convenient to have page->objects initialized before calling into
account_slab_page(). In particular, this information can be used to
pre-alloc the obj_cgroup vector.
Let's call account_slab_page() a bit later, after the initialization of
page->objects.
This commit doesn't bring any functional change, but is required for
further optimizations.
[akpm@linux-foundation.org: undo changes needed by forthcoming mm-memcg-slab-pre-allocate-obj_cgroups-for-slab-caches-with-slab_account.patch]
Link: https://lkml.kernel.org/r/20201110195753.530157-1-guro@fb.com
Signed-off-by: Roman Gushchin <guro@fb.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Decompressing zlib streams on s390 fails with "incorrect data check"
error.
Userspace zlib checks inflate_state.flags in order to byteswap checksums
only for zlib streams, and s390 hardware inflate code, which was ported
from there, tries to match this behavior. At the same time, kernel zlib
does not use inflate_state.flags, so it contains essentially random
values. For many use cases either zlib stream is zeroed out or checksum
is not used, so this problem is masked, but at least SquashFS is still
affected.
Fix by always passing a checksum to and from the hardware as is, which
matches zlib_inflate()'s expectations.
Link: https://lkml.kernel.org/r/20201215155551.894884-1-iii@linux.ibm.com
Fixes: 1261961000 ("lib/zlib: add s390 hardware support for kernel zlib_inflate")
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Tested-by: Christian Borntraeger <borntraeger@de.ibm.com>
Acked-by: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Cc: <stable@vger.kernel.org> [5.6+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Some graphic card has very big memory on chip, such as 32G bytes.
In the following case, it will cause overflow:
pool = gen_pool_create(PAGE_SHIFT, NUMA_NO_NODE);
ret = gen_pool_add(pool, 0x1000000, SZ_32G, NUMA_NO_NODE);
va = gen_pool_alloc(pool, SZ_4G);
The overflow occurs in gen_pool_alloc_algo_owner():
....
size = nbits << order;
....
The @nbits is "int" type, so it will overflow.
Then the gen_pool_avail() will return the wrong value.
This patch converts some "int" to "unsigned long", and
changes the compare code in while.
Link: https://lkml.kernel.org/r/20201229060657.3389-1-sjhuang@iluvatar.ai
Signed-off-by: Huang Shijie <sjhuang@iluvatar.ai>
Reported-by: Shi Jiasheng <jiasheng.shi@iluvatar.ai>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
I'm not sure if I'm completely missing something here, but AFAIKS the
reference to the mysterious "COW SMC race" confuses the issue. The
original changelog and mailing list thread didn't help me either.
This SMC race is where the problem was detected, but isn't the general
problem bigger and more obvious: that the new PTE could be picked up at
any time by any TLB while entries for the old PTE exist in other TLBs
before the TLB flush takes effect?
The case where the iTLB and dTLB of a CPU are pointing at different pages
is an interesting one but follows from the general problem.
The other (minor) thing with the comment I think it makes it a bit clearer
to say what the old code was doing (i.e., it avoids the race as opposed to
what?).
References: 4ce072f1fa ("mm: fix a race condition under SMC + COW")
Link: https://lkml.kernel.org/r/20201215121119.351650-1-npiggin@gmail.com
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Hugh Dickins <hughd@google.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suresh Siddha <sbsiddha@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
