reliability when tracing fast-moving processes or writing traces to
slow file systems by avoiding unbounded queueuing and dropped records.
Record loss was previously possible when the global pool of records
become depleted as a result of record generation outstripping record
commit, which occurred quickly in many common situations.
These changes partially restore the 4.x model of committing ktrace
records at the point of trace generation (synchronous), but maintain
the 5.x deferred record commit behavior (asynchronous) for situations
where entering VFS and sleeping is not possible (i.e., in the
scheduler). Records are now queued per-process as opposed to
globally, with processes responsible for committing records from their
own context as required.
- Eliminate the ktrace worker thread and global record queue, as they
are no longer used. Keep the global free record list, as records
are still used.
- Add a per-process record queue, which will hold any asynchronously
generated records, such as from context switches. This replaces the
global queue as the place to submit asynchronous records to.
- When a record is committed asynchronously, simply queue it to the
process.
- When a record is committed synchronously, first drain any pending
per-process records in order to maintain ordering as best we can.
Currently ordering between competing threads is provided via a global
ktrace_sx, but a per-process flag or lock may be desirable in the
future.
- When a process returns to user space following a system call, trap,
signal delivery, etc, flush any pending records.
- When a process exits, flush any pending records.
- Assert on process tear-down that there are no pending records.
- Slightly abstract the notion of being "in ktrace", which is used to
prevent the recursive generation of records, as well as generating
traces for ktrace events.
Future work here might look at changing the set of events marked for
synchronous and asynchronous record generation, re-balancing queue
depth, timeliness of commit to disk, and so on. I.e., performing a
drain every (n) records.
MFC after: 1 month
Discussed with: jhb
Requested by: Marc Olzheim <marcolz at stack dot nl>
- several TerraTec TValue [1]
- PixelView PlayTV Pro REV-4C [2]
In case you have the PixelView card, please tell us the "pciconf -v -l"
output on multimedia@FreeBSD.org if it works. There are revisions out there
which may not work and we need to know which ones work.
PR: 53383 [1], 76002 [2]
Submitted by: Tanja Wittke <tawi@gruft.de> [1], barner [1],
Dan Angelescu <mrhsaacdoh@yahoo.com> [2]
MFC after: 2 months
This is the only file of > 1700 files in a buildkernel here doing that.
It makes reproducible builds (same source => same binary) impossible.
Spotted by: devel/ccache
is that gdb knows SIGLWP and will pass it to program, otherwise gdb
will print out "unknown signal" and discard it, and then thread
cancellation won't work for libthr under gdb.
MFC: 3 days
- Add build_iovec_argf() helper function, for help converting old
mount options which used the mount_argf() function for the mount() syscall.
Discussed with: phk
The threshold for not being tiny was too small. Use the usual 2**-12
threshold. This change is not just an optimization, since the general
code that we fell into has accuracy problems even for tiny x. Avoiding
it fixes 2*1366 args with errors of more than 1 ulp, with a maximum
error of 1.167 ulps.
The magic number 22 is log(DBL_EPSILON)/2 plus slop. This is bogus
for float precision. Use 9 (~log(FLT_EPSILON)/2 plus less slop than
for double precision). The code for handling the interval
[2**-28, 9_was_22] has accuracy problems even for [9, 22], so this
change happens to fix errors of more than 1 ulp in about 2*17000
cases. It leaves such errors in about 2*1074000 cases, with a max
error of 1.242 ulps.
The threshold for switching from returning exp(x)/2 to returning
exp(x/2)^2/2 was a little smaller than necessary. As for coshf(),
This was not quite harmless since the exp(x/2)^2/2 case is inaccurate,
and fixing it avoids accuracy problems in 2*6 cases, leaving problems
in 2*19997 cases.
Fixed naming errors in pseudo-code in comments.
The threshold for not being tiny was confusing and too small. Use the
usual 2**-12 threshold and simplify the algorithm slightly so that
this threshold works (now use the threshold for sinhf() instead of one
for 1+expm1()). This is just a small optimization.
The magic number 22 is log(DBL_EPSILON)/2 plus slop. This is bogus
for float precision. Use 9 (~log(FLT_EPSILON)/2 plus less slop than
for double precision).
The threshold for switching from returning exp(x)/2 to returning
exp(x/2)^2/2 was a little smaller than necessary. This was not quite
harmless since the exp(x/2)^2/2 case is inaccurate. Fixing it happens
to avoid accuracy problems for 2*6 of the 2*151 args that were handled
by the exp(x)/2 case. This leaves accuracy problems for about 2*19997
args near the overflow threshold (~89); the maximum error there is
2.5029 ulps.
There are also accuracy probles for args in +-[0.5*ln2, 9] -- 2*188885
args with errors of more than 1 ulp, with a maximum error of 1.384 ulps.
Fixed a syntax error and naming errors in pseudo-code in comments.
whole name. This does not unnecessarily close the door that in some
future we want to test on something other than md(4) devices.
Also add a "conf" action so that we can check whether a gctl actually
did the right thing or not. It's one thing to check that the result
strings are as expected, but it doesn't tell us if the end result is
correct. This needs a bit more fleshing out, but for now a visual
(i.e. manual) check suffices.
- Teach the mount program to call the nmount() syscall directly
- Preserve existing method of calling mount() for UFS, until we clean things
up.
- Preserve existing method of forking and calling external mount programs for
mfs, msdosfs, nfs, nfs4, ntfs, nwfs, nullfs, portalfs, reiserfs, smbfs,
udf, umapfs, unionfs
- devfs, linprocfs, procfs, ext2fs call nmount() syscall directly, since
that is all those external mount programs were doing
Reviewed by: phk
Discussed on: arch
MD class. Previously only the DISK class was dumped. The only
consumer of this sysctl is libdisk (i.e. sysinstall) and it tests
explicitly for instances of the DISK class. Dumping other classes
is therefore harmless.
By also dumping the MD class regression tests can be written that
use the MD class for operations that would normally be done on the
DISK class. The sysctl can now be used to test if those operations
took an effect. An example is partitioning.
specialized for float precision. The new polynomial has degree 8
instead of 14, and a maximum error of 2**-34.34 (absolute) instead of
2**-30.66. This doesn't affect the final error significantly; the
maximum error was and is about 0.8879 ulps on amd64 -01.
The fdlibm expf() is not used on i386's (the "optimized" asm version
is used), but probably should be since it was already significantly
faster than the asm version on athlons. The asm version has the
advantage of being more accurate, so keep using it for now.
from the printer and discarding the data even if the ulpt device
was opened for reading. This resulted in crashes because two
conconcurrent read transfers were using the same transfer structure.
PR: usb/88886
Reported By: Alex Pivovarov
MFC after: 1 week
given prefix is also used recursively for the dependency packages,
if any. If the -P flag appears after any -p flag on the
command line, it overrides it's effect, causing pkg_add to use the
given prefix recursively.
PR: bin/75742
Submitted by: Frerich Raabe <raabe AT kde DOT org>
MFC after: 3 days
happiness, as well as correct other bugs:
- Replace notion of current and saved accounting credential/vnode with a
single credential/vnode and an acct_suspended flag. This simplifies the
accounting logic substantially.
- Replace acct_mtx with acct_sx, a sleepable lock held exclusively during
reconfiguration and space polling, but shared during log entry
generation. This avoids holding a mutex over sleepable VFS operations.
- Hold the sx lock over the duration of the I/O so that the vnode I/O
cannot occur after vnode close, which could occur previously if
accounting was disabled as a process exited.
- Write the accounting log entry with Giant conditionally acquired based
on the file system where the log is stored. Previously, the accounting
code relied on the caller acquiring Giant.
- Acquire Giant conditionally in the accounting callout based on the file
system where the accounting log is stored. Run the callout MPSAFE.
- Expose acct_suspended via a read-only sysctl so it is possibly to
programmatically determine whether accounting is suspended or not without
attempting to parse logs.
- Check both acct_vp and acct_suspended lock-free before entering the
accounting sx lock in acct().
- When accounting is disabled due to a VBAD vnode (i.e., forceable unmount),
generate a log message indicating accounting has been disabled.
- Correct a long-standing bug in how free space is calculated and compared
to the required space: generate and compare signed results, not unsigned
results, or negative free space will cause accounting to not be suspended
when required, or worse, incorrectly resumed once negative free space is
reached.
MFC after: 2 weeks
we included the length of the path in the returned size but not the length
of the associated Dl_serpath structure. Without this fix, programs
attempting to allocate a structure to hold the search path information
would allocate too small of a buffer and rtld would overrun the buffer
while filling it via a subsequent RTLD_DI_SERINFO request.
Submitted by: "William K. Josephson" wkj at morphisms dot net
Reviewed by: jdp
MFC after: 2 weeks
rather than in ifindex_table[]; all (except one) accesses are
through ifp anyway. IF_LLADDR() works faster, and all (except
one) ifaddr_byindex() users were converted to use ifp->if_addr.
- Stop storing a (pointer to) Ethernet address in "struct arpcom",
and drop the IFP2ENADDR() macro; all users have been converted
to use IF_LLADDR() instead.
