With kernel 4.14 there is no sign of the previous LOGICAL_INO performance
problems, so there seems to be no need to throttle threads using this
ioctl.
Increase the FD cache size limits and scan thread count. Let the kernel
figure out scheduling.
Signed-off-by: Zygo Blaxell <bees@furryterror.org>
This avoids PERFORMANCE warnings when large hash tables are used on slow
CPUs or with lots of worker threads. It also simplifies the code (no
locksets, only one object-wide mutex instead of two).
Fixed a few minor bugs along the way (e.g. we were not setting the dirty
flag on the right hash table extent when we detected hash table errors).
Simplified error handling: IO errors on the hash table are ignored,
instead of throwing an exception into the function that tried to use the
hash table.
Signed-off-by: Zygo Blaxell <bees@furryterror.org>
This commit adds an ebuild for Gentoo. Version 9999 is building live
from current git, currently using kakra:integration because it has some
installation and build fixes important for Gentoo.
Signed-off-by: Kai Krakow <kai@kaishome.de>
According to Gentoo packaging guide, -fPIC should only be used on shared
libraries, and not added unconditionally to every linker call.
Signed-off-by: Kai Krakow <kai@kaishome.de>
In preparation for Gentoo QA checks during ebuild merge phase, let's
make some more of the filesystem layout adjustable.
Signed-off-by: Kai Krakow <kai@kaishome.de>
In Gentoo, usage of DESTDIR is automatically handled by the build system
to support installation into a clean image from which the package is
created.
Thus, let's add DESTDIR to the install targets. One can now correctly
install bees with packaging systems simply by running:
$ DESTDIR=/tmp/bees-image make all install
This will no longer mess up with the PREFIX setting.
CC: Timofey Titovets <nefelim4ag@gmail.com>
Signed-off-by: Kai Krakow <kai@kaishome.de>
If using `scripts/beesd`, we need `blkid` which is part of util-linux.
It should be available on every distribution but let's document it
anyway.
Signed-off-by: Kai Krakow <kai@kaishome.de>
It happened more than once that I ran just "make install" only, which
doesn't install the scripts.
Let's fix this by renaming the previous install target to install_bees,
and then make a new install target which depends on each install target
and thus installs the complete distribution.
It doesn't hurt to install those few scripts. I don't see the point in
separating the install targets as it was previously done.
Signed-off-by: Kai Krakow <kai@kaishome.de>
Let's direct users to the support site when they ask systemd for help
about the service unit, or by looking at error messages.
Also, let's adjust the description to be more pleasing to the eyes. The
previous long description with uncommon formatting really stuck out in
the boot logs.
Signed-off-by: Kai Krakow <kai@kaishome.de>
Starting bees right after local-fs.target is probably not what we want,
as basic setup of the system might not have been done (like udev,
cryptsetup, sysctl, swap, etc).
Let's start only after sysinit.target instead which guarantees that all
basic setup has been done, most importantly, sysctl, udev, and swap have
been setup which may apply important tweaks, configuration, and tuning.
Signed-off-by: Kai Krakow <kai@kaishome.de>
Due to bees installing into the local-fs.target, bees also runs during
system-update.target. This should not be done, system-update.target is
meant as an isolated bootup mode for applying updates offline, that is:
Only essential services are running.
Fix this by making it WantedBy basic.target instead. According to
system-update.target and "man bootup", system-update.target pulls in
sysinit.target, as does basic.target. So essentially, basic.target is
not part of the system-update.target transaction.
Signed-off-by: Kai Krakow <kai@kaishome.de>
If two utilities are found, we get commands like
/usr/bin/markdown /usr/bin/markdown_py README.md > README.html
and that doesn't work.
Signed-off-by: Zygo Blaxell <bees@furryterror.org>
Read-only snapshots have always just worked. Remove them from the
"untested" list.
nodatasum (and therefore nodatacow) inodes are simply ignored. This seems
like the right thing to do since deduping a nodatacow extent turns it
into a datacow extent, which seems contrary to administrator wishes
implied by the nodatacow bit. We probably need an option to
override that assumption.
Clarify why converted ext[234] filesystems may cause problems and
the nature of those problems.
Assorted minor editorial changes.
Discuss calculation of the balance limit parameter when ensuring
sufficient metadata space.
Update kernel version bug/fix/feature lists, including LOGICAL_INO_V2.
Annotate kernel workaround list with known kernel versions that make
the workarounds necessary.
Remove reference to 'DEFRAG_RANGE' as bees requires much more control
over data placement than this interface can offer. It's easy enough to
create a new ioctl to implement bees requirements once it's known what
those requirements are.
Signed-off-by: Zygo Blaxell <bees@furryterror.org>
When a toxic extent is discovered, insert the offending hash/address/toxic
entry into the hash table.
When a previously discovered toxic extent is encountered, do nothing,
i.e. allow the offending hash/address/toxic entry in the hash table
to expire.
Previously both inserts were removed from the code, but the former one
is required. The latter prevents bees from forgiving toxic extents
(or any hash matching one) should they be relocated, deleted, or simply
become non-toxic.
Signed-off-by: Zygo Blaxell <bees@furryterror.org>
Interesting things happen when blindly swapping the release-build CCFLAGS
with the debug-build commented-out CCFLAGS. None of these things that
happen are good.
Signed-off-by: Zygo Blaxell <bees@furryterror.org>
Currently scheme lead to path like:
/tmp/makepkg/bees-git/pkg/bees-git/usr/lib/bees/bees
While packaging, so allow do:
make
make scripts
make install ...
make install_scripts ...
Signed-off-by: Timofey Titovets <nefelim4ag@gmail.com>
GCC 7 and higher turn a previous warning into an error for implicit
fallthrough. Let's hint the compiler that this is intentional here.
Signed-off-by: Kai Krakow <kai@kaishome.de>
(cherry picked from commit 270a91cf17783813edaa75fb9766858dcf8e1689)
