Adrian Chadd [Thu, 14 Mar 2013 06:20:02 +0000 (06:20 +0000)]
Implement "holding buffers" per TX queue rather than globally.
When working on TDMA, Sam Leffler found that the MAC DMA hardware
would re-read the last TX descriptor when getting ready to transmit
the next one. Thus the whole ATH_BUF_BUSY came into existance -
the descriptor must be left alone (very specifically the link pointer
must be maintained) until the hardware has moved onto the next frame.
He saw this in TDMA because the MAC would be frequently stopping during
active transmit (ie, when it wasn't its turn to transmit.)
Fast-forward to today. It turns out that this is a problem not with
a single MAC DMA instance, but with each QCU (from 0->9). They each
maintain separate descriptor pointers and will re-read the last
descriptor when starting to transmit the next.
So when your AP is busy transmitting from multiple TX queues, you'll
(more) frequently see one QCU stopped, waiting for a higher-priority QCU
to finsh transmitting, before it'll go ahead and continue. If you mess
up the descriptor (ie by freeing it) then you're short of luck.
Thanks to rpaulo for sticking with me whilst I diagnosed this issue
that he was quite reliably triggering in his environment.
This is a reimplementation; it doesn't have anything in common with
the ath9k or the Qualcomm Atheros reference driver.
Now - it in theory doesn't apply on the EDMA chips, as long as you
push one complete frame into the FIFO at a time. But the MAC can DMA
from a list of frames pushed into the hardware queue (ie, you concat
'n' frames together with link pointers, and then push the head pointer
into the TXQ FIFO.) Since that's likely how I'm going to implement
CABQ handling in hostap mode, it's likely that I will end up teaching
the EDMA TX completion code about busy buffers, just to be "sure"
this doesn't creep up.
Tested - iperf ap->sta and sta->ap (with both sides running this code):
* AR5416 STA
* AR9160/AR9220 hostap
To validate that it doesn't break the EDMA (FIFO) chips:
* AR9380, AR9485, AR9462 STA
Using iperf with the -S <tos byte decimal value> to set the TCP client
side DSCP bits, mapping to different TIDs and thus different TX queues.
TODO:
* Make this work on the EDMA chips, if we end up pushing lists of frames
to the hardware (eg how we eventually will handle cabq in hostap/ibss
mode.)
Tijl Coosemans [Wed, 13 Mar 2013 22:01:31 +0000 (22:01 +0000)]
- Fix two possible overflows when testing if ELF program headers are on
the first page:
1. Cast uint16_t operands in a multiplication to unsigned int because
otherwise the implicit promotion to int results in a signed
multiplication that can overflow and the behaviour on integer
overflow is undefined.
2. Replace (offset + size > PAGE_SIZE) with (size > PAGE_SIZE - offset)
because the sum may overflow.
- Use the same tests to see if the path to the interpreter is on the first
page. There's no overflow here because size is already limited by
MAXPATHLEN, but the compiler optimises the new tests better. Also fix an
off-by-one error.
- Simplify tests to see if an ELF note program header is on the first page.
This also fixes an off-by-one error.
- Make quirk for reading device descriptor from broken USB devices.
Else they won't enumerate at all:
hw.usb.full_ddesc=1
- Reduce the USB descriptor read timeout from 1000ms to
500ms. Typical value for LOW speed devices is 50-100ms.
- Enumerate USB device a maximum of 3 times when a port
connection change event is detected, before giving up.
- Make the FreeBSD's USB library compile under Linux.
- Fix a compile warning where the return value of a call
to a write() function was ignored.
- Remove redundant include files from userland USB header files.
- Add some now needed include files to various C-files.
Unlike OpenBSD's, our setusercontext() will intentionally ignore the user's
own umask setting (from ~/.login.conf) unless running with the user's UID.
Therefore, we need to call it again with LOGIN_SETUMASK after changing UID.
Pyun YongHyeon [Wed, 13 Mar 2013 01:40:01 +0000 (01:40 +0000)]
r241438 broke IPMI access on Sun Fire X2200 M2(BCM5715).
Fix the IPMI regression by sending BGE_FW_DRV_STATE_UNLOAD to
ASF/IPMI firmware in driver attach phase. Sending heartheat to
ASF/IPMI is enabled only after upping interface so
setting driver state to BGE_FW_DRV_STATE_START in attach phase
broke IPMI access.
While I'm here, add NVRAM arbitration lock before performing
controller reset. ASF/IPMI firmware may be able to access the NVRAM
while controller reset is in progress. Without the arbitration
lock before resetting the controller, ASF/IPMI may not initialize
properly.
Special thanks to Miroslav Lachman who provided full remote
debugging environments.
Gleb Smirnoff [Tue, 12 Mar 2013 13:42:47 +0000 (13:42 +0000)]
Functions m_getm2() and m_get2() have different order of arguments,
and that can drive someone crazy. While m_get2() is young and not
documented yet, change its order of arguments to match m_getm2().
John Baldwin [Tue, 12 Mar 2013 12:35:02 +0000 (12:35 +0000)]
Remove fortunes-o from the base system. Debating what does or does not
belong in this file is not a useful exercise or conducive to producing a
high quality advanced operating system.
While here, simplify the make rules to autocompute BLDS and FILES from a
single DB variable.
Gleb Smirnoff [Tue, 12 Mar 2013 12:15:24 +0000 (12:15 +0000)]
In kern_sendfile() use m_extadd() instead of MEXTADD() macro, supplying
appropriate wait argument and checking return value. Before this change
m_extadd() could fail, and kern_sendfile() ignored that.
Gleb Smirnoff [Tue, 12 Mar 2013 12:12:16 +0000 (12:12 +0000)]
The m_extadd() can fail due to memory allocation failure, thus:
- Make it return int, not void.
- Add wait parameter.
- Update MEXTADD() macro appropriately, defaults to M_NOWAIT, as
before this change.
Alexander Motin [Tue, 12 Mar 2013 06:58:49 +0000 (06:58 +0000)]
Make kern_nanosleep() and pause_sbt() to use per-CPU sleep queues.
This removes significant sleep queue lock congestion on multithreaded
microbenchmarks, making them scale to multiple CPUs almost linearly.
Neel Natu [Mon, 11 Mar 2013 17:36:37 +0000 (17:36 +0000)]
Convert the offset into the bar that contains the MSI-X table to an offset
into the MSI-X table before using it to calculate the table index.
In the common case where the MSI-X table is located at the begining of the
BAR these two offsets are identical and thus the code was working by accident.
This change will fix the case where the MSI-X table is located in the middle
or at the end of the BAR that contains it.
John Baldwin [Mon, 11 Mar 2013 16:33:05 +0000 (16:33 +0000)]
Fix the 'C' field for a running thread to match the behavior described
in the manpage by having it display the current CPU (ki_oncpu) rather
than the previously used CPU (ki_lastcpu). ki_lastcpu is still used for
all other thread states.
Reported by: Chris Ross <cross+freebsd@distal.com>
MFC after: 1 week
Gleb Smirnoff [Mon, 11 Mar 2013 13:05:11 +0000 (13:05 +0000)]
Fix for quite a special case when userland emulates a netgraph node, and
userland can reply to a message with NGM_HASREPLY bit set. In this case
we should not wait for a response to a responce.
PR: 176771
Submitted by: Keith Reynolds <keith.reynolds tidalscale.com>
Adrian Chadd [Mon, 11 Mar 2013 06:54:58 +0000 (06:54 +0000)]
Add a few new fields to the RX vendor radiotap header:
* a flags field that lets me know what's going on;
* the hardware ratecode, unmolested by conversion to a bitrate;
* the HAL rs_flags field, useful for debugging;
* specifically mark aggregate sub-frames.
This stuff sorely needs tidying up - it's missing some important
stuff (eg numdelims) and it would be nice to put the flags at the
beginning rather than at the end.
Tested:
* AR9380, STA mode, 2x2 HT40, monitoring RSSI and EVM values
Andrew Turner [Sun, 10 Mar 2013 03:52:35 +0000 (03:52 +0000)]
- Clang doesn't understand the -mno-thumb-interwork. Only use it with gcc.
- We need to add "-mllvm -arm-enable-ehabi" to clangs CFLAGS when
generating the unwind tables to tell it to add the required directives to
the assembly it generates.
Ian Lepore [Sun, 10 Mar 2013 00:43:01 +0000 (00:43 +0000)]
Attach the elf section headers to the loaded kernel as metadata, so
they can easily be used by later post-processing. When searching for
a compiled-in fdt blob, use the section headers to get the size and
location of the .dynsym section to do a symbol search.
This fixes a problem where the search could overshoot the symbol
table and wander into the string table. Sometimes that was harmless
and sometimes it lead to spurious panic messages about an offset
bigger than the module size.
Andrew Turner [Sat, 9 Mar 2013 23:55:23 +0000 (23:55 +0000)]
__FreeBSD_ARCH_armv6__ is undefined on clang. We can use __ARM_ARCH in
it's place. This makes 'uname -p' correctly output 'armv6' on a kernel
built with clang.
Ian Lepore [Sat, 9 Mar 2013 23:05:19 +0000 (23:05 +0000)]
Since ubldr doesn't necessarily load a kernel at the physical address in the
elf headers, mask out the high nibble of that address. This effectly makes
the entry point the offset from the load address, and it gets adjusted for
the actual load address before jumping to it.
Masking the high nibble makes assumptions about memory layout that are true
for all the arm platforms we support right now, but it makes me uneasy.
This needs to be revisited.
Davide Italiano [Sat, 9 Mar 2013 20:03:10 +0000 (20:03 +0000)]
Fixup r248032:
Change size requested to malloc(9) now that callwheel buckets are
callout_list and not callout_tailq anymore. This change was already
there but it seems it got lost after code churn in r248032.
Attilio Rao [Sat, 9 Mar 2013 15:31:19 +0000 (15:31 +0000)]
Improve UMTX_PROFILING:
- Use u_int values for length and max_length values
- Add a way to reset the max_length heuristic in order to have the
possibility to reuse the mechanism consecutively without rebooting
the machine
- Add a way to quick display top5 contented buckets in the system for
the max_length value.
This should give a quick overview on the quality of the hash table
distribution.
Davide Italiano [Sat, 9 Mar 2013 13:25:45 +0000 (13:25 +0000)]
smbfs_lookup() in the DOTDOT case operates on dvp->n_parent without
proper locking. This doesn't prevent in any case reclaim of the vnode.
Avoid this not going over-the-wire in this case and relying on subsequent
smbfs_getattr() call to restore consistency.
While I'm here, change a couple of SMBVDEBUG() in MPASS().
sbmfs_smb_lookup() doesn't and shouldn't know about '.' and '..'