Implement callout_drain_async(), inspired by the projects/hps_head
branch.
This function is used to drain a callout via a callback instead of
blocking the caller until the drain is complete. Refer to the
callout_drain_async() manual page for a detailed description.
Limitation: If a lock is used with the callout, the callout can only
be drained asynchronously one time unless the callout_init_mtx()
function is called again. This limitation is not present in
projects/hps_head and will require more invasive changes to the
timeout code, which was not in the scope of this patch.
To make driver programming easier the TSO limits are changed to
reflect the values used in the BUSDMA tag a network adapter driver is
using. The TCP/IP network stack will subtract space for all linklevel
and protocol level headers and ensure that the full mbuf chain passed
to the network adapter fits within the given limits.
Implementation notes:
If a network adapter driver needs to fixup the first mbuf in order to
support VLAN tag insertion, the size of the VLAN tag should be
subtracted from the TSO limit. Else not.
Network adapters which typically inline the complete header mbuf could
technically transmit one more segment. This patch does not implement a
mechanism to recover the last segment for data transmission. It is
believed when sufficiently large mbuf clusters are used, the segment
limit will not be reached and recovering the last segment will not
have any effect.
The current TSO algorithm tries to send MTU-sized packets, where the
MTU typically is 1500 bytes, which gives 1448 bytes of TCP data
payload per packet for IPv4. That means if the TSO length limitiation
is set to 65536 bytes, there will be a data payload remainder of
(65536 - 1500) mod 1448 bytes which is equal to 324 bytes. Trying to
recover total TSO length due to inlining mbuf header data will not
have any effect, because adding or removing the ETH/IP/TCP headers
to or from 324 bytes will not cause more or less TCP payload to be
TSO'ed.
Existing network adapter limits will be updated separately.
marius [Sun, 13 Sep 2015 21:59:56 +0000 (21:59 +0000)]
- Sanity check that the parent ranges given in the "ranges" property
of PCI-EBus-bridges actually match the BARs as specified in and
required by [1, p. 113 f.]. Doing so earlier would have simplified
diagnosing a bug in QEMU/OpenBIOS getting the mapping of child
addresses wrong, which still needs to be fixed there.
In theory, we could try to change the BARs accordingly if we hit
this problem. However, at least with real machines changing the
decoding likely won't work, especially if the PCI-EBus-bridge is
beneath an APB one. So implementing such functionality generally
is rather pointless.
- Actually change the allocation type of EBus resources if they
change from SYS_RES_MEMORY to SYS_RES_IOPORT when mapping them
to PCI ranges in ebus_alloc_resource() and passing them up to
bus_activate_resource(9). This may happen with the QEMU/OpenBIOS
PCI-EBus-bridge but not real ones. Still, this is only cleans up
the code and the result of resource allocation and activation is
unchanged.
- Change the remainder of printf(9) to device_printf(9) calls and
canonicalize their wording.
MFC after: 1 week
Peripheral Component Interconnect Input Output Controller,
Part No.: 802-7837-01, Sun Microelectronics, March 1997 [1]
adrian [Sun, 13 Sep 2015 19:17:26 +0000 (19:17 +0000)]
Disable mgmt frame sending in if_rsu.
The firmware in this NIC sends management frames. So far I'm not sure which
ones it handles and which ones it doesn't handle - but this is what openbsd
does.
The association messages are handled by the firmware; the key negotiation
for 802.1x and WPA are done as raw frames, not management frames.
This successfully allows it to associate to my home networks whereas it didn't
work beforehand.
Tested:
* RTL8712, cut 3, STA mode
TODO:
* The firmware does send a join response with a status code; that should be
logged in a more obvious way to assist with debugging. Ie, the firmware
is the thing that is saying "couldn't join, sorry!", not net80211.
Update em(4) with D3162 after testing further on hardware that failed
to attach with the last version of this commit. This commit fixes
attach failures on "ICH8" class devices via modifications to
e1000_init_nvm_params_ich8lan()
- Fix compiler warning in 80003es2lan.c
- Add return value handler for e1000_*_kmrn_reg_80003es2lan
- Fix usage of DEBUGOUT
- Remove unnecessary variable initializations.
- Removed unused variables (complaints from gcc).
- Edit defines in 82571.h.
- Add workaround for igb hw errata.
- Shared code changes for Skylake/I219 support.
- Remove unused OBFF and LTR functions.
Tested by some of the folks that reported breakage in previous incarnation.
Thanks to AllanJude, gjb, gnn, tijl for tempting fate with their machines.
MFV r287623: 5997 FRU field not set during pool creation and never
updated
ZFS already supports storing the vdev FRU in a vdev property. There
is code in libzfs to work with this property, and there is code in
the zfs-retire FMA module that looks for that information. But there
is no code actually setting or updating the FRU.
To address this, ZFS is changed to send a handful of new events
whenever a vdev is added, attached, cleared, or onlined, as well
as when a pool is created or imported.
Note that syseventd is not currently available on FreeBSD and thus
some work is needed to actually support the new ZFS events (e.g. in
zfsd) to actually use this capability, this changeset is mostly a
diff reduction from upstream.
adrian [Sun, 13 Sep 2015 04:12:51 +0000 (04:12 +0000)]
* fiddle with some more of the debugging output
* yes, when a "sta disconnect" message comes through we should, like,
disconnect things. We're not currently generating beacon miss messages,
and net80211 isn't disconnecting things via software beacon miss receive.
marius [Sun, 13 Sep 2015 00:08:04 +0000 (00:08 +0000)]
Merge r286374 from x86:
Formally pair store_rel(&smp_started) with load_acq(&smp_started).
Similarly to x86, this change is mostly a NOP due to the kernel
being run in total store order.
adrian [Sat, 12 Sep 2015 23:10:34 +0000 (23:10 +0000)]
if_rsu debug fixes:
* use an ath/iwn style debug bitmap - it's still global rather than per-device,
but it's better than debug levels
* disable bgscan - it just makes things unstable/unpredictable for now.
marius [Sat, 12 Sep 2015 22:49:32 +0000 (22:49 +0000)]
- Factor out the common and generic parts of the sparc64 host-PCI-bridge
drivers into the revived sys/sparc64/pci/ofw_pci.c, previously already
serving a similar purpose. This has been done with sun4v in mind, which
explains a) the otherwise not that obvious scheme employed and b) why
reusing sys/powerpc/ofw/ofw_pci.c was even lesser an option.
- Add a workaround for QEMU once again not emulating real machines, in
this case by not providing the OFW_PCI_CS_MEM64 range. [1]
Cleanup the handling of error causes for ERROR chunks. This fixes
an inconsistency of the padding handling. The final padding is
now considered to be a chunk padding.
In r286570 (MFV of r277426) an unprotected write to b_flags to
set the compression mode was introduced. This would open a race
window where data is partially decompressed, modified, checksummed
and written to the pool, resulting in pool corruption due to the
partial decompression.
How many demand read didn't have to wait for I/O
because of predictive prefetch. (more is better)
zfetch kstats have been similified to hits, misses, and max_streams,
with max_streams representing times when we were not able to create
new stream because we already have the maximum number of sequences
for a file.
The sysctl variable/loader tunable vfs.zfs.zfetch.block_cap have been
replaced by vfs.zfs.zfetch.max_distance, which controls maximum bytes
to prefetch per stream.
The <arch>/mkisoimages.sh script in release knows how to add
extra bits from an "xtra-bits-dir". This feature is unusable
from release/Makefile. Add an XTRADIR setting to use it.
radeon_suspend_kms: don't mess with pci state that's managed by the bus
The pci bus driver handles the power state, it also manages
configuration state saving and restoring for its child devices. Thus a
PCI device driver does not have to worry about those things. In fact, I
observe a hard system hang when trying to suspend a system with active
radeonkms driver where both the bus driver and radeonkms driver try to
do the same thing. I suspect that it could be because of an access to a
PCI configuration register after the device is placed into D3 state.
Close races between device close and request processing.
All requests arriving for processing after OFFLINE flag set are rejected
with BUSY status. Races around OFFLINE flag setting are closed by calling
taskqueue_drain_all().
Ensure that ERROR chunks are always padded by implementing this
in the routine, which queues an ERROR chunk, instead on relyinh
on the callers to do so. Since one caller missed this, this actially
fixes a bug.
Use strlcpy() in favor of strncpy() as it's defined to have a nul character
at the end of string buffer, and the code context do expects this to behave
correctly (e.g. strchr).
Note that we do not believe there is real-world impact for gstat(8)'s usage
because the strings are length checked, and the on-stack buffer belongs to
main() and we can expect to have zeros in them.
Add stack_save_td_running(), a function to trace the kernel stack of a
running thread.
It is currently implemented only on amd64 and i386; on these
architectures, it is implemented by raising an NMI on the CPU on which
the target thread is currently running. Unlike stack_save_td(), it may
fail, for example if the thread is running in user mode.
This change also modifies the kern.proc.kstack sysctl to use this function,
so that stacks of running threads are shown in the output of "procstat -kk".
This is handy for debugging threads that are stuck in a busy loop.
MFV r283513:
5930 fasttrap_pid_enable() panics when prfind() fails in forking process
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Gordon Ross <gordon.ross@nexenta.com>
Approved by: Richard Lowe <richlowe@richlowe.net>
Author: Bryan Cantrill <bryan@joyent.com>
Handle default MAKEOBJDIR for META_MODE.
If MAKEOBJDIRPREFIX is set, use it for default OBJROOT.
If MAKEOBJDIR is empty or not a suitable value (no '/')
set a default that works.
- Avoid accessing window properties directly, instead, use accessors.
This should be no-op for now, but allows the code to work if we
move to NCURSES_OPAQUE.
- Use calloc() instead of malloc+bzero.