Jens Axboe [Tue, 2 Aug 2016 14:45:44 +0000 (08:45 -0600)]
blk-mq: fix deadlock in blk_mq_register_disk() error path
If we fail registering any of the hardware queues, we call
into blk_mq_unregister_disk() with the hotplug mutex already
held. Since blk_mq_unregister_disk() attempts to acquire the
same mutex, we end up in a less than happy place.
Dan Williams [Sun, 31 Jul 2016 18:15:13 +0000 (11:15 -0700)]
block: fix bdi vs gendisk lifetime mismatch
The name for a bdi of a gendisk is derived from the gendisk's devt.
However, since the gendisk is destroyed before the bdi it leaves a
window where a new gendisk could dynamically reuse the same devt while a
bdi with the same name is still live. Arrange for the bdi to hold a
reference against its "owner" disk device while it is registered.
Otherwise we can hit sysfs duplicate name collisions like the following:
blk-mq: Allow timeouts to run while queue is freezing
In case a submitted request gets stuck for some reason, the block layer
can prevent the request starvation by starting the scheduled timeout work.
If this stuck request occurs at the same time another thread has started
a queue freeze, the blk_mq_timeout_work will not be able to acquire the
queue reference and will return silently, thus not issuing the timeout.
But since the request is already holding a q_usage_counter reference and
is unable to complete, it will never release its reference, preventing
the queue from completing the freeze started by first thread. This puts
the request_queue in a hung state, forever waiting for the freeze
completion.
This was observed while running IO to a NVMe device at the same time we
toggled the CPU hotplug code. Eventually, once a request got stuck
requiring a timeout during a queue freeze, we saw the CPU Hotplug
notification code get stuck inside blk_mq_freeze_queue_wait, as shown in
the trace below.
The fix is to allow the timeout work to execute in the window between
dropping the initial refcount reference and the release of the last
reference, which actually marks the freeze completion. This can be
achieved with percpu_refcount_tryget, which does not require the counter
to be alive. This way the timeout work can do it's job and terminate a
stuck request even during a freeze, returning its reference and avoiding
the deadlock.
Allowing the timeout to run is just a part of the fix, since for some
devices, we might get stuck again inside the device driver's timeout
handler, should it attempt to allocate a new request in that path -
which is a quite common action for Abort commands, which need to be sent
after a timeout. In NVMe, for instance, we call blk_mq_alloc_request
from inside the timeout handler, which will fail during a freeze, since
it also tries to acquire a queue reference.
I considered a similar change to blk_mq_alloc_request as a generic
solution for further device driver hangs, but we can't do that, since it
would allow new requests to disturb the freeze process. I thought about
creating a new function in the block layer to support unfreezable
requests for these occasions, but after working on it for a while, I
feel like this should be handled in a per-driver basis. I'm now
experimenting with changes to the NVMe timeout path, but I'm open to
suggestions of ways to make this generic.
It seems fairly obvious that device_create_file() is not being protected
from being run concurrently on the same nbd.
Quentin found the following relevant commits:
1a2ad21 nbd: add locking to nbd_ioctl 90b8f28 [PATCH] end of methods switch: remove the old ones d4430d6 [PATCH] beginning of methods conversion 08f8585 [PATCH] move block_device_operations to blkdev.h
It would seem that the race was introduced in the process of moving nbd
from BKL to unlocked ioctls.
By setting nbd->task_recv while the mutex is held, we can prevent other
processes from running concurrently (since nbd->task_recv is also checked
while the mutex is held).
This problem can occur in the following situation:
open()
- pread()
- .seq_start()
- iter = kmalloc() // succeeds
- seqf->private = iter
- .seq_stop()
- kfree(seqf->private)
- pread()
- .seq_start()
- iter = kmalloc() // fails
- .seq_stop()
- class_dev_iter_exit(seqf->private) // boom! old pointer
As the comment in disk_seqf_stop() says, stop is called even if start
failed, so we need to reinitialise the private pointer to NULL when seq
iteration stops.
An alternative would be to set the private pointer to NULL when the
kmalloc() in disk_seqf_start() fails.
Merge 4fc29c1aa375 included this extra line, but it's not needed (or
useful) since we'll bio_set_op_attrs() right after to properly set
the op and flags for the bio.
Paolo Valente [Wed, 27 Jul 2016 05:22:05 +0000 (07:22 +0200)]
block: add missing group association in bio-cloning functions
When a bio is cloned, the newly created bio must be associated with
the same blkcg as the original bio (if BLK_CGROUP is enabled). If
this operation is not performed, then the new bio is not associated
with any group, and the group of the current task is returned when
the group of the bio is requested.
Depending on the cloning frequency, this may cause a large
percentage of the bios belonging to a given group to be treated
as if belonging to other groups (in most cases as if belonging to
the root group). The expected group isolation may thereby be broken.
This commit adds the missing association in bio-cloning functions.
Hou Tao [Tue, 26 Jul 2016 09:13:42 +0000 (17:13 +0800)]
blkcg: kill unused field nr_undestroyed_grps
'nr_undestroyed_grps' in struct throtl_data was used to count
the number of throtl_grp related with throtl_data, but now
throtl_grp is tracked by blkcg_gq, so it is useless anymore.
Jan Kara [Tue, 26 Jul 2016 09:38:20 +0000 (11:38 +0200)]
writeback: Write dirty times for WB_SYNC_ALL writeback
Currently we take care to handle I_DIRTY_TIME in vfs_fsync() and
queue_io() so that inodes which have only dirty timestamps are properly
written on fsync(2) and sync(2). However there are other call sites -
most notably going through write_inode_now() - which expect inode to be
clean after WB_SYNC_ALL writeback. This is not currently true as we do
not clear I_DIRTY_TIME in __writeback_single_inode() even for
WB_SYNC_ALL writeback in all the cases. This then resulted in the
following oops because bdev_write_inode() did not clean the inode and
writeback code later stumbled over a dirty inode with detached wb.
Jiri Kosina [Thu, 16 Jun 2016 07:53:58 +0000 (09:53 +0200)]
floppy: fix open(O_ACCMODE) for ioctl-only open
Commit 09954bad4 ("floppy: refactor open() flags handling"), as a
side-effect, causes open(/dev/fdX, O_ACCMODE) to fail. It turns out that
this is being used setfdprm userspace for ioctl-only open().
Reintroduce back the original behavior wrt !(FMODE_READ|FMODE_WRITE)
modes, while still keeping the original O_NDELAY bug fixed.
James Hogan [Thu, 4 Aug 2016 16:36:08 +0000 (17:36 +0100)]
metag: Fix __cmpxchg_u32 asm constraint for CMP
The LNKGET based atomic sequence in __cmpxchg_u32 has slightly incorrect
constraints for the return value which under certain circumstances can
allow an address unit register to be used as the first operand of a CMP
instruction. This isn't a valid instruction however as the encodings
only allow a data unit to be specified. This would result in an
assembler error like the following:
Error: failed to assemble instruction: "CMP A0.2,D0Ar6"
Fix by changing the constraint from "=&da" (assigned, early clobbered,
data or address unit register) to "=&d" (data unit register only).
The constraint for the second operand, "bd" (an op2 register where op1
is a data unit register and the instruction supports O2R) is already
correct assuming the first operand is a data unit register.
Other cases of CMP in inline asm have had their constraints checked, and
appear to all be fine.
Moni Shoua [Thu, 16 Jun 2016 13:45:23 +0000 (16:45 +0300)]
Soft RoCE driver
Soft RoCE (RXE) - The software RoCE driver
ib_rxe implements the RDMA transport and registers to the RDMA core
device as a kernel verbs provider. It also implements the packet IO
layer. On the other hand ib_rxe registers to the Linux netdev stack
as a udp encapsulating protocol, in that case RDMA, for sending and
receiving packets over any Ethernet device. This yields a RDMA
transport over the UDP/Ethernet network layer forming a RoCEv2
compatible device.
The configuration procedure of the Soft RoCE drivers requires
binding to any existing Ethernet network device. This is done with
/sys interface.
A userspace Soft RoCE library (librxe) provides user applications
the ability to run with Soft RoCE devices. The use of rxe verbs ins
user space requires the inclusion of librxe as a device specifics
plug-in to libibverbs. librxe is packaged separately.
dm raid: fix use of wrong status char during resynchronization
During a resynchronization, device status char 'a' is output on the raid
status line for every device of a RAID set. It changes from 'a' to 'A'
(unless device failure) when the resynchronization completes.
Interrupting and restarting a resynchronization, by reloading the DM
table, erroneously lead to status char 'A'.
Fix this by avoiding setting the MD_RECOVERY_REQUESTED flag in
raid_preresume().
Linus Torvalds [Thu, 4 Aug 2016 13:14:38 +0000 (09:14 -0400)]
Merge tag 'modules-next-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux
Pull module updates from Rusty Russell:
"The only interesting thing here is Jessica's patch to add
ro_after_init support to modules. The rest are all trivia"
* tag 'modules-next-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux:
extable.h: add stddef.h so "NULL" definition is not implicit
modules: add ro_after_init support
jump_label: disable preemption around __module_text_address().
exceptions: fork exception table content from module.h into extable.h
modules: Add kernel parameter to blacklist modules
module: Do a WARN_ON_ONCE() for assert module mutex not held
Documentation/module-signing.txt: Note need for version info if reusing a key
module: Invalidate signatures on force-loaded modules
module: Issue warnings when tainting kernel
module: fix redundant test.
module: fix noreturn attribute for __module_put_and_exit()
Linus Torvalds [Thu, 4 Aug 2016 12:51:12 +0000 (08:51 -0400)]
Merge branch 'akpm' (patches from Andrew)
Merge even more updates from Andrew Morton:
- dma-mapping API cleanup
- a few cleanups and misc things
- use jump labels in dynamic-debug
* emailed patches from Andrew Morton <[email protected]>:
dynamic_debug: add jump label support
jump_label: remove bug.h, atomic.h dependencies for HAVE_JUMP_LABEL
arm: jump label may reference text in __exit
tile: support static_key usage in non-module __exit sections
sparc: support static_key usage in non-module __exit sections
powerpc: add explicit #include <asm/asm-compat.h> for jump label
drivers/media/dvb-frontends/cxd2841er.c: avoid misleading gcc warning
MAINTAINERS: update email and list of Samsung HW driver maintainers
block: remove BLK_DEV_DAX config option
samples/kretprobe: fix the wrong type
samples/kretprobe: convert the printk to pr_info/pr_err
samples/jprobe: convert the printk to pr_info/pr_err
samples/kprobe: convert the printk to pr_info/pr_err
dma-mapping: use unsigned long for dma_attrs
media: mtk-vcodec: remove unused dma_attrs
include/linux/bitmap.h: cleanup
tree-wide: replace config_enabled() with IS_ENABLED()
drivers/fpga/Kconfig: fix build failure
Jason Baron [Wed, 3 Aug 2016 20:46:39 +0000 (13:46 -0700)]
dynamic_debug: add jump label support
Although dynamic debug is often only used for debug builds, sometimes
its enabled for production builds as well. Minimize its impact by using
jump labels. This reduces the text section by 7000+ bytes in the kernel
image below. It does increase data, but this should only be referenced
when changing the direction of the branches, and hence usually not in
cache.
Jason Baron [Wed, 3 Aug 2016 20:46:36 +0000 (13:46 -0700)]
jump_label: remove bug.h, atomic.h dependencies for HAVE_JUMP_LABEL
The current jump_label.h includes bug.h for things such as WARN_ON().
This makes the header problematic for inclusion by kernel.h or any
headers that kernel.h includes, since bug.h includes kernel.h (circular
dependency). The inclusion of atomic.h is similarly problematic. Thus,
this should make jump_label.h 'includable' from most places.
Jason Baron [Wed, 3 Aug 2016 20:46:33 +0000 (13:46 -0700)]
arm: jump label may reference text in __exit
The jump table can reference text found in an __exit section. Thus,
instead of discarding it at build time, include EXIT_TEXT as part of
__init and it will be released when the system boots.
Chris Metcalf [Wed, 3 Aug 2016 20:46:30 +0000 (13:46 -0700)]
tile: support static_key usage in non-module __exit sections
Previously, all the __exit sections were just dropped by the link phase.
However, if there are static_key (jump label) constructs in __exit
sections that are not modules, the link fails with the message:
`.exit.text' referenced in section `__jump_table' of xxx.o:
defined in discarded section `.exit.text' of xxx.o
Support this usage by keeping the .exit.text sections in the final image
if JUMP_LABEL is defined, then discarding them once initialization is
complete.
Jason Baron [Wed, 3 Aug 2016 20:46:27 +0000 (13:46 -0700)]
sparc: support static_key usage in non-module __exit sections
The jump table can reference text found in an __exit section. Thus,
instead of discarding it at build/link time, include EXIT_TEXT as part
of __init and release it at system boot time.
Without this patch the link fails with:
`.exit.text' referenced in section `__jump_table' of xxx.o:
defined in discarded section `.exit.text' of xxx.o
The addition of jump label support in dynamic_debug caused an unexpected
warning in exactly one file in the kernel:
drivers/media/dvb-frontends/cxd2841er.c: In function 'cxd2841er_tune_tc':
include/linux/dynamic_debug.h:134:3: error: 'carrier_offset' may be used uninitialized in this function [-Werror=maybe-uninitialized]
__dynamic_dev_dbg(&descriptor, dev, fmt, \
^~~~~~~~~~~~~~~~~
drivers/media/dvb-frontends/cxd2841er.c:3177:11: note: 'carrier_offset' was declared here
int ret, carrier_offset;
^~~~~~~~~~~~~~
The problem seems to be that the compiler gets confused by the extra
conditionals in static_branch_unlikely, to the point where it can no
longer keep track of which branches have already been taken, and it
doesn't realize that this variable is now always initialized when it
gets used.
I have done lots of randconfig kernel builds and could not find any
other file with this behavior, so I assume it's a rare enough glitch
that we don't need to change the jump label support but instead just
work around the warning in the driver.
To achieve that, I'm moving the check for the return value into the
switch() statement, which is an obvious transformation, but is enough to
un-confuse the compiler here. The resulting code is not as nice to
read, but at least we retain the behavior of warning if it gets changed
to actually access an uninitialized carrier offset value in the future.
Ross Zwisler [Wed, 3 Aug 2016 20:46:15 +0000 (13:46 -0700)]
block: remove BLK_DEV_DAX config option
The functionality for block device DAX was already removed with commit acc93d30d7d4 ("Revert "block: enable dax for raw block devices"")
However, we still had a config option hanging around that was always
disabled because it depended on CONFIG_BROKEN. This config option was
introduced in commit 03cdadb04077 ("block: disable block device DAX by
default")
This change reverts that commit, removing the dead config option.
Huang Shijie [Wed, 3 Aug 2016 20:46:09 +0000 (13:46 -0700)]
samples/kretprobe: convert the printk to pr_info/pr_err
We prefer to use the pr_* to print out the log now, this patch converts
the printk to pr_info. In the error path, use the pr_err to replace the
printk.
Huang Shijie [Wed, 3 Aug 2016 20:46:06 +0000 (13:46 -0700)]
samples/jprobe: convert the printk to pr_info/pr_err
We prefer to use the pr_* to print out the log now, this patch converts
the printk to pr_info. In the error path, use the pr_err to replace the
printk.
Huang Shijie [Wed, 3 Aug 2016 20:46:03 +0000 (13:46 -0700)]
samples/kprobe: convert the printk to pr_info/pr_err
We prefer to use the pr_* to print out the log now, this patch converts
the printk to pr_info. In the error path, use the pr_err to replace the
printk.
The dma-mapping core and the implementations do not change the DMA
attributes passed by pointer. Thus the pointer can point to const data.
However the attributes do not have to be a bitfield. Instead unsigned
long will do fine:
1. This is just simpler. Both in terms of reading the code and setting
attributes. Instead of initializing local attributes on the stack
and passing pointer to it to dma_set_attr(), just set the bits.
2. It brings safeness and checking for const correctness because the
attributes are passed by value.
Semantic patches for this change (at least most of them):
Masahiro Yamada [Wed, 3 Aug 2016 20:45:50 +0000 (13:45 -0700)]
tree-wide: replace config_enabled() with IS_ENABLED()
The use of config_enabled() against config options is ambiguous. In
practical terms, config_enabled() is equivalent to IS_BUILTIN(), but the
author might have used it for the meaning of IS_ENABLED(). Using
IS_ENABLED(), IS_BUILTIN(), IS_MODULE() etc. makes the intention
clearer.
This commit replaces config_enabled() with IS_ENABLED() where possible.
This commit is only touching bool config options.
I noticed two cases where config_enabled() is used against a tristate
option:
Steve Capper [Wed, 3 Aug 2016 14:15:55 +0000 (15:15 +0100)]
arm64: Fix copy-on-write referencing in HugeTLB
set_pte_at(.) will set or unset the PTE_RDONLY hardware bit before
writing the entry to the table.
This can cause problems with the copy-on-write logic in hugetlb_cow:
*) hugetlb_cow(.) called to handle a write fault on read only pte,
*) Before the copy-on-write updates the new page table a call is
made to pte_same(huge_ptep_get(ptep), pte)), to check for a race,
*) Because set_pte_at(.) changed the pte, *ptep != pte, and the
hugetlb_cow(.) code erroneously assumes that it lost the race,
*) The new page is subsequently freed without being used.
On arm64 this problem only becomes apparent when we apply: 67961f9 mm/hugetlb: fix huge page reserve accounting for private
mappings
When one runs the libhugetlbfs test suite, there are allocation errors
and hugetlbfs pages become erroneously locked in memory as reserved.
(There is a high HugePages_Rsvd: count).
In this patch we introduce pte_same which ignores the PTE_RDONLY bit,
allowing for the libhugetlbfs test suite to pass as expected and
without leaking any reserved HugeTLB pages.
Bandan Das [Tue, 2 Aug 2016 20:32:36 +0000 (16:32 -0400)]
nvmx: mark ept single context invalidation as supported
Commit 4b855078601f ("KVM: nVMX: Don't advertise single
context invalidation for invept") removed advertising
single context invalidation since the spec does not mandate it.
However, some hypervisors (such as ESX) require it to be present
before willing to use ept in a nested environment. Advertise it
and fallback to the global case.
This can be reproduced steadily by kernel_irqchip=off.
We should not access preemption timer stuff if lapic is emulated in userspace.
This patch fix it by avoiding access preemption timer stuff when kernel_irqchip=off.
Paolo Bonzini [Thu, 4 Aug 2016 12:01:05 +0000 (14:01 +0200)]
KVM: documentation: fix KVM_CAP_X2APIC_API information
The KVM_X2APIC_API_USE_32BIT_IDS feature applies to both
KVM_SET_GSI_ROUTING and KVM_SIGNAL_MSI, but was not mentioned in the
documentation for the latter ioctl.
Paolo Bonzini [Thu, 9 Jun 2016 18:12:34 +0000 (20:12 +0200)]
x86: vdso: use __pvclock_read_cycles
The new simplified __pvclock_read_cycles does the same computation
as vread_pvclock, except that (because it takes the pvclock_vcpu_time_info
pointer) it has to be moved inside the loop. Since the loop is expected to
never roll, this makes no difference.
Paolo Bonzini [Thu, 9 Jun 2016 11:06:08 +0000 (13:06 +0200)]
pvclock: introduce seqcount-like API
The version field in struct pvclock_vcpu_time_info basically implements
a seqcount. Wrap it with the usual read_begin and read_retry functions,
and use these APIs instead of peppering the code with smp_rmb()s.
While at it, change it to the more pedantically correct virt_rmb().
With this change, __pvclock_read_cycles can be simplified noticeably.
Hui Wang [Thu, 4 Aug 2016 07:28:04 +0000 (15:28 +0800)]
ALSA: hda - Fix headset mic detection problem for two dell machines
One of the machines has ALC255 on it, another one has ALC298 on it.
On the machine with the codec ALC298, it also has the speaker volume
problem, so we add the fixup chained to ALC298_FIXUP_SPK_VOLUME rather
than adding a group of pin definition in the pin quirk table, since
the speak volume problem does not happen on other machines yet.
Ingo Molnar [Thu, 4 Aug 2016 09:02:38 +0000 (11:02 +0200)]
Merge tag 'perf-core-for-mingo-20160803' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent
Pull perf/core improvements and fixes from Arnaldo Carvalho de Melo:
New features:
- Add --sample-cpu to 'perf record', to explicitely ask for sampling
the CPU (Jiri Olsa)
Fixes:
- Fix processing of multi byte chunks in objdump output, fixing
disassemble processing for annotation on at least ARM64 (Jan Stancek)
- Use SyS_epoll_wait in a BPF 'perf test' entry instead of sys_epoll_wait, that
is not present in the DWARF info in vmlinux files (Arnaldo Carvalho de Melo)
- Add -wno-shadow when processing files using perl headers, fixing
the build on Fedora Rawhide and Arch Linux (Namhyung Kim)
Infrastructure changes:
- Annotate prep work to better catch and report errors related to
using objdump to disassemble DSOs (Arnaldo Carvalho de Melo)
- Add 'alloc', 'scnprintf' and 'and' methods for bitmap processing (Jiri Olsa)
Alex Vesker [Wed, 6 Jul 2016 13:36:35 +0000 (16:36 +0300)]
IB/core: Support for CMA multicast join flags
Added UCMA and CMA support for multicast join flags. Flags are
passed using UCMA CM join command previously reserved fields.
Currently supporting two join flags indicating two different
multicast JoinStates:
1. Full Member:
The initiator creates the Multicast group(MCG) if it wasn't
previously created, can send Multicast messages to the group
and receive messages from the MCG.
2. Send Only Full Member:
The initiator creates the Multicast group(MCG) if it wasn't
previously created, can send Multicast messages to the group
but doesn't receive any messages from the MCG.
IB: Send Only Full Member requires a query of ClassPortInfo
to determine if SM/SA supports this option. If SM/SA
doesn't support Send-Only there will be no join request
sent and an error will be returned.
ETH: When Send Only Full Member is requested no IGMP join
will be sent.
Alex Vesker [Wed, 6 Jul 2016 13:36:34 +0000 (16:36 +0300)]
IB/sa: Add cached attribute containing SM information to SA port
Added a new SA port attribute containing SM ClassPortInfo fields,
(ClassPortInfo fields: Table 126 IB Spec 1.3.). This is useful for
checking SM support for specific features. The attribute is cached
to avoid resending queries, caching is done when a successful
ClassPortInfo reply is received on the port. Invalidation of the
attribute is done on SM change events, SM re-registration events,
and SM LID change events. The fields in ClassPortInfo should not
change during SM runtime without an event.
Jason Gunthorpe [Sun, 3 Jul 2016 12:28:18 +0000 (15:28 +0300)]
IB/uverbs: Fix race between uverbs_close and remove_one
Fixes an oops that might happen if uverbs_close races with
remove_one.
Both contexts may run ib_uverbs_cleanup_ucontext, it depends
on the flow.
Currently, there is no protection for a case that remove_one
didn't make the cleanup it runs to its end, the underlying
ib_device was freed then uverbs_close will call
ib_uverbs_cleanup_ucontext and OOPs.
Above might happen if uverbs_close deleted the file from the list
then remove_one didn't find it and runs to its end.
Fixes to protect against that case by a new cleanup lock so that
ib_uverbs_cleanup_ucontext will be called always before that
remove_one is ended.
Markus Elfring [Sat, 23 Jul 2016 09:28:30 +0000 (11:28 +0200)]
IB/mthca: Clean up error unwind flow in mthca_reset()
The kfree() function was called in a few cases by the mthca_reset()
function during error handling even if the passed variables "bridge_header"
and "hca_header" contained a null pointer.
Adjust jump targets according to the Linux coding style convention.
Mark Bloch [Tue, 19 Jul 2016 17:54:58 +0000 (20:54 +0300)]
IB/mlx4: Add diagnostic hardware counters
Expose IB diagnostic hardware counters.
The counters count IB events and are applicable for IB and RoCE.
The counters can be divided into two groups, per device and per port.
Device counters are always exposed.
Port counters are exposed only if the firmware supports per port counters.
rq_num_dup and sq_num_to are only exposed if we have firmware support
for them, if we do, we expose them per device and per port.
rq_num_udsdprd and num_cqovf are device only counters.
rq - denotes responder.
sq - denotes requester.
|-----------------------|---------------------------------------|
| Name | Description |
|-----------------------|---------------------------------------|
|rq_num_lle | Number of local length errors |
|-----------------------|---------------------------------------|
|sq_num_lle | number of local length errors |
|-----------------------|---------------------------------------|
|rq_num_lqpoe | Number of local QP operation errors |
|-----------------------|---------------------------------------|
|sq_num_lqpoe | Number of local QP operation errors |
|-----------------------|---------------------------------------|
|rq_num_lpe | Number of local protection errors |
|-----------------------|---------------------------------------|
|sq_num_lpe | Number of local protection errors |
|-----------------------|---------------------------------------|
|rq_num_wrfe | Number of CQEs with error |
|-----------------------|---------------------------------------|
|sq_num_wrfe | Number of CQEs with error |
|-----------------------|---------------------------------------|
|sq_num_mwbe | Number of Memory Window bind errors |
|-----------------------|---------------------------------------|
|sq_num_bre | Number of bad response errors |
|-----------------------|---------------------------------------|
|sq_num_rire | Number of Remote Invalid request |
| | errors |
|-----------------------|---------------------------------------|
|rq_num_rire | Number of Remote Invalid request |
| | errors |
|-----------------------|---------------------------------------|
|sq_num_rae | Number of remote access errors |
|-----------------------|---------------------------------------|
|rq_num_rae | Number of remote access errors |
|-----------------------|---------------------------------------|
|sq_num_roe | Number of remote operation errors |
|-----------------------|---------------------------------------|
|sq_num_tree | Number of transport retries exceeded |
| | errors |
|-----------------------|---------------------------------------|
|sq_num_rree | Number of RNR NAK retries exceeded |
| | errors |
|-----------------------|---------------------------------------|
|rq_num_rnr | Number of RNR NAKs sent |
|-----------------------|---------------------------------------|
|sq_num_rnr | Number of RNR NAKs received |
|-----------------------|---------------------------------------|
|rq_num_oos | Number of Out of Sequence requests |
| | received |
|-----------------------|---------------------------------------|
|sq_num_oos | Number of Out of Sequence NAKs |
| | received |
|-----------------------|---------------------------------------|
|rq_num_udsdprd | Number of UD packets silently |
| | discarded on the Receive Queue due to |
| | lack of receive descriptor |
|-----------------------|---------------------------------------|
|rq_num_dup | Number of duplicate requests received |
|-----------------------|---------------------------------------|
|sq_num_to | Number of time out received |
|-----------------------|---------------------------------------|
|num_cqovf | Number of CQ overflows |
|-----------------------|---------------------------------------|
Mustafa Ismail [Mon, 18 Jul 2016 19:21:49 +0000 (14:21 -0500)]
Use smaller 512 byte messages for portmapper messages
Portmapper messages are short and do not occupy more than 512 bytes.
Lower portmapper message size to 512 bytes. This change significantly
reduces the amount of memory needed when trying to establish a large
number of connections simultaneously. The old value is based on page
size.
IB/ipoib: Report SG feature regardless of HW UD CSUM capability
Decouple SG support from HW ability to do UD checksum.
This coupling is for historical reasons and removed with 'commit ec5f06156423 ("net: Kill link between CSUM and SG features.")'
During driver load it is assumed that device does not supports SG. The
final decision is taken after creating UD QP based on device capability.
Roland Dreier [Fri, 29 Jul 2016 04:58:43 +0000 (21:58 -0700)]
IB/mlx4: Don't use GFP_ATOMIC for CQ resize struct
We allocate a small tracking structure as part of mlx4_ib_resize_cq().
However, we don't need to use GFP_ATOMIC -- immediately after the
allocation, we call mlx4_cq_resize(), which allocates a command
mailbox with GFP_KERNEL and then sleeps on a firmware command, so we
better not be in an atomic context.
This actually has a real impact, because when this GFP_ATOMIC
allocation fails (and GFP_ATOMIC does fail in practice) then a
userspace consumer resizing a CQ will get a spurious failure that we
can easily avoid.
Paul Gortmaker [Thu, 28 Jul 2016 03:11:47 +0000 (23:11 -0400)]
extable.h: add stddef.h so "NULL" definition is not implicit
While not an issue now, eventually we will have independent users of
the extable.h file and we will stop sourcing it via module.h header.
In testing that pending work, with very sparse builds, characteristic
of an "allnoconfig" on various architectures, we can sometimes hit an
instance where the very basic standard definitions aren't present,
resulting in:
include/linux/extable.h:26:9: error: 'NULL' undeclared (first use in this function)
To be clear, this isn't a regression, since currently extable.h is
only used by module.h -- however, we will need this addition present
before we start migrating exception table users off module.h and onto
extable.h during the next release cycle.
Add ro_after_init support for modules by adding a new page-aligned section
in the module layout (after rodata) for ro_after_init data and enabling RO
protection for that section after module init runs.
Rusty Russell [Wed, 27 Jul 2016 02:47:35 +0000 (12:17 +0930)]
jump_label: disable preemption around __module_text_address().
Steven reported a warning caused by not holding module_mutex or
rcu_read_lock_sched: his backtrace was corrupted but a quick audit
found this possible cause. It's wrong anyway...
Paul Gortmaker [Wed, 27 Jul 2016 02:36:34 +0000 (12:06 +0930)]
exceptions: fork exception table content from module.h into extable.h
For historical reasons (i.e. pre-git) the exception table stuff was
buried in the middle of the module.h file. I noticed this while
doing an audit for needless includes of module.h and found core
kernel files (both arch specific and arch independent) were just
including module.h for this.
The converse is also true, in that conventional drivers, be they
for filesystems or actual hardware peripherals or similar, do not
normally care about the exception tables.
Here we fork the exception table content out of module.h into a
new file called extable.h -- and temporarily include it into the
module.h itself.
Then we will work our way across the arch independent and arch
specific files needing just exception table content, and move
them off module.h and onto extable.h
Once that is done, we can remove the extable.h from module.h
and in doing it like this, we avoid introducing build failures
into the git history.
The gain here is that module.h gets a bit smaller, across all
modular drivers that we build for allmodconfig. Also the core
files that only need exception table stuff don't have an include
of module.h that brings in lots of extra stuff and just looks
generally out of place.
modules: Add kernel parameter to blacklist modules
Blacklisting a module in linux has long been a problem. The current
procedure is to use rd.blacklist=module_name, however, that doesn't
cover the case after the initramfs and before a boot prompt (where one
is supposed to use /etc/modprobe.d/blacklist.conf to blacklist
runtime loading). Using rd.shell to get an early prompt is hit-or-miss,
and doesn't cover all situations AFAICT.
This patch adds this functionality of permanently blacklisting a module
by its name via the kernel parameter module_blacklist=module_name.
[v2]: Rusty, use core_param() instead of __setup() which simplifies
things.
Steven Rostedt [Mon, 18 Jul 2016 20:29:24 +0000 (05:59 +0930)]
module: Do a WARN_ON_ONCE() for assert module mutex not held
When running with lockdep enabled, I triggered the WARN_ON() in the
module code that asserts when module_mutex or rcu_read_lock_sched are
not held. The issue I have is that this can also be called from the
dump_stack() code, causing us to enter an infinite loop...
In this commit, using system workqueue causes that the maximum parallel
executions of _Qxx can exceed 255. This violates the method reentrancy
limit in ACPICA and generates the following error log:
ACPI Error: Method reached maximum reentrancy limit (255) (20150818/dsmethod-341)
This patch creates a seperate workqueue and limits the number of parallel
_Qxx evaluations down to a configurable value (can be tuned against number
of online CPUs).
Since EC events are handled after driver probe, we can create the workqueue
in acpi_ec_init().
Something made the sys_epoll_wait() function alias not to be found in
the vmlinux DWARF info, being found only in /proc/kallsyms, which made
the BPF perf tests to fail:
[root@jouet ~]# perf test BPF
37: Test BPF filter :
37.1: Test basic BPF filtering : FAILED!
37.2: Test BPF prologue generation : Skip
37.3: Test BPF relocation checker : Skip
[root@jouet ~]#
Using -v we can see it is failing to find DWARF info for the probed function,
sys_epoll_wait, which we can find in /proc/kallsyms but not in vmlinux with
CONFIG_DEBUG_INFO:
[root@jouet ~]# perf probe sys_epoll_wait
Failed to find debug information for address ffffffffbd295b50
Probe point 'sys_epoll_wait' not found.
Error: Failed to add events.
[root@jouet ~]#
It all works if we use SyS_epoll_wait, that is just an alias to the probed
function:
[root@jouet ~]# grep -i sys_epoll_wait /proc/kallsyms ffffffffbd295b50 T SyS_epoll_wait ffffffffbd295b50 T sys_epoll_wait
[root@jouet ~]#
So use it:
[root@jouet ~]# perf test BPF
37: Test BPF filter :
37.1: Test basic BPF filtering : Ok
37.2: Test BPF prologue generation : Ok
37.3: Test BPF relocation checker : Ok
[root@jouet ~]#
Investigation as to why it fails is still underway, but it was always
going from sys_epoll_wait to SyS_epoll_wait when looking up the DWARF
info in vmlinux, and this is what is breaking now.
Switching to use SyS_epoll_wait allows this test to proceed and test the
BPF code it was designed for, so lets have this in to allow passing this
test while we fix the root cause.
Vegard Nossum [Sat, 21 May 2016 15:46:10 +0000 (17:46 +0200)]
um: Support kcov
This adds support for kcov to UML.
There is a small problem where UML will randomly segfault during boot;
this is because current_thread_info() occasionally returns an invalid
(non-NULL) pointer and we try to dereference it in
__sanitizer_cov_trace_pc(). I consider this a bug in UML itself and this
patch merely exposes it.
[v2: disable instrumentation in UML-specific code]
Linus Torvalds [Wed, 3 Aug 2016 22:11:24 +0000 (18:11 -0400)]
drm: i915: fix build when DEBUG_FS is disabled
This clearly had never gotten tested, probably because you need a fairly
minimal configuration in order to disable DEBUG_FS (several other
options select it).
The dummy inline functions that were used for the no-DEBUG_FS case were
missing the argument names in the declarations.
Currently UML sets up physical memory very early,
long before setup_arch() was called by the kernel main
function.
This can cause problems when code paths in UML's memory setup
code assume that the kernel is already running.
i.e. when kmemleak is enabled it will evaluate current()
in free_bootmem(). That early current() is undefined and
UML explodes.
Solve the problem by setting up physical memory in setup_arch(),
at this stage the kernel has materialized and basic infrastructure
such as current() works.
David Howells [Wed, 3 Aug 2016 16:57:39 +0000 (17:57 +0100)]
cachefiles: Fix race between inactivating and culling a cache object
There's a race between cachefiles_mark_object_inactive() and
cachefiles_cull():
(1) cachefiles_cull() can't delete a backing file until the cache object
is marked inactive, but as soon as that's the case it's fair game.
(2) cachefiles_mark_object_inactive() marks the object as being inactive
and *only then* reads the i_blocks on the backing inode - but
cachefiles_cull() might've managed to delete it by this point.
Fix this by making sure cachefiles_mark_object_inactive() gets any data it
needs from the backing inode before deactivating the object.
Linus Torvalds [Wed, 3 Aug 2016 16:50:06 +0000 (12:50 -0400)]
Merge tag 'trace-v4.8-1' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace
Pull tracing fixes from Steven Rostedt:
"A few updates and fixes:
- move the suppressing of the __builtin_return_address >0 warning to
the tracing directory only.
- metag recordmcount fix for newer glibc's
- two tracing histogram fixes that were reported by KASAN"
* tag 'trace-v4.8-1' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
tracing: Fix use-after-free in hist_register_trigger()
tracing: Fix use-after-free in hist_unreg_all/hist_enable_unreg_all
Makefile: Mute warning for __builtin_return_address(>0) for tracing only
ftrace/recordmcount: Work around for addition of metag magic but not relocations
dm raid: constructor fails on non-zero incompat_features
When lvm2 userspace requests a RaidLV repair, it sets the rebuild
constructor flag on the new replacement DataLVs but does not clear the
respective MetaLVs. Hence the superblock that is loaded from such new
MetaLVs may have a non-zero incompat_features member and the constructor
will fail with false-positive on incompat_features.
Solve by initializing the incompat_features member properly.
projects / linux.git / log