Some filesystems do not provide inode numbers through readdir (FUSE mounts
without -o use_ino). We therefore have to lookup each directory entry to
get the missing numbers.
dot and double-dot are exceptions, as we already know the values. Moreover,
the lookup code does not expect to get requests for dot and will abort
perfused(8) when it gets some. In order to fix that, we just check for
dot and double-dot special case and use the known values instead of sending
a lookup.
NetBSD's PTHREAD_KEYS_MAX is set to 256, which is low compared to
other systems like Linux (1024) or MacOS X (512). As a result some
setups tested on Linux will exhibit problems on NetBSD because of
pthread_keys usage beyond the limit. This happens for instance on
Apache with various module loaded, and in this case no particular
developper can be blamed for going beyond the limit, since several
modules from different sources contribute to the problem.
This patch makes the limit conigurable through the PTHREAD_KEYS_MAX
environement variable. If undefined, the default remains unchanged
(256). In any case, the value cannot be lowered below POSIX-mandated
_POSIX_THREAD_KEYS_MAX (128).
While there:
- use EXIT_FAILURE instead of 1 when calling err(3) in libpthread.
- Reset _POSIX_THREAD_KEYS_MAX to POSIX mandated 128, instead of 256.
depends on locale settings and the result might not be accepted
or even misinterpreted by the strunvis decoder. As a workaround
encode manually as a string of octal numbers.
strvisx should learn how enforce such an encoding by itself.
Fixes hijacking processes that vfork and exec. Symptom was the child
would spin with read/EAGAIN <-> kevent/EBADF because the inheritance
mechanism relied on setting the holyfd to -1 on fork...which didn't
happen if we didn't hijack vfork.
ok pooka@
This is essential for fork-safety, so don't merely #warn about it.
Attaining fork-safety without it requires restructuring things -- in
particular, there's no clear way to make it per-thread and fork-safe
without some global list of states to zero on fork.
checks. Compared to OpenBSD's reallocarray, makes it easier to avoid memory
leaks on allocation failures and it doesn't depend on malloc(0) != NULL
for correct error checking. Compared to plain realloc, it also avoids
the problem of intermediate integer overflows. The trade-off is the use
of void * to side step C type system with regard to generic pointer to
pointer.
FUSE filesystems do not expect to get metadata updates for [amc]time
and size, they updates the value on their own after operations.
The PUFFS PUFFS_KFLAG_NOFLUSH_META option prevents regular metadata cache
flushes to the filesystem , and libperfuse uses it to match Linux FUSE
behavior.
While there, fix a bug in SETATTR: do not update kernel metadata cache
from SETATTR reply when the request is asynchronous, as we do not have
the reply yet.
Not including non-renamed symbols in libc caused all kinds of configure
scripts to do the wrong thing when they just tested linkage without
including headers. So, go for the "least moving parts" bandaid for now.
tzalloc now scrubs time zone abbreviations compatibly with the way
that tzset always has, by replacing invalid bytes with '_' and by
shortening too-long abbreviations.
When fts_read() gets an error on fchdir(), it exited with sp->fts_cur
set to a freed structure. fts_close() would later attempt to free it
again, crashing the program.
FUSE filesystems assume that SETATTR with atime is the result of utiimes()
being called. As a result, atime and mtime will be updated. This happens
with MooseFS and glusterFS. atime is supposed to be updated by the
filesystem itself when it gets read operations.
We fix the problem in SETATTR operations by
1) do not create a mtime update when we have an atime update (and vice
versa), just fill the fields to avoid the filesystem restting the
missing field to Epoch, but do not pretend we want to update it.
2) If the change is limited to atime, iscard it, as updates should be
done by READ operations
3) Kernel part of PUFFS has been fixed to make sure reads on empty file
are sent to the filesystem:
http://mail-index.netbsd.org/source-changes/2015/01/13/msg062364.html
Thanks to Tom Ivar Helbekkmo for reporting this issue.
with inline code which does what __cerror() was doing. #ifdef
that code (i.e. all code) out of cerror.S; __cerror() is no more.
This seems to be necessary to fix the link of rescue/rescue, and
should have the pleasant side effect of making all other workarounds
done to keep the 'b __cerror' working unnecessary.
* Document error handling of el_init(), el_set(), el_source(),
and history_init().
* Fix a typo an improve punctuation below H_SETUNIQUE.
* The ellipsis already implies "optional", no need for [].
* Sort options in editrc(5).
* Prevent e.g. rom being misconstrued as the end of a sentence.
* Drop a useless duplicate .Ar macro.
* Put telltc in its proper place in the alphabetical order.
* A few typos in vi editor command names.
* Some missing vi editor command names.
* Some missing author macros.
libpuffs calls realpath() to obtain an absolute path to use for mounting.
If the obtained path is different from the one given by the caller, a
warning is issued. This included the situation where the path passed by
the caller just have trailing slashes, a situation where we just want them
to be striped without a warning.
A concrete result is enabling unpatched libpthread to run on the
rumprun stacks (e.g. Xen and bare metal) with a non-NetBSD scheduler.
Those schedulers hook into the existing _lwp_frobnitz() NetBSD syscall
interfaces (well, "syscall" interfaces in that scenario ;)
More specifically about the change itself:
1) instead of calling _lwp_makecontext() followed by _lwp_create()
and passing the entry point in ucontext_t (MD) through the calls, roll
the calls into pthread__makelwp() and allow alternate implementations
for that MI interface.
2) allow compile-time overriding of __lwp_gettcb_fast() or
__lwp_getprivate_fast, which are inline and leak MD scheduler/thread
details into libpthread
Additionally, two small nits:
I) define LIB=pthread before including mk.conf so that it's possible
to test for LIB==pthread in mk.conf
II) make it possible to leave out pthread_cancelstub.c. This is required
by the current implementation of rumprun-posix (i.e. rumprun on
POSIX hosts) due to symbol collisions. It needs to be fixed properly
some day, but for now allows an almost-correct libpthread to run.
I am sure @justin will be happy to explain the details ;)
no change to NetBSD
tested: anita+atf
to unwind the stack. Add a temporary workaround where we simply don't
allow the thread to exit (a kernel thread exit is a relatively uncommon
event in a rump kernel anyway).
* do not build _errno.c (different thread/tls handling)
* do not build either phk- or jemalloc (different backing page allocation)
Somehow I missed these in the previous commit, but now libc built with
RUMPRUN=yes works also with rumprun-posix, so there's reasonably high
confidence that I didn't miss anything anymore.
If "yes", does the following (default "no"):
* prevents ASM syscalls from being built (librump provides syscalls)
* does not include compat (useless when application is linked with libc)
* does not build tags (no /var/db to install them to)
* does not include tls
==> libc for rumprun can now be built against unmodified NetBSD sources
in private mail, it broke rcp(1).
To achieve the documented behavior and to fix long standing incorrect
rsh(1) behavior which I've tried to fix in rev. 1.36, rcmd(1) should have
two operation mode; whether it should relay signal information on
auxiliary channel or not, depending on the argument `fd2p' passed to rcmd(3).
So, make rcmd(1) behave differntly depending on the environment variable and
set it when necessary in rcmd(3) according to how auxiliary channel
is set up by rcmd(3).
that.
- Move the stat() call to fdopendir() and change it's error handling so that
it does not hide errors.
- According to POSIX, fdopendir() transfers ownership of the fd only on
success, so don't close it on failure. XXX: We still make it non-blocking
on failure, but that's nitpicking.
XXX: pullup-7?
querybuf in net/gethnamaddr.c defines MAXPACKET to 64K. This in turn
gets passed down until it reached res_nopt(..., answer, anslen), where
the size of the buffer must fit in 16 bits. Unfortunately we end up
being one more than the max so we end up sending a 0 as the size and
unbound does not like that. Instead we clip now to 64K - 1, and everyone
is happy.
XXX: Pullup to 7.
namely:
* src/lib is used only when building for POSIX'y platforms, but
the man pages have their use for all platforms
* rumpuser.3 is a function of the rump kernel, not one of the of
the POSIX'y implementation hosted in src/lib/librumpuser
no functional change
These were not strictly needed before, as lseek would error on negative
arguments, but having added open_memstream we have a virtual file pointer
that assumes that it gets sane values, so we get an assertion triggered
on a negative value. Best to check in one place rather than at all the
relevant points.
manually. When tzload() fails called from zoneinit(), when trying to set the
local timezone for the first time in tzsetlcl(), we end up with a lclptr
that contains garbage, so settzname() core-dumps.
Thanks Andreas for the analysis!
http://pubs.opengroup.org/onlinepubs/009695399/utilities/\
xcu_chap02.html#tag_02_13_01
1. A [...] pattern containing a slash is not a pattern; the [ ]'s become regular
characters
2. A [] or a [!] is not an empty pattern, why would it? The first would never
match and the second would always match which makes it equivalent to ?
In those cases the ] is taken as a literal character and does not have
special meaning.
* A magic value USE_LOCAL_TIME (defined as 99999) may be passed as the
Timezone to Convert(), instructing it to use mktime() to work
in the local time zone, instead of using mktime_z to work in UTC
(and then adding the specified timezone offset).
* Some old code is removed now that there's no need to find the local
timezone offset.
* Allow either one or both of the now and zone arguments to
parsedate() to be NULL, treating them independently. Previously,
if either one was NULL, the other was ignored.
* If the zone argument is specified, then the current date is calculated
in the specified zone, not in local time.
Also add some disabled debug code.
This should fix PR lib/47916.
a numeric timezone. Move some productions from "time" to "time_numericzone".
Increment yyHaveZone when encountering one of these.
Previously, input of the form "HH:MM:SS +ZZZZ" would not have set the
yyhaveZone flag.
that was designed for handling two-digit abbreviated years.
For example, "1/2/70" still refers to the year 1970, as before,
but "70-01-02" now refers to the year 70.
* Add a new yyHaveFullYear member to struct dateinfo, to record whether
or not the year needs to be adjusted.
* Code that parses years sets yyHaveFullYear=1 if they know that the
year should not be adjusted (as is the case for ISO 8601 dates), or if
they perform their own adjustment (as is the case for CVS timestamps).
* Move the year adjustment code into a new function, AdjustYear,
instead of inline in Convert().
* Make Convert() assume the year doesn't need to be adjusted,
and make Convert's callers first call AdjustYear() if appropriate.