Files created under macOS cannot be opened under linux if their names
contain Korean characters, and vice versa.
The Korean alphabet is special because its normalization is done without a
table. The module deals with it correctly when composing, but forgets
about it for the decomposition.
Fix this using the Hangul decomposition function provided in the Unicode
Standard. The code fits a bit awkwardly because it requires a buffer,
while all the other normalizations are returned as pointers to the
decomposition table. This is actually also a bug because reordering may
still be needed, but for now leave it as it is.
The patch will cause trouble for Hangul filenames already created by the
module in the past. This shouldn't really be concern because its main
purpose was always sharing with macOS. If a user actually needs to access
such a file the nodecompose mount option should be enough.
Link: http://lkml.kernel.org/r/20180717220951.p6qqrgautc4pxvzu@eaf
Signed-off-by: Ernesto A. Fernández <[email protected]>
Reported-by: Ting-Chang Hou <[email protected]>
Tested-by: Ting-Chang Hou <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
After an extent is removed from the extent tree, the corresponding bits
are also cleared from the block allocation file. This is currently done
without releasing the tree lock.
The problem is that the allocation file has extents of its own; if it is
fragmented enough, some of them may be in the extent tree as well, and
hfsplus_get_block() will try to take the lock again.
To avoid deadlock, only hold the extent tree lock during the actual tree
operations.
Link: http://lkml.kernel.org/r/20180709202549.auxwkb6memlegb4a@eaf
Signed-off-by: Ernesto A. Fernández <[email protected]>
Reported-by: Anatoly Trosinenko <[email protected]>
Cc: Viacheslav Dubeyko <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
The userspace automount(8) daemon is meant to perform a forced expire when
sent a SIGUSR2.
But since the expiration is routed through the kernel and the kernel
doesn't send an expire request if the mount is busy this hasn't worked at
least since autofs version 5.
Add an AUTOFS_EXP_FORCED flag to allow implemention of the feature and
bump the protocol version so user space can check if it's implemented if
needed.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ian Kent <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Make the usage of the expire flags consistent by naming the expire flags
the same as it is named in the version 5 miscelaneous ioctl parameters and
only check the bit flags when needed.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ian Kent <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
The expire flag AUTOFS_EXP_LEAVES is cleared before the second call to
should_expire() in autofs_expire_indirect() but the parameter passed in
the second call is incorrect.
Fortunately AUTOFS_EXP_LEAVES expire flag has not been used for a long
time but might be needed in the future so fix it rather than remove the
expire leaves functionality.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ian Kent <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
The global variable "now" in fs/autofs/expire.c is used in an inconsistent
way, sometimes using jiffies directly, and sometimes using the "now"
variable, and setting it isn't done consistently either.
But the autofs dentry info last_used field is only updated during path
walks or during expire so jiffies can be used directly and the global
variable "now" removed.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ian Kent <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Depending on how it is configured the autofs user space daemon can leave
in use mounts mounted at exit and re-connect to them at start up. But for
this to work best the state of the autofs file system needs to be left
intact over the restart.
Also, at system shutdown, mounts in an autofs file system might be
umounted exposing a mount point trigger for which subsequent access can
lead to a hang. So recent versions of automount(8) now does its best to
set autofs file system mounts catatonic at shutdown.
When autofs file system mounts are catatonic it's currently possible to
create and remove directories and symlinks which can be a problem at
restart, as described above.
So return EACCES in the directory, symlink and unlink methods if the
autofs file system is catatonic.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ian Kent <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Add a log message to `run_init_process()`.
This log message serves two purposes.
1. If the init process is not specified on the Linux Kernel command
line, the user sees, what file was chosen.
2. The time stamps shows exactly, when the Linux kernel handed over
control to the init process.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Paul Menzel <[email protected]>
Reviewed-by: Andrew Morton <[email protected]>
Cc: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Sparse checking used to be disabled on init/do_mounts.c and a few related
files because "Many of the syscalls used in this file expect some of the
arguments to be __user pointers not __kernel pointers".
However since 28128c61e ("kconfig.h: Include compiler types to avoid
missed struct attributes") the checks are, in fact, not disabled anymore
because of the more early include of "linux/compiler_types.h"
So remove the now ineffective #undefery that was done to disable these
warnings, as well as the associated comment.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Luc Van Oostenryck <[email protected]>
Cc: Dominik Brodowski <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Kees Cook <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Similar to other calls, ep_poll() is not called with interrupts disabled,
and we can therefore avoid the irq save/restore dance and just disable
local irqs. In fact, the call should never be called in irq context at
all, considering that the only path is
epoll_wait(2) -> do_epoll_wait() -> ep_poll().
When running on a 2 socket 40-core (ht) IvyBridge a common pipe based
epoll_wait(2) microbenchmark, the following performance improvements are
seen:
# threads vanilla dirty
1 1805587 2106412
2 1854064 2090762
4 1805484 2017436
8 1751222 1974475
16 1725299 1962104
32 1378463 1571233
64 787368 900784
Which is a pretty constantly near 15%.
Also add a lockdep check such that we detect any mischief before
deadlocking.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Reviewed-by: Andrew Morton <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Jason Baron <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
On Sun, 2018-08-05 at 08:52 -0700, Linus Torvalds wrote:
> "long unsigned int" isn't _technically_ wrong. But we normally
> call that type "unsigned long".
So add a checkpatch test for it.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Joe Perches <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Current checkpatch implementation permits notation like
} else{
in kernel code. It looks like oversight and inconsistency in checkpatch
rules (e.g. instruction like 'do' is tested).
Add regex for checking space after 'else' keyword and trigger error if
space is not present.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Michal Zylowski <[email protected]>
Acked-by: Joe Perches <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
checkpatch uses the in-kernel script spdxcheck.py to validate the specific
license in a file or script.
This check can currently fail for a couple reasons:
o spdxcheck.py assumes the existence of git tree that may not
exist for a bare source tree from something like a tarball
o the spdxcheck.py must be run from the top level root directory
So add a git existence test and set the subprocess subdirectory.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Joe Perches <[email protected]>
Reported-by: Charlemagne Lasse <[email protected]>
Tested-by: Charlemagne Lasse <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
The current krealloc test does not function correctly when the temporary
pointer return name contains the original pointer name.
Fix that by maximally matching the return pointer name and the original
pointer name and doing a separate comparison of the both names.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Joe Perches <[email protected]>
Reported-by: Lars-Peter Clausen <[email protected]>
Cc: Manish Narani <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Multiple line macro definitions where the arguments are separated by line
continuations can cause checkpatch to emit invalid syntax regex tests.
This can occur when a single argument is modified in a part of a patch.
For example: (to not add a diff in the commit message)
$ ./scripts/checkpatch.pl --git db023296f0
Unterminated \g... pattern in regex; <very long regex omitted>
And, the test does not work correctly when these arguments are all new as
the initial patch line addition "+" is used in the argument name.
Fix this by stripping the line continuations and any "+" from the list of
arguments.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Joe Perches <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
checkpatch repeatedly uses a runtime minimum version check that validates
the minimum perl version required for a regex match by using a "$^V ge
5.10.0" runtime string match.
Only perform that minimum version test once and store the result to reduce
string matching time.
This reduces runtime execution time for patches or files with high line
counts.
An example runtime improvement:
new: $ time ./scripts/checkpatch.pl -f drivers/net/ethernet/intel/i40e/i40e_main.c > /dev/null
real 0m11.856s
user 0m11.831s
sys 0m0.025s
old: $ time ./scripts/checkpatch.pl -f drivers/net/ethernet/intel/i40e/i40e_main.c > /dev/null
real 0m13.330s
user 0m13.282s
sys 0m0.049s
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Joe Perches <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
A struct with a bool member can have different sizes on various
architectures because neither bool size nor alignment is standardized.
So emit a message on the use of bool in structs only in .h files and not
.c files.
There is the real possibility that this test could have a false positive
when a bool is declared as an automatic, so limit the test to .h files
where the only false positive is for declarations in static inline
functions.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Joe Perches <[email protected]>
Suggested-by: Peter Zijlstra <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Patch series "add crc64 calculation as kernel library", v5.
This patchset adds basic implementation of crc64 calculation as a Linux
kernel library. Since bcache already does crc64 by itself, this patchset
also modifies bcache code to use the new crc64 library routine.
Currently bcache is the only user of crc64 calculation, another potential
user is bcachefs which is on the way to be in mainline kernel. Therefore
it makes sense to make crc64 calculation to be a public library.
bcache uses crc64 as storage checksum, if a change of crc lib routines
results an inconsistent result, the unmatched checksum may make bcache
'think' the on-disk is corrupted, such a change should be avoided or
detected as early as possible. Therefore a patch is being prepared which
adds a crc test framework, to check consistency of different calculations.
This patch (of 2):
Add the re-write crc64 calculation routines for Linux kernel. The CRC64
polynomical arithmetic follows ECMA-182 specification, inspired by CRC
paper of Dr. Ross N. Williams (see
http://www.ross.net/crc/download/crc_v3.txt) and other public domain
implementations.
All the changes work in this way,
- When Linux kernel is built, host program lib/gen_crc64table.c will be
compiled to lib/gen_crc64table and executed.
- The output of gen_crc64table execution is an array called as lookup
table (a.k.a POLY 0x42f0e1eba9ea369) which contain 256 64-bit long
numbers, this table is dumped into header file lib/crc64table.h.
- Then the header file is included by lib/crc64.c for normal 64bit crc
calculation.
- Function declaration of the crc64 calculation routines is placed in
include/linux/crc64.h
Currently bcache is the only user of crc64_be(), another potential user is
bcachefs which is on the way to be in mainline kernel. Therefore it makes
sense to move crc64 calculation into lib/crc64.c as public code.
[[email protected]: fix review comments from v4]
Link: http://lkml.kernel.org/r/[email protected]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Coly Li <[email protected]>
Co-developed-by: Andy Shevchenko <[email protected]>
Signed-off-by: Andy Shevchenko <[email protected]>
Reviewed-by: Hannes Reinecke <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: Andy Shevchenko <[email protected]>
Cc: Michael Lyle <[email protected]>
Cc: Kent Overstreet <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Kate Stewart <[email protected]>
Cc: Eric Biggers <[email protected]>
Cc: Randy Dunlap <[email protected]>
Cc: Noah Massey <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
The pointer foo is local to the source and does not need to be
in global scope, so make it static.
Cleans up sparse warning:
symbol 'foo' was not declared. Should it be static?
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Colin Ian King <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
There is an external use case for multiple private MAINTAINER style files
in a separate directory. Allow it.
--mpath has a default of "./MAINTAINERS".
The value entered can be either a file or a directory.
The behaviors are now:
--mpath <file> Read only the specific file as <MAINTAINER_TYPE> file
--mpath <directory> Read all files in <directory> as <MAINTAINER_TYPE> files
--mpath <directory> --find-maintainer-files
Recurse through <directory> and read all files named MAINTAINERS
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Joe Perches <[email protected]>
Tested-by: Don Zickus <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Add option '--no-tree' to get_maintainer.pl script to allow using this
script in projects that aren't the Linux kernel if they use the same
format for their MAINTAINERS file. This command is also available in
checkpatch.pl, for example.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Antonio Nino Diaz <[email protected]>
Cc: Joe Perches <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Both functions are similar to the context of ep_modify(), called via
epoll_ctl(2). Just like ep_modify(), saving and restoring interrupts is
an overkill in these calls as it will never be called with irqs disabled.
While ep_remove() can be called directly from EPOLL_CTL_DEL, it can also
be called when releasing the file, but this also complies with the above.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Cc: Jason Baron <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Patch series "fs/epoll: loosen irq safety when possible".
Both patches replace saving+restoring interrupts when taking the ep->lock
(now the waitqueue lock), with just disabling local irqs. This shows
immediate performance benefits in patch 1 for an epoll workload running on
Xen. The main concern we need to have with this sort of changes in epoll
is the ep_poll_callback() which is passed to the wait queue wakeup and is
done very often under irq context, this patch does not touch this call.
Patches have been tested pretty heavily with the customer workload,
microbenchmarks, ltp testcases and two high level workloads that use epoll
under the hood: nginx and libevent benchmarks.
This patch (of 2):
Saving and restoring interrupts in ep_scan_ready_list() is an
overkill as it is never called with interrupts disabled. Loosen
this to simply disabling local irqs such that archs where managing
irqs is expensive or virtual environments. This patch yields
some throughput improvements on a workload that is epoll intensive
running on a single Xen DomU.
1 Job 7500 --> 8800 enq/s (+17%)
2 Jobs 14000 --> 15200 enq/s (+8%)
3 Jobs 20500 --> 22300 enq/s (+8%)
4 Jobs 25000 --> 28000 enq/s (+8-12)%
On bare metal:
For a 2-socket 40-core (ht) IvyBridge on a few workloads, unfortunately I
don't have a xen environment and the results for Xen I do have (which
numbers are in patch 1) I don't have the actual workload, so cannot
compare them directly.
1) Different configurations were used for a epoll_wait (pipes io)
microbench (http://linux-scalability.org/epoll/epoll-test.c) and shows
around a 7-10% improvement in overall total number of times the
epoll_wait() loops when using both regular and nested epolls, so very
raw numbers, but measurable nonetheless.
# threads vanilla dirty
1 1677717 1805587
2 1660510 1854064
4 1610184 1805484
8 1577696 1751222
16 1568837 1725299
32 1291532 1378463
64 752584 787368
Note that stddev is pretty small.
2) Another pipe test, which shows no real measurable improvement.
(http://www.xmailserver.org/linux-patches/pipetest.c)
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Davidlohr Bueso <[email protected]>
Cc: Jason Baron <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
The userfaultfd code currently uses the unlocked waitqueue helpers for
managing fault_wqh, but instead of holding the waitqueue lock for this
waitqueue around these calls, it the waitqueue lock of
fault_pending_wq, which is a different waitqueue instance. Given that
the waitqueue is not exposed to the rest of the kernel this actually
works ok at the moment, but prevents the userfaultfd locking rules from
being enforced using lockdep.
Switch to the internally locked waitqueue helpers instead. This means
that the lock inside fault_wqh now nests inside the fault_pending_wqh
lock, but that's not a problem since it was entirely unused before.
[[email protected]: slight changelog updates]
[[email protected]: spotted changelog spellos]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Matthew Wilcox <[email protected]>
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Mike Rapoport <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Andrea Arcangeli <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Jason Baron <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Davidlohr Bueso <[email protected]>
Cc: Matthew Wilcox <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Patch series "waitqueue lockdep annotation", v3.
This series adds a strategic lockdep_assert_held to __wake_up_common to
ensure callers really do hold the wait_queue_head lock when calling the
unlocked wake_up variants. It turns out epoll did not do this for a
fairly common path (hit all the time by systemd during bootup), so the
second patch fixed this instance as well.
This patch (of 3):
The epoll code currently uses the unlocked waitqueue helpers for managing
ep->wq, but instead of holding the waitqueue lock around these calls, it
uses its own ep->lock spinlock. Given that the waitqueue is not exposed
to the rest of the kernel this actually works ok at the moment, but
prevents the epoll locking rules from being enforced using lockdep.
Remove ep->lock and use the waitqueue lock to not only reduce the size of
struct eventpoll but also to make sure we can assert locking invariants in
the waitqueue code.
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Jason Baron <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Andrea Arcangeli <[email protected]>
Cc: Mike Rapoport <[email protected]>
Cc: Jason Baron <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Matthew Wilcox <[email protected]>
Cc: Davidlohr Bueso <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
