Alexander Motin [Sun, 17 Feb 2013 18:42:30 +0000 (18:42 +0000)]
MFcalloutng:
Microoptimize i8254 one-shot operation mode (disabled by default to allow
timecounter functionality) by not writing to mode and MSB registers when
it is not required. This saves several microseconds of CPU time per call,
reducing minimal measured interrupts interval to 19.5us.
David Xu [Sun, 17 Feb 2013 02:52:42 +0000 (02:52 +0000)]
Make more code be protected by internal mutex, and now it is fork-safe, in
error case, the file exclusive lock is now released as soon as possible,
in previous code, child process can still hold the exclusive lock.
Marcel Moolenaar [Sun, 17 Feb 2013 02:15:19 +0000 (02:15 +0000)]
In kthr.c, obtain the address of the PCB for threads that were running
on a core, when the core was stopped, by calling kgdb_trgt_core_pcb().
This has 2 advantages:
1. We don't need to include a machine-specific header anymore and as
such kthr.c is truly machine independent. This allows the code to
be used in a cross-debugger.
2. We don't need to lookup stoppcbs in generic code when it's an
inherently target-spicific symbol. It does not exist for ia64.
Implement kgdb_trgt_core_pcb() for all architectures, except ia64, by
calling a common function called kgdb_trgt_stop_pcb(). This function
differs from kgdb_trgt_core_pcb() in that it gets the size of the PCB
structure as an argument and as such remains machine independent.
On ia64 the PCB for stopped cores is in the PCPU structure itself.
This for better scaling. The implementation of kgdb_trgt_core_pcb()
for ia64 uses the cpuid_to_pcpu[] array to to obtain the address of
the PCB structure.
Marcel Moolenaar [Sun, 17 Feb 2013 00:51:34 +0000 (00:51 +0000)]
Close a race relating to setting the PCPU pointer (r13). Register r13
points to the TLS in user space and points to the PCPU structure in
the kernel. The race is the result of having the exception handler on
the one hand and the RPC system call entry on the other. The EPC
syscall path is non-atomic in that interrupts are enabled while the
two stacks are switched. The register stack is switched last as that
is the stack used to determine whether we're going back to user space
by the exception handler. If we go back to user space, we restore r13,
otherwise we leave r13 alone. The EPC syscall path however set r13 to
the PCPU structure *before* switching the register stack, which means
that there was a window in which the exception handler would restore
r13 when it was already pointing to the PCPU structure. This is fatal
when the exception happened on CPU x, but left from the exception on
anotehr CPU. In that case r13 would point to the PCPU of the CPU the
thread was running on. This immediately results in getting the wrong
value for curthread.
The fix is to make sure we assign r13 *after* we set ar.bspstore to
point to the kernel register stack for the thread.
- Add hw.bcm2835.sdhci.hs tunable to enable/disable highspeed mode in
SDHCI driver
Suggested by: Daisuke Aoyama
- Set initilization sequence frequency to 8MHz. It should fix Data CRC
errors. Standard requires initialization sequence to be executed
at 400KHz but on this hardware low frequncies seems to cause
Data CRC errors.
Value was derived from analyzing hardware signals after
Raspberry Pi is powered up. Before any data is read though DATA line
adapter's clock frequency is changed to 8MHz.
Modern cards should function fine at 8MHz but for older MMC cards it
can be overriden by setting hw.bcm2835.sdhci.min_freq tunable.
- Replace divisor numbers with more descirptive names
- Properly calculate minimum frequency for SDHCI 3.0
- Properly calculate frequency for SDHCI 3.0 in mmcbr_set_clock
- Add min_freq method to sdhci_if.m and provide default
implementation. By re-implementing this method hardware
drivers can control frequency controller operates when
executing initialization sequence
Ian Lepore [Sat, 16 Feb 2013 20:43:16 +0000 (20:43 +0000)]
In _bus_dmamap_addseg(), the return value must be zero for error, or the size
actually added to the segment (possibly smaller than the requested size if
boundary crossings had to be avoided).
Dimitry Andric [Sat, 16 Feb 2013 20:17:31 +0000 (20:17 +0000)]
Since clang 3.2 now has an option to suppress warnings about implicitly
promoted K&R parameters, remove the workarounds added for sendmail
components in r228558.
Adrian Chadd [Sat, 16 Feb 2013 19:11:57 +0000 (19:11 +0000)]
* Reduce the PCU lock overhead a little by only re-acquiring it if we
actually do have to reinitialise the RX side of things after an RX
descriptor EOL error.
* Revert a change of mine from quite a while ago - don't shortcut the
RX initialisation path. There's a RX FIFO bug in the earlier chips
(I'm not sure when it was fixed in this series, but it's fixed
with the AR9380 and later) which causes the same RX descriptor to
be written to over and over. This causes the descriptor to be
marked as "done", and this ends up causing the whole RX path to
go very strange. This should fixed the "kickpcu; handled X packets"
message spam where "X" is consistently small.
Kirk McKusick [Sat, 16 Feb 2013 15:11:40 +0000 (15:11 +0000)]
The UFS2 filesystem allocates new blocks of inodes as they are needed.
When a cylinder group runs short of inodes, a new block for inodes is
allocated, zero'ed, and written to the disk. The zero'ed inodes must
be on the disk before the cylinder group can be updated to claim them.
If the cylinder group claiming the new inodes were written before the
zero'ed block of inodes, the system could crash with the filesystem in
an unrecoverable state.
Rather than adding a soft updates dependency to ensure that the new
inode block is written before it is claimed by the cylinder group
map, we just do a barrier write of the zero'ed inode block to ensure
that it will get written before the updated cylinder group map can
be written. This change should only slow down bulk loading of newly
created filesystems since that is the primary time that new inode
blocks need to be created.
Reported by: Robert Watson
Reviewed by: kib
Tested by: Peter Holm
Kirk McKusick [Sat, 16 Feb 2013 14:51:30 +0000 (14:51 +0000)]
Add barrier write capability to the VFS buffer interface. A barrier
write is a disk write request that tells the disk that the buffer
being written must be committed to the media along with any writes
that preceeded it before any future blocks may be written to the drive.
Barrier writes are provided by adding the functions bbarrierwrite
(bwrite with barrier) and babarrierwrite (bawrite with barrier).
Following a bbarrierwrite the client knows that the requested buffer
is on the media. It does not ensure that buffers written before that
buffer are on the media. It only ensure that buffers written before
that buffer will get to the media before any buffers written after
that buffer. A flush command must be sent to the disk to ensure that
all earlier written buffers are on the media.
Dimitry Andric [Fri, 15 Feb 2013 23:12:58 +0000 (23:12 +0000)]
Pull in r175057 from upstream llvm trunk:
X86: Disable generation of rep;movsl when %esi is used as a base pointer.
This happens when there is both stack realignment and a dynamic alloca in the
function. If we overwrite %esi (rep;movsl uses fixed registers) we'll lose the
base pointer and the next register spill will write into oblivion.
Fixes PR15249 and unbreaks firefox on i386/freebsd. Mozilla uses dynamic allocas
and freebsd a 4 byte stack alignment.
Ian Lepore [Fri, 15 Feb 2013 18:30:32 +0000 (18:30 +0000)]
Add PPS_CANWAIT support for time_pps_fetch(). This adds support for all three
blocking modes described in section 3.4.3 of RFC 2783, allowing the caller
to retrieve the most recent values without blocking, to block for a specified
time, or to block forever.
Manually add the correct text for the LDNS and LDNS_UTILS options.
There seems to be a bug in makeman that causes it to misunderstand
the relationship between LDNS_UTILS and BIND_UTILS.
Xin LI [Thu, 14 Feb 2013 21:02:18 +0000 (21:02 +0000)]
Eliminate real_LZ4_uncompress. It's unused and does not perform sufficient
check against input stream (i.e. it could read beyond specified input
buffer).
* The following bit flags where incroccetly defined:
o Mesh Control Present
o Mesh Power Save Level
o RSPI
This is now corrected according to Table 8.4 as per IEEE 802.11 2012;
In case of failure of the pkg boostrap advice the user to either change the
PACKAGESITE they use or install from ports directly indicating where the port
is localted in the port collection
Hiroki Sato [Thu, 14 Feb 2013 08:32:07 +0000 (08:32 +0000)]
- Fix libmd dependency. It is needed in the bootstrap library list because
usr.bin/xinstall depends on it.
- Remove libutil from usr.bin/xinstall/Makefile. No symbol was actually used.
Xin LI [Thu, 14 Feb 2013 02:00:41 +0000 (02:00 +0000)]
Simplify r243637 and make sure that nfsdargs.{min,max}threads are always set
to meaningful value:
- When nfsdcnt is set, it dictates all values;
- Otherwise, nfsdargs.minthreads is set to user specified value, or the
automatically detected value if there is no one specified;
nfsdargs.maxthreads is set to the user specified value, or the value
of nfsdargs.minthreads if there is no one specified; when it is smaller
than nfsdargs.minthreads, the latter's value is always used.
Niclas Zeising [Wed, 13 Feb 2013 15:46:33 +0000 (15:46 +0000)]
Add strchrnul(), a GNU function similar to strchr(), except that it returns
a pointer to the end of the string, rather than NULL, if the character was
not found.
Ian Lepore [Wed, 13 Feb 2013 15:09:16 +0000 (15:09 +0000)]
Make the F_READAHEAD option to fcntl(2) work as documented: a value of zero
now disables read-ahead. It used to effectively restore the system default
readahead hueristic if it had been changed; a negative value now restores
the default.
Resolve a LOR after r246616. Protect control requests using the USB device
enumeration lock. Make sure all callers of usbd_enum_lock() check the return
value. Remove the control transfer specific lock. Bump the FreeBSD version
number, hence external USB modules may need to be recompiled due to a USB
device structure change.
Adrian Chadd [Wed, 13 Feb 2013 05:32:19 +0000 (05:32 +0000)]
Pull out the if_transmit() work and revert back to ath_start().
My changed had some rather significant behavioural changes to throughput.
The two issues I noticed:
* With if_start and the ifnet mbuf queue, any temporary latency
would get eaten up by some mbufs being queued. With ath_transmit()
queuing things to ath_buf's, I'd only get 512 TX buffers before I
couldn't queue any further frames.
* There's also some non-zero latency involved with TX being pushed
into a taskqueue via direct dispatch. Any time the scheduler didn't
immediately schedule the ath TX task would cause extra latency.
Various 1ge/10ge drivers implement both direct dispatch (if the TX
lock can be acquired) and deferred task transmission (if the TX lock
can't be acquired), with frames being pushed into a drbd queue.
I'll have to do this at some point, but until I figure out how to
deal with 802.11 fragments, I'll have to wait a while longer.
So what I saw:
* lots of extra latency, specially under load - if the taskqueue
wasn't immediately scheduled, things went pear shaped;
* any extra latency would result in TX ath_buf's taking their sweet time
being replenished, so any further calls to ath_transmit() would drop
mbufs.
* .. yes, there's no explicit backpressure here - things are just dropped.
Eek.
With this, the general performance has gone up, but those subtle if_start()
related race conditions are back. For some reason, this is doubly-obvious
with the AR5416 NIC and I don't quite understand why yet.
There's an unrelated issue with AR5416 performance in STA mode (it's
fine in AP mode when bridging frames, weirdly..) that requires a little
further investigation. Specifically - it works fine on a Lenovo T40
(single core CPU) running a March 2012 9-STABLE kernel, but a Lenovo T60
(dual core) running an early November 2012 kernel behaves very poorly.
The same hardware with an AR9160 or AR9280 behaves perfectly.