mav [Sat, 18 Feb 2017 06:03:16 +0000 (06:03 +0000)]
Change XCOPY memory allocations.
Before this change XCOPY code could allocate memory in chunks up to 16-32MB
(VMware does XCOPY in 4MB chunks by default), that could be difficult for
VM subsystem to do due to KVA fragmentation, that sometimes created huge
allocation delays, blocking any I/O for respective LU for that time.
This change limits allocations down to TPC_MAX_IO_SIZE, which is 1MB now.
1MB is also not a cookie, but ZFS also can do that for large blocks, so
it should be less dramatic. As drawback this increases CPU overhead, but
it still look acceptable comparing to time consumed by ZFS read/write.
ken [Fri, 17 Feb 2017 20:04:22 +0000 (20:04 +0000)]
Add task attribute support to camcontrol(8).
Users can use the new generic argument, -Q task_attr, to specify a task
attribute (simple, ordered, head of queue, aca) for the commands issued.
The the default is simple, which works with all SCSI devices that support
tagged queueing.
This will mostly be useful for debugging target behavior in certain
situations.
You can try it out by compiling CTL with CTL_IO_DELAY turned on (in
sys/cam/ctl/ctl_io.h) and then do something like this with one of the CTL
LUNs:
And at then before the 10 second timer is up, in another terminal:
camcontrol inquiry da34 -Q ordered -v
The Inquiry should complete just after the TUR completes. Ordinarily
it would complete first because of the delay injection, but because the
task attribute is set to ordered in this case, CTL holds it up until the
previous command has completed.
sbin/camcontrol/camcontrol.c:
Add the new generic argument, -Q, which allows the user to specify
a SCSI task attribute. The user can specify task attributes by
name or numerically.
Add a new task_attr arguments to SCSI sub-functions.
sbin/camcontrol/attrib.c,
sbin/camcontrol/camcontrol.h,
sbin/camcontrol/fwdownload.c,
sbin/camcontrol/modeedit.c,
sbin/camcontrol/persist.c,
sbin/camcontrol/timestamp.c,
sbin/camcontrol/zone.c:
Add the new task_attr argument to SCSI sub-functions.
sbin/camcontrol/camcontrol.8:
Document the new -Q option, and add an example.
sbruno [Fri, 17 Feb 2017 20:02:40 +0000 (20:02 +0000)]
Update OLD_DIRS for various targets so that some of the branches of
directories are pruned when the appropriate knobs are turned.
Specifically, turning off bsdconfig, locales, examples, i18n, man,
ntp, syscons. It may not seem like a lot, but it helps when trying
to keep an x86 image under 96MB for MFSRoot netbooting.
adrian [Fri, 17 Feb 2017 18:49:18 +0000 (18:49 +0000)]
[asmc] Add support for MacBook Pro 11,2
This patch will add support for MacBookPro 11.2.
For the macros, the MBP11_* macros (for the existing MacBookPro11.3) did not
match so they have been renamed to MBP113_* and a new MBP112_* has been
added (modified copy of MBP11_*).
Some trailing whitespaces may have been removed automatically.
PR: kern/214836
Obtained from: Johannes Lundberg <johannes@brilliantservice.co.jp
We see long spa_sync(). We are waiting to hold dp_config_rwlock for writer. Some
other thread holds dp_config_rwlock for reader, then calls arc_get_data_buf(),
which finds that arc_is_overflowing()==B_TRUE. So it waits (while holding
dp_config_rwlock for reader) for arc_reclaim_thread to signal arc_reclaim_waiters_cv.
Before signaling, arc_reclaim_thread does arc_kmem_reap_now(), which takes ~seconds.
Author: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Prakash Surya <prakash.surya@delphix.com>
Approved by: Dan McDonald <danmcd@omniti.com>
mjg [Fri, 17 Feb 2017 15:40:24 +0000 (15:40 +0000)]
mtx: get rid of file/line args from slow paths if they are unused
This denotes changes which went in by accident in r313877.
On most production kernels both said parameters are zeroed and have nothing
reading them in either __mtx_lock_sleep or __mtx_unlock_sleep. Thus this change
stops passing them by internal consumers which this is the case.
Kernel modules use _flags variants which are not affected kbi-wise.
royger [Fri, 17 Feb 2017 12:47:51 +0000 (12:47 +0000)]
x86: fix MTRR initialization if EARLY_AP_STARTUP is used
MTRR handlers are set in {amd64/i686}_mem_drvinit, which is called at
SI_SUB_DRIVERS, and that's too late when EARLY_AP_STARTUP is set because APs
have already started at this point. {amd64/i686}_mrinit is also called too late
for the BSP, since that happens when the memory device is attached, also after
APs have already started.
Move the position to SI_SUB_CPU, and also initialize the state for the BSP, so
that the APs can correctly get to the same state as the BSP.
mmel [Fri, 17 Feb 2017 11:16:19 +0000 (11:16 +0000)]
Publish __aeabi_uidiv and __aeabi_idiv as compatible symbols from libc.
Due to bug[1] in libcompiler_rt, all symbols declared by
DEFINE_AEABI_FUNCTION_ALIAS() are not hidden. All these but two
are explicitly exported from libc and don't causes problems.
Remaining two, __aeabi_uidiv and __aeabi_idiv, infecting all
non-versioned shared libraries. And these symbols are consumed
by many (if not all) packages[2].
As workaround, export these from libc as compatible symbols,
in global namespace. With this, these are still visible for
rtld, but static linker doesn't use then.
[1]
DEFINE_AEABI_FUNCTION_ALIAS() macro uses '.set' directive for
declaration of aliased symbol. Unfortunately, '.set' doesn't
inherit visibility of base symbol, and macro don't explicitly
sets visibility for aliased one.
[2]
Given symbols are exported from non-versioned libraries only if
library itself uses them. So, if world is built for CPU with
HW divide, these function are not used and given symbols are
not exported. By this, contents of these libraries is not stable,
and all packages fails to run.
Note: Due to r313823 I'm forced to commit this too early, without
leave enough time for proper review.
mjg [Fri, 17 Feb 2017 05:39:40 +0000 (05:39 +0000)]
locks: let primitives for modules unlock without always goging to the slsow path
It is only needed if the LOCK_PROFILING is enabled. It has to always check if
the lock is about to be released which requires an avoidable read if the option
is not specified..
mav [Fri, 17 Feb 2017 05:22:58 +0000 (05:22 +0000)]
Change the way MaxCmdSN is used.
Before this change MaxCmdSN was reported as CmdSN + delta, that made it
limit number of requests in transmission from the initiator to target,
that was pretty useless. After this change MaxCmdSN limits number of
requests queued to CTL, i.e. maximal queue depth for the initiator.
The default limit is 256 outstanding requests per initiator at a time.
This code uses existing cs_outstanding_ctl_pdus counter to track queue
depth. It's semantics doen't perfectly match, but close enough to not
add another counter. Just don't set the maxtags below 2.
mjg [Fri, 17 Feb 2017 05:09:51 +0000 (05:09 +0000)]
locks: remove SCHEDULER_STOPPED checks from primitives for modules
They all fallback to the slow path if necessary and the check is there.
This means a panicked kernel executing code from modules will be able to
succeed doing actual lock/unlock, but this was already the case for core code
which has said primitives inlined.
markj [Thu, 16 Feb 2017 23:05:20 +0000 (23:05 +0000)]
Prevent CPU migration when checking the DTrace nofault flag on x86.
dtrace_trap() consumes page and protection faults triggered by code running
in DTrace probe context. Such faults occur with interrupts disabled and are
detected using a per-CPU flag. Regular faults cause dtrace_trap() to be
called with interrupts enabled, and nothing was ensuring that the flag was
read from the correct CPU. This may result in dtrace_trap() consuming
unrelated page and protection faults when DTrace is enabled, causing the
fault handler to return without actually having handled the fault.
Diagnosed by: Ryan Libby <rlibby@gmail.com>
MFC after: 3 days
Sponsored by: Dell EMC Isilon
imp [Thu, 16 Feb 2017 21:57:35 +0000 (21:57 +0000)]
Remove EISA bus support for add-in cards. Remove related kernel and
compile options. Remove doxygen pointers to now deleted files. Remove
EISA and VME as examples in bus_space.9.
Retained EISA mode code for IO PIC and MPTABLES because that's not
EISA bus, per se, and some people have abused EISA to mean "EISA-like
behavior as opposed to ISA" rather than using it for EISA add-in
cards.
imp [Thu, 16 Feb 2017 21:57:25 +0000 (21:57 +0000)]
Remove EISA support from Buslogic (bt) cards. Remove known models form
man page. Tweak comments to not refer to EISA scenarios now that it is
no longer supported. Remove unused enum.
imp [Thu, 16 Feb 2017 21:57:19 +0000 (21:57 +0000)]
Remove EISA support from ahc driver. The AIC-7770 chip can be on ISA,
VesaLocalBus or EISA. Internally, EISA and ISA are handled the same,
with VL being handled slightly differently. To avoid too much code
churn, retain the EISA name, despite it being used only for ISA
bus. When it is on the ISA bus, weird gymnastics are required with
EISA-space address accesses as well. Remove known models from the ahc
man page. Remove ahc_eisa module.
imp [Thu, 16 Feb 2017 21:57:08 +0000 (21:57 +0000)]
Remove EISA attachment (fea) from pdq driver. Remove vestiges of
TurboChannel and Q-Bus support while I'm here. Remove obsolete
diagnostics from man page.
imp [Thu, 16 Feb 2017 21:57:02 +0000 (21:57 +0000)]
Remove EISA support from adv driver. Remove references to it from man
page. Remove comment about EISA dual channel card. Remove trivial
references in advlib to avoid false positives with grep. Remove stray
MCA reference not worth a seperate commit.
imp [Thu, 16 Feb 2017 21:56:51 +0000 (21:56 +0000)]
Remove EISA support for ep driver. Left in place EISA strings that are
still relevant (ISA cards can still be in EISA mode, and we're still
ignoring those in the identify routine). Notes about cards in EISA
mode have been left in the manual since they aren't relevant to EISA
support, but instruct how to properly configure an ISA card in a mode
when it is in a ISA bus slot.
imp [Thu, 16 Feb 2017 21:56:46 +0000 (21:56 +0000)]
Remove references to EISA support from the vx driver, along with EISA
support. Fix a comment block that's shared with both vx and ep. Remove
obsolete refernce to statically compiling a kernel with a fixed number
of vx devices. Have not removed EISA from the title of the document
the register definitions were originally derived from (though no doubt
more recent docments were also consulted).
rstone [Thu, 16 Feb 2017 21:18:31 +0000 (21:18 +0000)]
Revert r313814 and r313816
Something evidently got mangled in my git tree in between testing and
review, as an old and broken version of the patch was apparently submitted
to svn. Revert this while I work out what went wrong.
vangyzen [Thu, 16 Feb 2017 20:47:41 +0000 (20:47 +0000)]
Use inet_ntoa_r() instead of inet_ntoa() throughout the kernel
inet_ntoa() cannot be used safely in a multithreaded environment
because it uses a static local buffer. Instead, use inet_ntoa_r()
with a buffer on the caller's stack.
vangyzen [Thu, 16 Feb 2017 20:44:44 +0000 (20:44 +0000)]
pf: use inet_ntoa_r() instead of inet_ntoa(); maybe fix IPv6 OS fingerprinting
inet_ntoa() cannot be used safely in a multithreaded environment
because it uses a static local buffer. Instead, use inet_ntoa_r()
with a buffer on the caller's stack.
This code had an INET6 conditional before this commit, but opt_inet6.h
was not included, so INET6 was never defined. Apparently, pf's OS
fingerprinting hasn't worked with IPv6 for quite some time.
This commit might fix it, but I didn't test that.
Reviewed by: gnn, kp
MFC after: 2 weeks
Relnotes: yes (if I/someone can test pf OS fingerprinting with IPv6)
Sponsored by: Dell EMC
Differential Revision: https://reviews.freebsd.org/D9625
rstone [Thu, 16 Feb 2017 20:06:21 +0000 (20:06 +0000)]
Fix a typo in my previous commit
Somehow in the late stages of testing my sched_ule patch, a character was
accidentally deleted from the file. Correct this.
While I'm committing anyway, the previous commit message requires some
clarification: in the normal case of unlending priority after releasing
a mutex, the thread that was doing the lending will be woken up and
immediately become the highest-priority thread, and in that case no
priority inversion would take place. However, if that thread is pinned
to a different CPU, then the currently running thread that just had its
priority lowered will not be preempted and then priority inversion can
occur.
Reported by: O. Hartmann (typo), jhb (scheduler clarification)
MFC after: 1 month
Pointy hat to: rstone
rstone [Thu, 16 Feb 2017 19:41:13 +0000 (19:41 +0000)]
Check for preemption after lowering a thread's priority
When a high-priority thread is waiting for a mutex held by a
low-priority thread, it temporarily lends its priority to the
low-priority thread to prevent priority inversion. When the mutex
is released, the lent priority is revoked and the low-priority
thread goes back to its original priority.
When the priority of that thread is lowered (through a call to
sched_priority()), the schedule was not checking whether
there is now a high-priority thread in the run queue. This can
cause threads with real-time priority to be starved in the run
queue while the low-priority thread finishes its quantum.
Fix this by explicitly checking whether preemption is necessary
when a thread's priority is lowered.
https://www.illumos.org/issues/7500
With the integration of:
commit 0f6d88aded0d165f5954688a9b13bac76c38da84
Author: Alex Reece <alex@delphix.com>
Date: Sat Jul 26 13:40:04 2014 -0800
4873 zvol unmap calls can take a very long time for larger datasets
the dnode's dn_bufs field was changed from a list to a tree. As a result,
the dn_unlisted_l0_blkid field is no longer necessary.
Author: Stephen Blinick <stephen.blinick@delphix.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Dan Kimmel <dan.kimmel@delphix.com>
Approved by: Gordon Ross <gordon.w.ross@gmail.com>
ed [Thu, 16 Feb 2017 06:52:53 +0000 (06:52 +0000)]
Remove unnecessary #includes from the kqueue(2) man page.
Now that <sys/event.h> can be included on its own, adjust the manual
page accordingly. Remove both unnecessary #include statements from the
synopsis and the example code.
While there, also add a note to the BUGS section to mention that
previous versions of this header file still depend on <sys/types.h>.
mjg [Wed, 15 Feb 2017 23:33:14 +0000 (23:33 +0000)]
rwlock: tidy up r313392
While a new bit was added and thread alignment got shifted to accomodate it,
RW_READERS_SHIFT was not modified accordingly and clashed with the new flag.
This was surprisingly harmless. If the lock was taken for writing, other flags
were tested. If the lock was taken for reading, it would correctly work for
readers > 1 and this was the only relevant test performed.
imp [Wed, 15 Feb 2017 23:04:25 +0000 (23:04 +0000)]
Remove Micro Channel Architecture support. Of the commonly available
machines, only a few 486 machines that used it, and those haven't had
enough memory to run FreeBSD for quite some time (often limited to
16MB).
Not to be confused with the Machine Check Architecture, which is still
very much alive and used (and untouched by this commit).
mav [Wed, 15 Feb 2017 19:46:00 +0000 (19:46 +0000)]
Fix handling of negative sbspace() return values.
I found that at least with Chelsio NICs TOE sockets quite often report
negative sbspace() values. Using unsigned variable to store it resulted
in attempts to aggregate too much data in one sosend() call, that caused
errors and following connection termination.
trasz [Wed, 15 Feb 2017 16:52:21 +0000 (16:52 +0000)]
Change the "devfs_fsync: vop_stdfsync failed" from panic to a printf.
It's not a proper fix, but should be better than what we have now.
Since it got broken some six months ago it results in an incredibly
annoying and trivially reproducible panic every time eg an USB disk
gets disconnected.