This adds on to 446ec6b092 ("docs: mention that libclang is
required"). That commit updated the dependency list for Debian. This
adds the packages needed to get llvm-config for arch and fedora.
Signed-off-by: Thomas Bertschinger <tahbertschinger@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
passing a parent closure is totally unnecessary when using
BCH_WRITE_SYNC - and pops an assert in debug mode, because the write
path no longer supports delivering completions that way.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
now that we have
- shrinkers running in the background via a thread
- run_shrinkers() considering swap usage
this can probably be safely dropped - and it causes problems when
running as non root, so let's do so.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
- make deb fails with dh: error: Unknown sequence --with
(options should not come before the sequence)
- fix deb build deps
- add missing rpm files
Signed-off-by: Alexander Fougner <fougner89@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
6% of physical RAM is info.totalram >> 4 not info.totalram << 4.
Signed-off-by: Chris Webb <chris@arachsys.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
clang's default warnings include -Wgnu-variable-sized-type-not-at-end,
producing a lot of false alarms about the GNU extension for libbcachefs.
Disable these using cc-disable-warning when building with clang.
Signed-off-by: Chris Webb <chris@arachsys.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Our include/linux/byteorder.h uses the system <asm/byteorder.h>
because we don't provide our own in include/asm. This then pulls in
system <linux/stddef.h> which defines __struct_group. That definition
collides with the unconditional #define __struct_group in our local
include/linux/kernel.h, provoking preprocessor warnings.
Signed-off-by: Chris Webb <chris@arachsys.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Before 326d7c1, the shrinker used freeram and totalram from a struct
sysinfo (constructed from /proc/meminfo) to target 25% free physical
memory. As well as the slowness of repeatedly reading /proc/meminfo,
this was a problem as freeram rises when the system starts to swap.
We don't want swapping to reduce our estimate of memory pressure.
To work around this, in 326d7c1 the shrinker started to use the total
allocated heap from a glibc-specific interface mallinfo2(), aiming to
shrink such that our heap is less than 80% of physical memory, unless
overall free memory is less than 6% so that becomes the determining factor.
Unfortunately, a sign error in the calculation means this heuristic
never worked. It would shrink aggressively when the process was small,
and not at all when the process grew beyond 80% of physical RAM. Only the
fallback test ensuring the free physical RAM doesn't fall below 6% would
actually kick in under memory pressure. It also breaks portability to
anything other than recent glibc.
Later, in 2440469 the mallinfo2() was replaced with the older mallinfo()
to improve compatibility with older glibc. This is even more problematic:
it's still not portable but also struct mallinfo has (signed) int fields
which overflow for large processes on 32-bit machines with a 3G/1G split.
Rather than trying to use libc-specific debug interfaces and our own heap
to inform the shrinker, use the information about free and total swap
we already have from sysinfo(2) to explicitly compensate for swapping
in our estimate of free physical memory. Target free memory of 6% of
physical RAM adjusted for zero swap use when calculating the pressure
on the shrinker, based on the effective behaviour of 326d7c1 in practice
given the sign error.
As well as fixing portability to non-glibc systems, this loosens the
assumption that we are the only process using significant memory when
setting the shrinker target. It wouldn't be unreasonable to run two
fsck jobs against independent devices on a large RAM machine and want to
balance physical RAM between them.
Signed-off-by: Chris Webb <chris@arachsys.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Use a single sysinfo(2) call to fill out struct sysinfo instead of
multiple libc sysconf(3) requests, which will only make sysinfo(2) calls
internally anyway. This also enables us to access other struct sysinfo
fields, not just the three filled-out previously.
As we provide our own definition of struct sysinfo in include/linux/mm.h
to match the kernel, which is not guaranteed to align with the definition
libc provides in <sys/sysinfo.h>, use syscall(SYS_sysinfo, ...) directly
instead of the libc wrapper.
Signed-off-by: Chris Webb <chris@arachsys.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Although glibc also provides basename(3) from <string.h>, this is
non-standard whereas it is available from <libgen.h> portably.
Signed-off-by: Chris Webb <chris@arachsys.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
"fsck/bcachefsck_all:" coming before "all:" in the Makefile
causes the bcachefs binary not being built by default. Fix
this by explicitly setting the .DEFAULT_GOAL=all.
Signed-off-by: Tero Roponen <tero.roponen@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This hooks up cmd_fsck to the new BCH_IOCTL_FSCK_ONLINE: if a filesystem
is mounted, we now use that instead of bailing out with an error.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Create a script that looks for mounted bcachefs filesystems and tries to
run as many of them in parallel as possible without two different
scrubbers hitting the same low level block device.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Add some systemd service files so that bcachefs can automatically fsck
mounted filesystems in the background. Hopefully with minimal
disruption to frontend operations.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
* changed in Makefile to use recomended fuse api 35;
* force foreground mode, because fuse_daemonize is forking after initalization other bcachefs threads;
* fix fuse_link, there where mistype in fuse_log (the paremeter need to be inum.inum);
* The write_aligned is fixed, by wrapping struct bch_write_op into container with single closure and using end_io callback to release
that closure.
- include rules file in rpmspec
- update install docs with udev build requirement
Signed-off-by: Alexander Fougner <fougner89@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We seem to be hitting a rare crash in the exit path of fsck - when
shutting down the shrinker thread. Disable exiting the shrinker thread
as a workaround.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
The unending stream of syzbot bug reports and overwrought filing of CVEs
for corner case handling (i.e. things that distract from actual user
complaints) in XFS has generated all sorts of of overheated rhetoric
about how every bug is a Serious Security Issue(tm) because anyone can
craft a malicious filesystem on a USB stick, insert the stick into a
victim machine, and mount will trigger a bug in the kernel driver that
leads to some compromise or DoS or something.
I thought that nobody would be foolish enough to automount an XFS
filesystem. What a fool I was! It turns out that udisks can be told
that it's okay to automount things, and then GNOME will do exactly that.
Including mounting mangled XFS filesystems! Same with bcachefs!
<delete angry rant about poor decisionmaking and armchair fs developers
blasting us on X while not actually doing any of the work>
Turn off /this/ idiocy by adding a udev rule to tell udisks not to
automount bcachefs filesystems.
This will not stop a logged in user from unwittingly inserting a
malicious storage device and pressing [mount] and getting breached.
This is not a substitute for a thorough audit. This is not a substitute
for lklfuse. This does not solve the general problem of in-kernel fs
drivers being a huge attack surface. I just want to give Kent a break
from some of the oceans of bu******.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Even though the blkid wipe has been fixed to detect and wipe old
superblocks, we still incorrectly skip the wipe when in force mode.
Update the force logic in open_for_format() to bypass the user
request and otherwise proceed with the wipe.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
In list_journal, we note which journal entries are being ignored - this
was forgetting to mark entries newer than the newest flush.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
