The IO_REPARSE_TAG_LX_ tags originally were used by WSL but they
are preferred by the Linux client in some cases since, unlike
the NFS reparse tag (or EAs), they don't require an extra query
to determine which type of special file they represent.
Add support for readdir to recognize special file types of
FIFO, SOCKET, CHAR, BLOCK and SYMLINK. This can be tested
by creating these special files in WSL Linux and then
sharing that location on the Windows server and mounting
to the Windows server to access them.
Prior to this patch all of the special files would show up
as being of type 'file' but with this patch they can be seen
with the correct file type as can be seen below:
brwxr-xr-x 1 root root 0, 0 Oct 21 17:10 block
crwxr-xr-x 1 root root 0, 0 Oct 21 17:46 char
drwxr-xr-x 2 root root 0 Oct 21 18:27 dir
prwxr-xr-x 1 root root 0 Oct 21 16:21 fifo
-rwxr-xr-x 1 root root 0 Oct 21 15:48 file
lrwxr-xr-x 1 root root 0 Oct 21 15:52 symlink-to-file
TODO: go through all documented reparse tags to see if we can
reasonably map some of them to directories vs. files vs. symlinks
and also add support for device numbers for block and char
devices.
Signed-off-by: Steve French <stfrench@microsoft.com>
Reviewed-by: Aurelien Aptel <aaptel@suse.com>
The "end" pointer is either NULL or it points to the next byte to parse.
If there isn't a next byte then dereferencing "end" is an off-by-one out
of bounds error. And, of course, if it's NULL that leads to an Oops.
Printing "*end" doesn't seem very useful so let's delete this code.
Also for the last debug statement, I noticed that it should be printing
"sequence_end" instead of "end" so fix that as well.
Reported-by: Dominik Maier <dmaier@sect.tu-berlin.de>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
This and related patches which move mount related
code to fs_context.c has the advantage of
shriking the code in fs/cifs/connect.c (which had
the second most lines of code of any of the files
in cifs.ko and was getting harder to read due
to its size) and will also make it easier to
switch over to the new mount API in the future.
Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
Reviewed-by: Aurelien Aptel <aaptel@suse.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Helps to shrink connect.c and make it more readable
by moving mount related code to fs_context.c and
fs_context.h
Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Reviewed-by: Aurelien Aptel <aaptel@suse.com>
This patch moves the parsing of security mount options into
fs_context.ch. There are no changes to any logic.
Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Reviewed-by: Aurelien Aptel <aaptel@suse.com>
This will make it easier in the future, but also will allow us to
shrink connect.c which is getting too big, and harder to read
Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Reviewed-by: Aurelien Aptel <aaptel@suse.com>
To servers which do not support directory leases (e.g. Samba)
it is wasteful to try to open_shroot (ie attempt to cache the
root directory handle). Skip attempt to open_shroot when
server does not indicate support for directory leases.
Cuts the number of requests on mount from 17 to 15, and
cuts the number of requests on stat of the root directory
from 4 to 3.
Signed-off-by: Steve French <stfrench@microsoft.com>
CC: Stable <stable@vger.kernel.org> # v5.1+
When mounting with modefromsid mount option, it was possible to
get the error on stat of a fifo or char or block device:
"cannot stat <filename>: Operation not supported"
Special devices can be stored as reparse points by some servers
(e.g. Windows NFS server and when using the SMB3.1.1 POSIX
Extensions) but when the modefromsid mount option is used
the client attempts to get the ACL for the file which requires
opening with OPEN_REPARSE_POINT create option.
Signed-off-by: Steve French <stfrench@microsoft.com>
CC: Stable <stable@vger.kernel.org>
Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com>
Reviewed-by: Shyam Prasad N <sprasad@microsoft.com>
Can be helpful in debugging mount and reconnect issues
Signed-off-by: Samuel Cabrero <scabrero@suse.de>
Reviewed-by: Shyam Prasad N <sprasad@microsoft.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
TCP server info field server->total_read is modified in parallel by
demultiplex thread and decrypt offload worker thread. server->total_read
is used in calculation to discard the remaining data of PDU which is
not read into memory.
Because of parallel modification, server->total_read can get corrupted
and can result in discarding the valid data of next PDU.
Signed-off-by: Rohith Surabattula <rohiths@microsoft.com>
Reviewed-by: Aurelien Aptel <aaptel@suse.com>
Reviewed-by: Pavel Shilovsky <pshilov@microsoft.com>
CC: Stable <stable@vger.kernel.org> #5.4+
Signed-off-by: Steve French <stfrench@microsoft.com>
Don't populate const array smb3_create_tag_posix on the stack but
instead make it static. Makes the object code smaller by 50 bytes.
Before:
text data bss dec hex filename
150184 47167 0 197351 302e7 fs/cifs/smb2pdu.o
After:
text data bss dec hex filename
150070 47231 0 197301 302b5 fs/cifs/smb2pdu.o
(gcc version 10.2.0)
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Aurelien Aptel <aaptel@suse.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
We were setting the uid/gid to the default in each dir entry
in the parsing of the POSIX query dir response, rather
than attempting to map the user and group SIDs returned by
the server to well known SIDs (or upcall if not found).
CC: Stable <stable@vger.kernel.org>
Reviewed-by: Aurelien Aptel <aaptel@suse.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
SMB3 crediting is used for flow control, and it can be useful to
trace for problem determination how many credits were acquired
and for which operation.
Here is an example ("trace-cmd record -e *add_credits"):
cifsd-9522 [010] .... 5995.202712: smb3_add_credits:
server=localhost current_mid=0x12 credits=373 credits_to_add=10
cifsd-9522 [010] .... 5995.204040: smb3_add_credits:
server=localhost current_mid=0x15 credits=400 credits_to_add=30
Reviewed-by: Aurelien Aptel <aaptel@suse.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
There are cases where the server can return a cipher type of 0 and
it not be an error. For example server supported no encryption types
(e.g. server completely disabled encryption), or the server and
client didn't support any encryption types in common (e.g. if a
server only supported AES256_CCM). In those cases encryption would
not be supported, but that can be ok if the client did not require
encryption on mount and it should not return an error.
In the case in which mount requested encryption ("seal" on mount)
then checks later on during tree connection will return the proper
rc, but if seal was not requested by client, since server is allowed
to return 0 to indicate no supported cipher, we should not fail mount.
Reported-by: Pavel Shilovsky <pshilov@microsoft.com>
Reviewed-by: Pavel Shilovsky <pshilov@microsoft.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
In crypt_message, when smb2_get_enc_key returns error, we need to
return the error back to the caller. If not, we end up processing
the message further, causing a kernel oops due to unwarranted access
of memory.
Call Trace:
smb3_receive_transform+0x120/0x870 [cifs]
cifs_demultiplex_thread+0xb53/0xc20 [cifs]
? cifs_handle_standard+0x190/0x190 [cifs]
kthread+0x116/0x130
? kthread_park+0x80/0x80
ret_from_fork+0x1f/0x30
Signed-off-by: Shyam Prasad N <sprasad@microsoft.com>
Reviewed-by: Pavel Shilovsky <pshilov@microsoft.com>
Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com>
CC: Stable <stable@vger.kernel.org>
Signed-off-by: Steve French <stfrench@microsoft.com>
Now that 256 bit encryption can be negotiated, update
names of the nonces to match the updated official protocol
documentation (e.g. AES_GCM_NONCE instead of AES_128GCM_NONCE)
since they apply to both 128 bit and 256 bit encryption.
Signed-off-by: Steve French <stfrench@microsoft.com>
Reviewed-by: Pavel Shilovsky <pshilov@microsoft.com>
If server does not support AES-256-GCM and it was required on mount, print
warning message. Also log and return a different error message (EOPNOTSUPP)
when encryption mechanism is not supported vs the case when an unknown
unrequested encryption mechanism could be returned (EINVAL).
Signed-off-by: Steve French <stfrench@microsoft.com>
Reviewed-by: Pavel Shilovsky <pshilov@microsoft.com>
Add new module load parameter enable_gcm_256. If set, then add
AES-256-GCM (strongest encryption type) to the list of encryption
types requested. Put it in the list as the second choice (since
AES-128-GCM is faster and much more broadly supported by
SMB3 servers). To make this stronger encryption type, GCM-256,
required (the first and only choice, you would use module parameter
"require_gcm_256."
Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Add new module load parameter require_gcm_256. If set, then only
request AES-256-GCM (strongest encryption type).
Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
This is basically the same as STATUS_LOGON_FAILURE,
but after the account is locked out.
Signed-off-by: Stefan Metzmacher <metze@samba.org>
Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Currently there are three supported signing algorithms
Reviewed-by: Aurelien Aptel <aaptel@suse.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
RHBZ: 1848178
Some calls that set attributes, like utimensat(), are not supposed to return
-EINTR and thus do not have handlers for this in glibc which causes us
to leak -EINTR to the applications which are also unprepared to handle it.
For example tar will break if utimensat() return -EINTR and abort unpacking
the archive. Other applications may break too.
To handle this we add checks, and retry, for -EINTR in cifs_setattr()
Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Currently STATUS_IO_TIMEOUT is not treated as retriable error.
It is currently mapped to ETIMEDOUT and returned to userspace
for most system calls. STATUS_IO_TIMEOUT is returned by server
in case of unavailability or throttling errors.
This patch will map the STATUS_IO_TIMEOUT to EAGAIN, so that it
can be retried. Also, added a check to drop the connection to
not overload the server in case of ongoing unavailability.
Signed-off-by: Rohith Surabattula <rohiths@microsoft.com>
Reviewed-by: Aurelien Aptel <aaptel@suse.com>
Reviewed-by: Pavel Shilovsky <pshilov@microsoft.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Cleanup patch for followon to cache additional information for the root directory
when directory lease held.
Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Cleanup patch for followon to cache additional information for the root directory
when directory lease held.
Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
MS-SMB2 was updated recently to include new protocol definitions for
updated compression payload header and new RDMA transform capabilities
Update structure definitions in smb2pdu.h to match
Signed-off-by: Steve French <stfrench@microsoft.com>
Acked-by: Ronnie Sahlberg <lsahlber@redhat.com>
In encryption capabilities negotiate context can now request
AES256 GCM or CCM
Signed-off-by: Steve French <stfrench@microsoft.com>
Acked-by: Ronnie Sahlberg <lsahlber@redhat.com>
When converting trailing spaces and periods in paths, do so
for every component of the path, not just the last component.
If the conversion is not done for every path component, then
subsequent operations in directories with trailing spaces or
periods (e.g. create(), mkdir()) will fail with ENOENT. This
is because on the server, the directory will have a special
symbol in its name, and the client needs to provide the same.
Signed-off-by: Boris Protopopov <pboris@amazon.com>
Acked-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Pull vfs fix from Al Viro:
"Fixes an obvious bug (memory leak introduced in 5.8)"
* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
pipe: Fix memory leaks in create_pipe_files()
On setxattr() syscall path due to an apprent typo the size of a dynamically
allocated memory chunk for storing struct smb2_file_full_ea_info object is
computed incorrectly, to be more precise the first addend is the size of
a pointer instead of the wanted object size. Coincidentally it makes no
difference on 64-bit platforms, however on 32-bit targets the following
memcpy() writes 4 bytes of data outside of the dynamically allocated memory.
=============================================================================
BUG kmalloc-16 (Not tainted): Redzone overwritten
-----------------------------------------------------------------------------
Disabling lock debugging due to kernel taint
INFO: 0x79e69a6f-0x9e5cdecf @offset=368. First byte 0x73 instead of 0xcc
INFO: Slab 0xd36d2454 objects=85 used=51 fp=0xf7d0fc7a flags=0x35000201
INFO: Object 0x6f171df3 @offset=352 fp=0x00000000
Redzone 5d4ff02d: cc cc cc cc cc cc cc cc cc cc cc cc cc cc cc cc ................
Object 6f171df3: 00 00 00 00 00 05 06 00 73 6e 72 75 62 00 66 69 ........snrub.fi
Redzone 79e69a6f: 73 68 32 0a sh2.
Padding 56254d82: 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZ
CPU: 0 PID: 8196 Comm: attr Tainted: G B 5.9.0-rc8+ #3
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.13.0-1 04/01/2014
Call Trace:
dump_stack+0x54/0x6e
print_trailer+0x12c/0x134
check_bytes_and_report.cold+0x3e/0x69
check_object+0x18c/0x250
free_debug_processing+0xfe/0x230
__slab_free+0x1c0/0x300
kfree+0x1d3/0x220
smb2_set_ea+0x27d/0x540
cifs_xattr_set+0x57f/0x620
__vfs_setxattr+0x4e/0x60
__vfs_setxattr_noperm+0x4e/0x100
__vfs_setxattr_locked+0xae/0xd0
vfs_setxattr+0x4e/0xe0
setxattr+0x12c/0x1a0
path_setxattr+0xa4/0xc0
__ia32_sys_lsetxattr+0x1d/0x20
__do_fast_syscall_32+0x40/0x70
do_fast_syscall_32+0x29/0x60
do_SYSENTER_32+0x15/0x20
entry_SYSENTER_32+0x9f/0xf2
Fixes: 5517554e43 ("cifs: Add support for writing attributes on SMB2+")
Signed-off-by: Vladimir Zapolskiy <vladimir@tuxera.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The afs filesystem has a lock[*] that it uses to serialise I/O operations
going to the server (vnode->io_lock), as the server will only perform one
modification operation at a time on any given file or directory. This
prevents the the filesystem from filling up all the call slots to a server
with calls that aren't going to be executed in parallel anyway, thereby
allowing operations on other files to obtain slots.
[*] Note that is probably redundant for directories at least since
i_rwsem is used to serialise directory modifications and
lookup/reading vs modification. The server does allow parallel
non-modification ops, however.
When a file truncation op completes, we truncate the in-memory copy of the
file to match - but we do it whilst still holding the io_lock, the idea
being to prevent races with other operations.
However, if writeback starts in a worker thread simultaneously with
truncation (whilst notify_change() is called with i_rwsem locked, writeback
pays it no heed), it may manage to set PG_writeback bits on the pages that
will get truncated before afs_setattr_success() manages to call
truncate_pagecache(). Truncate will then wait for those pages - whilst
still inside io_lock:
# cat /proc/8837/stack
[<0>] wait_on_page_bit_common+0x184/0x1e7
[<0>] truncate_inode_pages_range+0x37f/0x3eb
[<0>] truncate_pagecache+0x3c/0x53
[<0>] afs_setattr_success+0x4d/0x6e
[<0>] afs_wait_for_operation+0xd8/0x169
[<0>] afs_do_sync_operation+0x16/0x1f
[<0>] afs_setattr+0x1fb/0x25d
[<0>] notify_change+0x2cf/0x3c4
[<0>] do_truncate+0x7f/0xb2
[<0>] do_sys_ftruncate+0xd1/0x104
[<0>] do_syscall_64+0x2d/0x3a
[<0>] entry_SYSCALL_64_after_hwframe+0x44/0xa9
The writeback operation, however, stalls indefinitely because it needs to
get the io_lock to proceed:
# cat /proc/5940/stack
[<0>] afs_get_io_locks+0x58/0x1ae
[<0>] afs_begin_vnode_operation+0xc7/0xd1
[<0>] afs_store_data+0x1b2/0x2a3
[<0>] afs_write_back_from_locked_page+0x418/0x57c
[<0>] afs_writepages_region+0x196/0x224
[<0>] afs_writepages+0x74/0x156
[<0>] do_writepages+0x2d/0x56
[<0>] __writeback_single_inode+0x84/0x207
[<0>] writeback_sb_inodes+0x238/0x3cf
[<0>] __writeback_inodes_wb+0x68/0x9f
[<0>] wb_writeback+0x145/0x26c
[<0>] wb_do_writeback+0x16a/0x194
[<0>] wb_workfn+0x74/0x177
[<0>] process_one_work+0x174/0x264
[<0>] worker_thread+0x117/0x1b9
[<0>] kthread+0xec/0xf1
[<0>] ret_from_fork+0x1f/0x30
and thus deadlock has occurred.
Note that whilst afs_setattr() calls filemap_write_and_wait(), the fact
that the caller is holding i_rwsem doesn't preclude more pages being
dirtied through an mmap'd region.
Fix this by:
(1) Use the vnode validate_lock to mediate access between afs_setattr()
and afs_writepages():
(a) Exclusively lock validate_lock in afs_setattr() around the whole
RPC operation.
(b) If WB_SYNC_ALL isn't set on entry to afs_writepages(), trying to
shared-lock validate_lock and returning immediately if we couldn't
get it.
(c) If WB_SYNC_ALL is set, wait for the lock.
The validate_lock is also used to validate a file and to zap its cache
if the file was altered by a third party, so it's probably a good fit
for this.
(2) Move the truncation outside of the io_lock in setattr, using the same
hook as is used for local directory editing.
This requires the old i_size to be retained in the operation record as
we commit the revised status to the inode members inside the io_lock
still, but we still need to know if we reduced the file size.
Fixes: d2ddc776a4 ("afs: Overhaul volume and server record caching and fileserver rotation")
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix missing result check of exfat_build_inode().
And use PTR_ERR_OR_ZERO instead of PTR_ERR.
Signed-off-by: Tetsuhiro Kohada <kohada.t2@gmail.com>
Signed-off-by: Namjae Jeon <namjae.jeon@samsung.com>
Tetsuo Handa reports that splice() can return 0 before the real EOF, if
the data in the splice source pipe is an empty pipe buffer. That empty
pipe buffer case doesn't happen in any normal situation, but you can
trigger it by doing a write to a pipe that fails due to a page fault.
Tetsuo has a test-case to show the behavior:
#define _GNU_SOURCE
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
int main(int argc, char *argv[])
{
const int fd = open("/tmp/testfile", O_WRONLY | O_CREAT, 0600);
int pipe_fd[2] = { -1, -1 };
pipe(pipe_fd);
write(pipe_fd[1], NULL, 4096);
/* This splice() should wait unless interrupted. */
return !splice(pipe_fd[0], NULL, fd, NULL, 65536, 0);
}
which results in
write(5, NULL, 4096) = -1 EFAULT (Bad address)
splice(4, NULL, 3, NULL, 65536, 0) = 0
and this can confuse splice() users into believing they have hit EOF
prematurely.
The issue was introduced when the pipe write code started pre-allocating
the pipe buffers before copying data from user space.
This is modified verion of Tetsuo's original patch.
Fixes: a194dfe6e6 ("pipe: Rearrange sequence in pipe_write() to preallocate slot")
Link:https://lore.kernel.org/linux-fsdevel/20201005121339.4063-1-penguin-kernel@I-love.SAKURA.ne.jp/
Reported-by: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Acked-by: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull epoll fixes from Al Viro:
"Several race fixes in epoll"
* 'work.epoll' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
ep_create_wakeup_source(): dentry name can change under you...
epoll: EPOLL_CTL_ADD: close the race in decision to take fast path
epoll: replace ->visited/visited_list with generation count
epoll: do not insert into poll queues until all sanity checks are done
-----BEGIN PGP SIGNATURE-----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=MT2i
-----END PGP SIGNATURE-----
Merge tag 'for-5.9-rc7-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux
Pull btrfs fixes from David Sterba:
"Two more fixes.
One is for a lockdep warning/lockup (also caught by syzbot), that one
has been seen in practice. Regarding the other syzbot reports
mentioned last time, they don't seem to be urgent and reliably
reproducible so they'll be fixed later.
The second fix is for a potential corruption when device replace
finishes and the in-memory state of trim is not copied to the new
device"
* tag 'for-5.9-rc7-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
btrfs: fix filesystem corruption after a device replace
btrfs: move btrfs_rm_dev_replace_free_srcdev outside of all locks
btrfs: move btrfs_scratch_superblocks into btrfs_dev_replace_finishing
The pipe splice code still used the old model of waiting for pipe IO by
using a non-specific "pipe_wait()" that waited for any pipe event to
happen, which depended on all pipe IO being entirely serialized by the
pipe lock. So by checking the state you were waiting for, and then
adding yourself to the wait queue before dropping the lock, you were
guaranteed to see all the wakeups.
Strictly speaking, the actual wakeups were not done under the lock, but
the pipe_wait() model still worked, because since the waiter held the
lock when checking whether it should sleep, it would always see the
current state, and the wakeup was always done after updating the state.
However, commit 0ddad21d3e ("pipe: use exclusive waits when reading or
writing") split the single wait-queue into two, and in the process also
made the "wait for event" code wait for _two_ wait queues, and that then
showed a race with the wakers that were not serialized by the pipe lock.
It's only splice that used that "pipe_wait()" model, so the problem
wasn't obvious, but Josef Bacik reports:
"I hit a hang with fstest btrfs/187, which does a btrfs send into
/dev/null. This works by creating a pipe, the write side is given to
the kernel to write into, and the read side is handed to a thread that
splices into a file, in this case /dev/null.
The box that was hung had the write side stuck here [pipe_write] and
the read side stuck here [splice_from_pipe_next -> pipe_wait].
[ more details about pipe_wait() scenario ]
The problem is we're doing the prepare_to_wait, which sets our state
each time, however we can be woken up either with reads or writes. In
the case above we race with the WRITER waking us up, and re-set our
state to INTERRUPTIBLE, and thus never break out of schedule"
Josef had a patch that avoided the issue in pipe_wait() by just making
it set the state only once, but the deeper problem is that pipe_wait()
depends on a level of synchonization by the pipe mutex that it really
shouldn't. And the whole "wait for any pipe state change" model really
isn't very good to begin with.
So rather than trying to work around things in pipe_wait(), remove that
legacy model of "wait for arbitrary pipe event" entirely, and actually
create functions that wait for the pipe actually being readable or
writable, and can do so without depending on the pipe lock serializing
everything.
Fixes: 0ddad21d3e ("pipe: use exclusive waits when reading or writing")
Link: https://lore.kernel.org/linux-fsdevel/bfa88b5ad6f069b2b679316b9e495a970130416c.1601567868.git.josef@toxicpanda.com/
Reported-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-and-tested-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Calling pipe2() with O_NOTIFICATION_PIPE could results in memory
leaks unless watch_queue_init() is successful.
In case of watch_queue_init() failure in pipe2() we are left
with inode and pipe_inode_info instances that need to be freed. That
failure exit has been introduced in commit c73be61ced ("pipe: Add
general notification queue support") and its handling should've been
identical to nearby treatment of alloc_file_pseudo() failures - it
is dealing with the same situation. As it is, the mainline kernel
leaks in that case.
Another problem is that CONFIG_WATCH_QUEUE and !CONFIG_WATCH_QUEUE
cases are treated differently (and the former leaks just pipe_inode_info,
the latter - both pipe_inode_info and inode).
Fixed by providing a dummy wacth_queue_init() in !CONFIG_WATCH_QUEUE
case and by having failures of wacth_queue_init() handled the same way
we handle alloc_file_pseudo() ones.
Fixes: c73be61ced ("pipe: Add general notification queue support")
Signed-off-by: Qian Cai <cai@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
We use a device's allocation state tree to track ranges in a device used
for allocated chunks, and we set ranges in this tree when allocating a new
chunk. However after a device replace operation, we were not setting the
allocated ranges in the new device's allocation state tree, so that tree
is empty after a device replace.
This means that a fitrim operation after a device replace will trim the
device ranges that have allocated chunks and extents, as we trim every
range for which there is not a range marked in the device's allocation
state tree. It is also important during chunk allocation, since the
device's allocation state is used to determine if a range is already
allocated when allocating a new chunk.
This is trivial to reproduce and the following script triggers the bug:
$ cat reproducer.sh
#!/bin/bash
DEV1="/dev/sdg"
DEV2="/dev/sdh"
DEV3="/dev/sdi"
wipefs -a $DEV1 $DEV2 $DEV3 &> /dev/null
# Create a raid1 test fs on 2 devices.
mkfs.btrfs -f -m raid1 -d raid1 $DEV1 $DEV2 > /dev/null
mount $DEV1 /mnt/btrfs
xfs_io -f -c "pwrite -S 0xab 0 10M" /mnt/btrfs/foo
echo "Starting to replace $DEV1 with $DEV3"
btrfs replace start -B $DEV1 $DEV3 /mnt/btrfs
echo
echo "Running fstrim"
fstrim /mnt/btrfs
echo
echo "Unmounting filesystem"
umount /mnt/btrfs
echo "Mounting filesystem in degraded mode using $DEV3 only"
wipefs -a $DEV1 $DEV2 &> /dev/null
mount -o degraded $DEV3 /mnt/btrfs
if [ $? -ne 0 ]; then
dmesg | tail
echo
echo "Failed to mount in degraded mode"
exit 1
fi
echo
echo "File foo data (expected all bytes = 0xab):"
od -A d -t x1 /mnt/btrfs/foo
umount /mnt/btrfs
When running the reproducer:
$ ./replace-test.sh
wrote 10485760/10485760 bytes at offset 0
10 MiB, 2560 ops; 0.0901 sec (110.877 MiB/sec and 28384.5216 ops/sec)
Starting to replace /dev/sdg with /dev/sdi
Running fstrim
Unmounting filesystem
Mounting filesystem in degraded mode using /dev/sdi only
mount: /mnt/btrfs: wrong fs type, bad option, bad superblock on /dev/sdi, missing codepage or helper program, or other error.
[19581.748641] BTRFS info (device sdg): dev_replace from /dev/sdg (devid 1) to /dev/sdi started
[19581.803842] BTRFS info (device sdg): dev_replace from /dev/sdg (devid 1) to /dev/sdi finished
[19582.208293] BTRFS info (device sdi): allowing degraded mounts
[19582.208298] BTRFS info (device sdi): disk space caching is enabled
[19582.208301] BTRFS info (device sdi): has skinny extents
[19582.212853] BTRFS warning (device sdi): devid 2 uuid 1f731f47-e1bb-4f00-bfbb-9e5a0cb4ba9f is missing
[19582.213904] btree_readpage_end_io_hook: 25839 callbacks suppressed
[19582.213907] BTRFS error (device sdi): bad tree block start, want 30490624 have 0
[19582.214780] BTRFS warning (device sdi): failed to read root (objectid=7): -5
[19582.231576] BTRFS error (device sdi): open_ctree failed
Failed to mount in degraded mode
So fix by setting all allocated ranges in the replace target device when
the replace operation is finishing, when we are holding the chunk mutex
and we can not race with new chunk allocations.
A test case for fstests follows soon.
Fixes: 1c11b63eff ("btrfs: replace pending/pinned chunks lists with io tree")
CC: stable@vger.kernel.org # 5.2+
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
autofs got broken in some configurations by commit 13c164b1a1
("autofs: switch to kernel_write") because there is now an extra LSM
permission check done by security_file_permission() in rw_verify_area().
autofs is one if the few places that really does want the much more
limited __kernel_write(), because the write is an internal kernel one
that shouldn't do any user permission checks (it also doesn't need the
file_start_write/file_end_write logic, since it's just a pipe).
There are a couple of other cases like that - accounting, core dumping,
and splice - but autofs stands out because it can be built as a module.
As a result, we need to export this internal __kernel_write() function
again.
We really don't want any other module to use this, but we don't have a
"EXPORT_SYMBOL_FOR_AUTOFS_ONLY()". But we can mark it GPL-only to at
least approximate that "internal use only" for licensing.
While in this area, make autofs pass in NULL for the file position
pointer, since it's always a pipe, and we now use a NULL file pointer
for streaming file descriptors (see file_ppos() and commit 438ab720c6:
"vfs: pass ppos=NULL to .read()/.write() of FMODE_STREAM files")
This effectively reverts commits 9db9775224 ("fs: unexport
__kernel_write") and 13c164b1a1 ("autofs: switch to kernel_write").
Fixes: 13c164b1a1 ("autofs: switch to kernel_write")
Reported-by: Ondrej Mosnacek <omosnace@redhat.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Acked-by: Acked-by: Ian Kent <raven@themaw.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The async buffered reads feature is not working when readahead is
turned off. There are two things to concern:
- when doing retry in io_read, not only the IOCB_WAITQ flag but also
the IOCB_NOWAIT flag is still set, which makes it goes to would_block
phase in generic_file_buffered_read() and then return -EAGAIN. After
that, the io-wq thread work is queued, and later doing the async
reads in the old way.
- even if we remove IOCB_NOWAIT when doing retry, the feature is still
not running properly, since in generic_file_buffered_read() it goes to
lock_page_killable() after calling mapping->a_ops->readpage() to do
IO, and thus causing process to sleep.
Fixes: 1a0a7853b9 ("mm: support async buffered reads in generic_file_buffered_read()")
Fixes: 3b2a4439e0 ("io_uring: get rid of kiocb_wait_page_queue_init()")
Signed-off-by: Hao Xu <haoxu@linux.alibaba.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Highlights include:
Bugfixes:
- NFSv4.2: copy_file_range needs to invalidate caches on success
- NFSv4.2: Fix security label length not being reset
- pNFS/flexfiles: Ensure we initialise the mirror bsizes correctly on read
- pNFS/flexfiles: Fix signed/unsigned type issues with mirror indices
-----BEGIN PGP SIGNATURE-----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=cNi/
-----END PGP SIGNATURE-----
Merge tag 'nfs-for-5.9-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs
Pull NFS client bugfixes from Trond Myklebust:
"Highlights include:
- NFSv4.2: copy_file_range needs to invalidate caches on success
- NFSv4.2: Fix security label length not being reset
- pNFS/flexfiles: Ensure we initialise the mirror bsizes correctly
on read
- pNFS/flexfiles: Fix signed/unsigned type issues with mirror
indices"
* tag 'nfs-for-5.9-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs:
pNFS/flexfiles: Be consistent about mirror index types
pNFS/flexfiles: Ensure we initialise the mirror bsizes correctly on read
NFSv4.2: fix client's attribute cache management for copy_file_range
nfs: Fix security label length not being reset
-----BEGIN PGP SIGNATURE-----
iQJEBAABCAAuFiEEwPw5LcreJtl1+l5K99NY+ylx4KYFAl9upV4QHGF4Ym9lQGtl
cm5lbC5kawAKCRD301j7KXHgpuPrD/9K1UQLv38K2nPYclLymOi+GIsukpjgwzdY
SM38GNXU5vYkFhylH/bXfckNQ/gja0/whNpcr/UVCgTWleMnss9UiZaCgysyuIOL
vnBxT4yDZIxtkOwF/790NiwV2FrsmrLFdNZU4LkmfbmmrAlNtjOElKyJsOyNNMzJ
UMzHH2Z1vvUwKz+Yq4fPyZCJbpNHN6ABwkSXY/Nz8oWsKfw728fZztLsP57gOtkl
yYVFO2z1n7VaWp5ZzVFYG51DFuMCIDXgN6mMlaKfnQ6auQZxjR+jg69HSRKLjIx7
ZEG1gl/DzwH1+751P7HnuI3U7BtBYolyErHW4j4a6Ko4XX8PPhG9ODKOmsEMPrEq
gCUGcGgWUsEyz+pyullTEt7ea/oLGJ5N86qtNOdviXETZZTghm47QlzxFWr1/GWy
wH++ctBf/Ekk0dbCBF6mJiqDl/PrVSDSClTVhsGJESEmk4BOoC9zd9zT/EfsiR9m
vA8wLE2g1/5oU+irQ0Dlc/EENVWISiigOFFvTPJJjma9iGXAW3kV2/aYW6DKZSwM
w/va7zTlzt89O+L0AT+Rg8btaiTiaZcs3op1AFa1z5Gut3b5YhWL/e95wlaOI6Nv
Tudm4GX06BaN1QdUDV9g0Pr5iNOaCvuOArjNOU3j7ySusJxiJ8GdA3WFqJ/XUlIV
pne8hC/+7A==
=mBw7
-----END PGP SIGNATURE-----
Merge tag 'io_uring-5.9-2020-09-25' of git://git.kernel.dk/linux-block
Pull io_uring fixes from Jens Axboe:
"Two fixes for regressions in this cycle, and one that goes to 5.8
stable:
- fix leak of getname() retrieved filename
- remove plug->nowait assignment, fixing a regression with btrfs
- fix for async buffered retry"
* tag 'io_uring-5.9-2020-09-25' of git://git.kernel.dk/linux-block:
io_uring: ensure async buffered read-retry is setup properly
io_uring: don't unconditionally set plug->nowait = true
io_uring: ensure open/openat2 name is cleaned on cancelation
