Readd _el_fn_sh_complete for backward compatibility
This function is not needed anymore, it allows old sh binary to continue
to run and avoid breaking backward compatibility.
Note that is now just calls the regular _el_fn_complete which does a proper
job at quoting.
ian [Tue, 10 Sep 2019 22:08:34 +0000 (22:08 +0000)]
In am335x_dmtpps, use a spin mutex to interlock between PPS capture and PPS
ioctl(2) handling. This allows doing the pps_event() work in the polling
routine, instead of using a taskqueue task to do that work.
Also, add PNPINFO, and switch to using make_dev_s() to create the cdev.
Using a spin mutex and calling pps_event() from the polling function works
around the situation which requires more than 2 sets of timecounter
timehands in a single-core system to get reliable PPS capture. That problem
would happen when a single-core system is idle in cpu_idle() then gets woken
up with an event timer event which was scheduled to handle a hardclock tick.
That processing path would end up calling tc_windup 3 or 4 times between
when the tc polling function was called and when the taskqueue task would
eventually run, and with only two sets of timehands, the th_generation count
would always be too old to allow the captured PPS data to be used.
lualoader: Revert to ASCII menu frame for serial console
The box drawing characters we use aren't necessarily safe with a serial
console; for instance, in the report by npn@, these were causing his xterm
to send back a sequence that lua picked up as input and halted the boot.
This is less than ideal.
Fallback to ASCII frames for console with 'comconsole' in it. This is a
partial revert r338108 by imp@ -- instead of removing the menu entirely and
disabling color/cursor sequences, just reverting the default frame to ASCII
is enough to not break in this setup.
Reported by: npn
Triaged and recommended by: tsoome
cache: avoid excessive relocking on entry removal during lookup
Due to lock ordering issues (bucket lock held, vnode locks wanted) the code
starts with trylocking which in face of contention often fails. Prior to
the change it would loop back with a possible yield.
Instead note we know what locks are needed and can take them in the right
order, avoiding retries. Then we can safely re-lookup and see if the entry
we are looking for is still there.
On a 104-way box poudriere would result in constant retries during an 11h
run as seen in the vfs.cache.zap_and_exit_bucket_fail counter.
cache: change the formula for calculating lock array sizes
It used to be mp_ncpus * 64, but this gives unnecessarily big values for small
machines and at the same time constraints bigger ones. In particular this helps
on a 104-way box for which the count is now doubled.
While here make cache_purgevfs less likely. Currently it is not efficient in
face of contention due to lock ordering issues. These are fixable but not worth
it at the moment.
jeff [Tue, 10 Sep 2019 19:08:01 +0000 (19:08 +0000)]
Replace redundant code with a few new vm_page_grab facilities:
- VM_ALLOC_NOCREAT will grab without creating a page.
- vm_page_grab_valid() will grab and page in if necessary.
- vm_page_busy_acquire() automates some busy acquire loops.
jeff [Tue, 10 Sep 2019 18:27:45 +0000 (18:27 +0000)]
Use the sleepq lock rather than the page lock to protect against wakeup
races with page busy state. The object lock is still used as an interlock
to ensure that the identity stays valid. Most callers should use
vm_page_sleep_if_busy() to handle the locking particulars.
While here, mark all non-standard ones as FreeBSD-only as
other systems (at least, GNU/Linux and illumos) do not handle
them, so we should not encourage their use.
Compared to current version in base:
- great improvements on the Unicode support
- full support for filename completion including quoting
which means we do not need anymore our custom addition)
- Improved readline compatiblity
Upgrading libedit has been a pain in the past, because somehow we never
managed to properly cleanup the tree in lib/libedit and each merge has always
been very painful. After years of fighting give up and refresh a merge from
scrarch properly in contrib.
Note that the switch to this version will be done in another commit.
Remove mklocale(1) and colldef(1) which are deprecated since FreeBSD 11
In FreeBSD 11 along with the rework on the collation, mklocale(1) and colldef(1)
has been replaced by localedef(1) (a note has been added to the manpage to state
it).
mklocale(1) and colldef(1) has been kept around to be able to build older
versions of FreeBSD. None of the version requiring those tools are supported
anymore so it is time to remove them from base
o Do not run any iconv() processing in -a. The locale of root user is not
what is desired by most of the users who receive their calendar mail.
Just assume that users store their calendars in a format that is readable
to them. This fixes regression from r344340.
o fork() and setusercontext(LOGIN_SETALL) for every user. This makes LANG
set inside a calendar file mostly excessive, as we will pick up user's
login class LANG.
o This also executes complex function cal() that parses user owned files
with appropriate user privileges.
Previously it was run with privileges dropped only temporary for execution
of cal(), and fully dropped only before invoking sendmail (see r22473).
cem [Mon, 9 Sep 2019 22:54:27 +0000 (22:54 +0000)]
ddb(4): Add 'show route <dest>' and 'show routetable [<af>]'
These commands show the route resolved for a specified destination, or
print out the entire routing table for a given address family (or all
families, if none is explicitly provided).
Change synchonization rules for vm_page reference counting.
There are several mechanisms by which a vm_page reference is held,
preventing the page from being freed back to the page allocator. In
particular, holding the page's object lock is sufficient to prevent the
page from being freed; holding the busy lock or a wiring is sufficent as
well. These references are protected by the page lock, which must
therefore be acquired for many per-page operations. This results in
false sharing since the page locks are external to the vm_page
structures themselves and each lock protects multiple structures.
Transition to using an atomically updated per-page reference counter.
The object's reference is counted using a flag bit in the counter. A
second flag bit is used to atomically block new references via
pmap_extract_and_hold() while removing managed mappings of a page.
Thus, the reference count of a page is guaranteed not to increase if the
page is unbusied, unmapped, and the object's write lock is held. As
a consequence of this, the page lock no longer protects a page's
identity; operations which move pages between objects are now
synchronized solely by the objects' locks.
The vm_page_wire() and vm_page_unwire() KPIs are changed. The former
requires that either the object lock or the busy lock is held. The
latter no longer has a return value and may free the page if it releases
the last reference to that page. vm_page_unwire_noq() behaves the same
as before; the caller is responsible for checking its return value and
freeing or enqueuing the page as appropriate. vm_page_wire_mapped() is
introduced for use in pmap_extract_and_hold(). It fails if the page is
concurrently being unmapped, typically triggering a fallback to the
fault handler. vm_page_wire() no longer requires the page lock and
vm_page_unwire() now internally acquires the page lock when releasing
the last wiring of a page (since the page lock still protects a page's
queue state). In particular, synchronization details are no longer
leaked into the caller.
The change excises the page lock from several frequently executed code
paths. In particular, vm_object_terminate() no longer bounces between
page locks as it releases an object's pages, and direct I/O and
sendfile(SF_NOCACHE) completions no longer require the page lock. In
these latter cases we now get linear scalability in the common scenario
where different threads are operating on different files.
__FreeBSD_version is bumped. The DRM ports have been updated to
accomodate the KPI changes.
This turns into a warning in GCC 4.2 that 'reverse' may be used
uninitialized in this function. While I don't immediately see where it's
deciding this from (there's only two paths that make column != NULL, and
they both set reverse), initializing reverse earlier is good for clarity.
Fix number of problems found while testing on SAT devices.
- Remove incomplete and dangerous ata_res decoding from ata_do_cmd().
Instead switch all functions that need the result to use get_ata_status(),
doing the same, but more careful, also reducing code duplication.
- Made get_ata_status() to also decode fixed format sense. In many cases
it is still not enough to make it useful, since it can only report results
of 28-bit command, but it is slightly better then nothing.
- Organize error reporting in ata_do_cmd(), so that if caller specified
AP_FLAG_CHK_COND, it is responsible for command errors (non-ioctl ones).
- Make HPA/AMA errors not fatal for `identify` subcommand.
- Fix reprobe() not being called on HPA/AMA when in quiet mode.
- Remove not very useful messages from `format` and `sanitize` commands
with -y flag. Once they started, they often can't be stopped any way.
cem [Mon, 9 Sep 2019 16:32:23 +0000 (16:32 +0000)]
ddb(4): Add some support for lexing IPv6 addresses
Allow commands to specify that (hex) numbers may start with A-F, by adding
the DRT_HEX flag for db_read_token_flags(). As before, numbers containing
invalid digits for the current radix are rejected.
Also, lex ':' and '::' tokens as tCOLON and tCOLONCOLON respectively.
There is a mild conflict here with lexed "identifiers" (tIDENT): ddb
identifiers may contain arbitrary colons, and the ddb lexer is greedy. So
the identifier lex will swallow any colons it finds inside identifiers, and
consumers are still unable to expect the token sequence 'tIDENT tCOLON'.
That limitation does not matter for IPv6 addresses, because the lexer always
attempts to lex numbers before identifiers.
When softdep_fsync() is running, a caller must already started write
for the mount point. Since unmount or remount to ro suspends mount
point, it cannot run in parallel with softdep_fsync(), which makes
vfs_busy() call there not needed.
Doing blocking vfs_busy() there effectively causes lock order reversal
between vn_start_write() and setting MNTK_UNMOUNT, because
vfs_busy(mp, 0) sleeps waiting for MNTK_UNMOUNT becoming clear, while
unmount sets the flag and starts the suspension.
Note that all other uses of vfs_busy() in SU code are non-blocking.
Reported by: chs by mckusick
Reviewed by: mckusick
Tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
projects / freebsd.git / log