Alan Somers [Wed, 4 May 2016 22:34:11 +0000 (22:34 +0000)]
Improve performance and functionality of the bitstring(3) api
Two new functions are provided, bit_ffs_at() and bit_ffc_at(), which allow
for efficient searching of set or cleared bits starting from any bit offset
within the bit string.
Performance is improved by operating on longs instead of bytes and using
ffsl() for searches within a long. ffsl() is a compiler builtin in both
clang and gcc for most architectures, converting what was a brute force
while loop search into a couple of instructions.
All of the bitstring(3) API continues to be contained in the header file.
Some of the functions are large enough that perhaps they should be uninlined
and moved to a library, but that is beyond the scope of this commit.
sys/sys/bitstring.h:
Convert the majority of the existing bit string implementation from
macros to inline functions.
Properly protect the implementation from inadvertant macro expansion
when included in a user's program by prefixing all private
macros/functions and local variables with '_'.
Add bit_ffs_at() and bit_ffc_at(). Implement bit_ffs() and
bit_ffc() in terms of their "at" counterparts.
Provide a kernel implementation of bit_alloc(), making the full API
usable in the kernel.
Improve code documenation.
share/man/man3/bitstring.3:
Add pre-exisiting API bit_ffc() to the synopsis.
Document new APIs.
Document the initialization state of the bit strings
allocated/declared by bit_alloc() and bit_decl().
Correct documentation for bitstr_size(). The original code comments
indicate the size is in bytes, not "elements of bitstr_t". The new
implementation follows this lead. Only hastd assumed "elements"
rather than bytes and it has been corrected.
etc/mtree/BSD.tests.dist:
tests/sys/Makefile:
tests/sys/sys/Makefile:
tests/sys/sys/bitstring.c:
Add tests for all existing and new functionality.
include/bitstring.h
Include all headers needed by sys/bitstring.h
lib/libbluetooth/bluetooth.h:
usr.sbin/bluetooth/hccontrol/le.c:
Include bitstring.h instead of sys/bitstring.h.
sbin/hastd/activemap.c:
Correct usage of bitstr_size().
sys/dev/xen/blkback/blkback.c
Use new bit_alloc.
sys/kern/subr_unit.c:
Remove hard-coded assumption that sizeof(bitstr_t) is 1. Get rid of
unrb.busy, which caches the number of bits set in unrb.map. When
INVARIANTS are disabled, nothing needs to know that information.
callapse_unr can be adapted to use bit_ffs and bit_ffc instead.
Eliminating unrb.busy saves memory, simplifies the code, and
provides a slight speedup when INVARIANTS are disabled.
sys/net/flowtable.c:
Use the new kernel implementation of bit-alloc, instead of hacking
the old libc-dependent macro.
sys/sys/param.h
Update __FreeBSD_version to indicate availability of new API
Enji Cooper [Wed, 4 May 2016 21:15:28 +0000 (21:15 +0000)]
Default NO_INSTALLEXTRAKERNELS to yes, not no
The old (^/stable/9) default was yes, not no ("no" was the new default
introduced recently that broke POLA). Restore it to keep POLA like
glebius intended in r299077
MFC after: 3 days
X-MFC with: r299086
Pointyhat to: ngie (research before assuming and committing next time)
Sponsored by: EMC / Isilon Storage Division
Alan Somers [Wed, 4 May 2016 20:20:55 +0000 (20:20 +0000)]
Allow setextattr(8) to take attribute values from stdin
Add the -i option to setextattr. This option allow extended attribute data
to be provided via stdin. Add a -qq option to getextattr, which omits the
trailing newline. Together these options can be used to work with extended
attributes whose values are large and/or binary.
usr.sbin/extattr/Makefile:
Link against libsbuf which is used for processing stdin data.
usr.sbin/extattr/rmextattr.8:
Document setextattr's -i option, getextattr's -qq option, and remove
the BUG about setextattr only being useful for strings.
usr.sbin/extattr/rmextattr.c:
For setextattr operations, buffer attribute data in an sbuf. If -i
is specified, pull the data from stdin, otherwise from the
appropriate argurment.
Update usage text and argument validation code for setextattr's -i
option.
usr.sbin/extattr/tests/extattr_test.sh
Add tests for -q and -i.
Alan Cox [Wed, 4 May 2016 17:54:13 +0000 (17:54 +0000)]
Explain why pmap_copy(), pmap_enter_pde(), and pmap_enter_quick_locked()
call pmap_invalidate_page() even though they are not destroying a leaf-
level page table entry.
Eliminate some bogus white-space characters in a comment.
Bjoern A. Zeeb [Wed, 4 May 2016 16:24:12 +0000 (16:24 +0000)]
While gem5 is not qemu, we treat it as "simulators" or "virtual environments".
Add the needed hardcoded gem5 attachments for the UART there, re-using all
the other bits.
In collaboration with: andrew
Sponsored by: DARPA/AFRL
Reviewed by: andrew
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D6204
Bjoern A. Zeeb [Wed, 4 May 2016 16:15:39 +0000 (16:15 +0000)]
The virtual timer is optional on ARM64. Properly handle that condition. [1]
In case we do not have an interrupt assignment for the virtual timer,
force the physical timer.
Also skip resource allocation for any timer we do not have an interrupt
assignment for.
In collaboration with: andrew
Submitted by: br ([1] from his gem5 arm64 work)
Sponsored by: DARPA/AFRL
Reviewed by: andrew
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D6203
Bjoern A. Zeeb [Wed, 4 May 2016 16:09:51 +0000 (16:09 +0000)]
The ARM generic timer keeps ticking even if disabled or it expired.
In case of updating it with a very low value it might expire again
after writing the tval but before updating ctrl. In that case we do
lose the status bit saying that the timer expired and we will consequently
not get an interrupt for it, leaving the timer in a "dead" state.
In order to solve this increase the minimum period with what the timer
can be loaded to something higher.
Found & analysed with: gem5
Debugged with: andrew
Sponsored by: DARPA/AFRL
Reviewed by: andrew
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D6202
Roger Pau Monné [Wed, 4 May 2016 13:48:59 +0000 (13:48 +0000)]
rtc: fix inverted resolution check
The current code in clock_register checks if the newly added clock has a
resolution value higher than the current one in order to make it the
default, which is wrong. Clocks with a lower resolution value should be
better than ones with a higher resolution value, in fact with the current
code FreeBSD is always selecting the worse clock.
Adrian Chadd [Wed, 4 May 2016 02:11:45 +0000 (02:11 +0000)]
[bwn] fix rate control feedback.
* Add a debug print for the xmit completion status fields.
Yes, I like staring at a stream of DWORDS.
* Set the retrycnt to the number of full frame retries for now;
I'll figure out how to factor rts/cts failures into it when
I figure out what the difference is.
Adrian Chadd [Wed, 4 May 2016 01:36:19 +0000 (01:36 +0000)]
[bwn] accurately(ish) account transmit/recieve failures for rate control.
I noticed that it'd associate fine, but it'd quickly stop exchanging traffic.
Receive was okay, but transmit just failed.
Then I went "wlandebug +rate". I discovered it started at 36M OFDM, and then
quickly rose to 54M, which then showed 0% transmit success.
Then, I dug into how the completion path works. We are reading 'ack=0'
in the TX status side, so .. then I discovered we were only processing the
TX completion status /if/ ack=1. So, we'd only ever count successes;
we'd never count failures, and thus the rate control code thought
everything was a-ok.
We also have to set retrycnt to something non-zero so it indeed does
bring the rate down upon failure.
So:
* Delete the rate control completion code from the tx completion
routine, it's just duplicate and never worked. Putting it behind
'if (status->ack) was pointless.
* Move it to the PIO and DMA completion routines which actually
do free the node reference and mbuf. We know at that point
what the status is, so do it there.
* Fake a retrycnt of 1 for now, so we at least count failures.
Also:
* Start adding comments about weird stuff I find with rate selection.
In this instance, we shouldn't be selecting a fallback rate that
doesn't match the currently configured mode (11a, 11b, 11g, etc.)
This isn't perfect - AMRR does try 54mbit and takes a few packets
before it figures out it's a bad idea - but it's better than nothing.
This makes the bwn(4) driver actually useful for the first time since
I've tried using it - and that dates back to 2011. I've resisted
successfully until now.
Tested:
* Broadcom BCM4312 802.11b/g Wireless, STA mode
WLAN (chipid 0x4312 rev 15) PHY (analog 6 type 5 rev 1) RADIO (manuf 0x17f ver 0x2062 rev 2)
TODO:
* See if the fallback rate actually /is/ working
* Question my own sanity over touching this driver in the first place.
Adrian Chadd [Wed, 4 May 2016 01:29:00 +0000 (01:29 +0000)]
[bwn] ensure the fallback rate stays in the same PHY mode.
Falling back from 6MB OFDM to 5MB CCK (a) may not work well in the
11bg PHYs, (b) won't work at all if you're 11g only, and (c) plainly
won't work for the 11a PHY.
So, don't do that!
Tested:
* BCM4312 802.11b/g Wireless, STA mode
WLAN (chipid 0x4312 rev 15) PHY (analog 6 type 5 rev 1) RADIO (manuf 0x17f ver 0x2062 rev 2)
Reported and tested by: Bill O'Hanlon (bill.ohanlon at gmail.com)
Reviewed by: jhb
MFC after: 1 week
Relnotes: If dmesg shows "SRAT: Duplicate local APIC ID",
try updating your BIOS to fix NUMA support.
Sponsored by: Dell Inc.
John Baldwin [Tue, 3 May 2016 19:45:24 +0000 (19:45 +0000)]
Save and restore SRIOV-related config registers.
Save the value of the IOV control and page size registers and restore
them (along with the VF count) in pci_cfg_save/pci_cfg_restore. This
ensures ARI remains enabled if a PF driver resets itself during the
PCI_IOV_INIT callback. This might also properly restore SRIOV state
across suspend/resume.
Conrad Meyer [Tue, 3 May 2016 17:06:33 +0000 (17:06 +0000)]
ioat(4): Add CRC descriptor structure
Add CRC/MOVECRC operations, as well as the TEST and STORE variants.
With these operations, a CRC32C can be computed over one or more
descriptors' source data. When the STORE operation is encountered, the
accumulated CRC32C is emitted to memory. A TEST operations triggers an
IOAT channel error if the accumulated CRC32C does not match one in
memory.
These operations are not exposed through any API yet.
Conrad Meyer [Tue, 3 May 2016 17:05:58 +0000 (17:05 +0000)]
ioat(4): Limit descriptor allocation to low 40 bits
The IOAT engine can only address the low 40 bits (1 TB) of physmem via
the 'next descriptor' pointer. Restrict acceptable range given to
bus_dma_tag_create to match.
John Baldwin [Tue, 3 May 2016 00:35:11 +0000 (00:35 +0000)]
Fix an off by one error when remapping MSI-X vectors.
pci_remap_msix() can be used to alter the mapping of allocated
MSI-X vectors to the MSI-X table. The code had an off by one error
when adding the IRQ resources after performing a remap. This was
fatal for any vectors in the table that used the "last" valid IRQ as
those vectors were assigned a garbage IRQ value.
Peter Wemm [Tue, 3 May 2016 00:09:13 +0000 (00:09 +0000)]
Change a rounding operation that had missing braces into a roundup2()
macro. Adjust the buffer clipping code to work as expected.
This prevented a number of machines in the FreeBSD.org cluster from
booting due to "ZFS: i/o error - all block copies unavailable"
after an unclean shutdown.
Adrian Chadd [Mon, 2 May 2016 22:58:11 +0000 (22:58 +0000)]
[bwn] break out the 'g' phy code into a separate source file.
* Break out the 'g' phy code;
* Break out the debugging bits into a separate source file, since
some debugging prints are done in the phy code;
* Make some more chip methods in if_bwn.c public.
This brings the size of if_bwn.c down to 6,805 lines which is now
approaching managable.
Adrian Chadd [Mon, 2 May 2016 21:06:02 +0000 (21:06 +0000)]
[bwn] break out the LP PHY code into a separate file.
This (and eventually migrating the other PHY code out) is in preparation
for adding the 11n PHY. No, the 11ac PHY (for the BCM4260 softmac part) isn't
yet open source, so we can't grow that. Yet.
This trims ~3,700 lines of code from if_bwn.c, bringing it down to a slightly
less crazy sounding 10,446 lines of code.
* Change x/a to work similar to gdb. The content of the memory is
treated as an address, printed symbolically and the address is advanced.
This way you can x/a <stack_address> and then just hit return a bunch
of times to locate useful data on the stack.
* Add x/p. The content of the memory is treated as an address and
printed as hex.
This is based on the similar commit from DragonFlyBSD without the
cosmetic changes.
Roger Pau Monné [Mon, 2 May 2016 18:23:48 +0000 (18:23 +0000)]
xen/control: improve suspend/resume
Implement several small improvements to the suspend/resume Xen sequence:
- Call the power_suspend_early event before stopping all processes.
- Stop all processes. This was done implicitly previously by putting all
the CPUs in a known IPI handler.
- Warm up the timecounter.
- Re-initialize the time of day register.
John Baldwin [Mon, 2 May 2016 18:00:38 +0000 (18:00 +0000)]
Add a new bus method to fetch device-specific CPU sets.
bus_get_cpus() returns a specified set of CPUs for a device. It accepts
an enum for the second parameter that indicates the type of cpuset to
request. Currently two valus are supported:
- LOCAL_CPUS (on x86 this returns all the CPUs in the package closest to
the device when DEVICE_NUMA is enabled)
- INTR_CPUS (like LOCAL_CPUS but only returns 1 SMT thread for each core)
For systems that do not support NUMA (or if it is not enabled in the kernel
config), LOCAL_CPUS fails with EINVAL. INTR_CPUS is mapped to 'all_cpus'
by default. The idea is that INTR_CPUS should always return a valid set.
Device drivers which want to use per-CPU interrupts should start using
INTR_CPUS instead of simply assigning interrupts to all available CPUs.
In the future we may wish to add tunables to control the policy of
INTR_CPUS (e.g. should it be local-only or global, should it ignore
SMT threads or not).
The x86 nexus driver exposes the internal set of interrupt CPUs from the
the x86 interrupt code via INTR_CPUS.
The ACPI bus driver and PCI bridge drivers use _PXM to return a suitable
LOCAL_CPUS set when _PXM exists and DEVICE_NUMA is enabled. They also and
the global INTR_CPUS set from the nexus driver with the per-domain set from
_PXM to generate a local INTR_CPUS set for child devices.
Roger Pau Monné [Mon, 2 May 2016 16:16:08 +0000 (16:16 +0000)]
xen/time: fix PV clock resolution
The current resolution of the Xen PV clock is too high, which causes an
adjustment of 5s to be applied to it. Reduce the resolution to be the same
as the RTC plus one, so it's always selected as the best source when
available on x86.
Also don't reset the clock on resume, it's pointless and discards any
previous adjustments.
Roger Pau Monné [Mon, 2 May 2016 16:15:28 +0000 (16:15 +0000)]
xen/time: allow Dom0 to set the host time
Dom0 should be able to set the host time. This is implemented by first
writing to the RTC (as would be done on bare metal), and then using the
XENPF_settime64 hypercall in order to force Xen to update the wallclock
shared page of all domains.
Roger Pau Monné [Mon, 2 May 2016 16:14:55 +0000 (16:14 +0000)]
atrtc: export function to set RTC
This is going to be used by the Xen clock on Dom0 in order to set the RTC of
the host. The current logic in atrtc_settime is moved to atrtc_set and the
unused device_t parameter is removed from the atrtc_set function call so it
can be safely used by other callers.
Roger Pau Monné [Mon, 2 May 2016 16:13:55 +0000 (16:13 +0000)]
xen/timer: remove the timer setup loop
With the removal of the usage of the VCPU_SSHOTTMR_future flag, now
all errors from xentimer_vcpu_start_timer should be considered fatal, and
the loop is no longer needed since in case of setting the timer in the past
we will get an event interrupt right away (instead of returning ETIME).
Sponsored by: Citrix Systems R&D
MFC after : 2 weeks
Roger Pau Monné [Mon, 2 May 2016 16:13:11 +0000 (16:13 +0000)]
xen/x86: don't lose event interrupts
On slow platforms with unreliable TSC, such as QEMU emulated machines,
it is possible for the FreeBSD kernel to request the next event in the
past. In that case, in the current implementation of
xentimer_vcpu_start_timer, we simply return -ETIME. To be precise Xen
returns -ETIME and we pass it on. As a consequence we need to loop
around to function to make sure that the timer is properly set.
Instead it is better to always ask the hypervisor for a timer event,
even if the timeout is past. To do that, remove the VCPU_SSHOTTMR_future
flag.