In 10.1-RELEASE the default number of MSI vectors used was changed from one
to as many vectors as the HW supports.
This change resulted in an ahci timeouts regression when running on AMD
SB7x0/SB8x0/SB9x0 hardware, so its now limited to 1 MSI by default using
this new quirk.
MFC after: 2 weeks
Sponsored by: Multiplay
Previously ahci_attach returned a hard coded ENXIO instead of the value
from ahci_setup_interrupt. This is effectively a NOOP change as currently
ahci_setup_interrupt only ever returns 0 or ENXIO, so just there to protect
against any future changes to that.
Differential Revision: D838
MFC after: 2 weeks
Sponsored by: Multiplay
This prevents the possiblity of any overruns on the statically allocated
struct irqs field.
Differential Revision: D838
MFC after: 2 weeks
X-MFC-With: r276012
Sponsored by: Multiplay
Previously, any timeout value for which (timeout * hz) will overflow the
signed integer, will give weird results, since callout(9) routines will
convert negative values of ticks to '1'. For unsigned integer overflow we
will get sufficiently smaller timeout values than expected.
Switch from callout_reset, which requires conversion to int based ticks
to callout_reset_sbt to avoid this.
Also correct isci to correctly resolve ccb timeout.
This was based on the original work done by Eygene Ryabinkin
<rea@freebsd.org> back in 5 Aug 2011 which used a macro to help avoid
the overlow.
Differential Revision: https://reviews.freebsd.org/D1157
Reviewed by: mav, davide
MFC after: 1 month
Sponsored by: Multiplay
It was found that VirtualBox' AHCI does not allow nterrupt to be cleared
before the interrupt status register is read, causing interrupt storm.
AHCI specification allows to skip this register use when multi-vector MSI
is enabled and so interrupting port is known. For single-vector MSI that
is not stated explicitly, but if the port is only one, it is obviously
known too.
Previously it was possible for issues e.g. use after free, to result
from processing the done queue while not holding the channel lock.
While this should never happen in practice, unexpected code flows
which result in two threads processing from the same queue may
be possible.
We now use a local STAILQ to prevent this ever being an issue.
Sponsored by: Multiplay
For controllers with only one port (like PCIe or M.2 SSDs) interrupt can
come from only one source, and skipping read saves few percents of CPU time.
MFC after: 1 month
H/W donated by: I/O Switch
I gave up to update list of Marvell chips that require this quirk.
The final nail was growing number of PCIe/M.2 SSDs where Marvell chips
have PCI IDs of different vendors.
MFC after: 1 week
H/W donated by: I/O Switch
PCI IDs into quirks, which mostly fit (though you'd get no argument
from me that AHCI_Q_SATA1_UNIT0 is oddly specific). Set these quirks
in the PCI attachment. Make some shared functions public so that PCI
and possibly other bus attachments can use them.
The split isn't perfect yet, but it is functional. The split will be
perfected as other bus attachments for AHCI are written.
Sponsored by: Netflix
Reviewed by: kan, mav
Differential Revision: https://reviews.freebsd.org/D699
initializing 16. Initialize all 20 so we don't send garbage in the
Auxiliary register. The SATA standard mandates a 5 dword length for
the Host to Device FIS.
Sponsored by: Netflix
Using unmapped BIOs causes failure inside bus_dmamap_sync, since
this function requires valid MVA address, which is not present
if mapping is not set up.
Submitted by: Wojciech Macek <wma@semihalf.com>
Obtained from: Semihalf
shifts into the sign bit. Instead use (1U << 31) which gets the
expected result.
This fix is not ideal as it assumes a 32 bit int, but does fix the issue
for most cases.
A similar change was made in OpenBSD.
Discussed with: -arch, rdivacky
Reviewed by: cperciva
reduce lock congestion and improve SMP scalability of the SCSI/ATA stack,
preparing the ground for the coming next GEOM direct dispatch support.
Replace big per-SIM locks with bunch of smaller ones:
- per-LUN locks to protect device and peripheral drivers state;
- per-target locks to protect list of LUNs on target;
- per-bus locks to protect reference counting;
- per-send queue locks to protect queue of CCBs to be sent;
- per-done queue locks to protect queue of completed CCBs;
- remaining per-SIM locks now protect only HBA driver internals.
While holding LUN lock it is allowed (while not recommended for performance
reasons) to take SIM lock. The opposite acquisition order is forbidden.
All the other locks are leaf locks, that can be taken anywhere, but should
not be cascaded. Many functions, such as: xpt_action(), xpt_done(),
xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM
lock to be held.
To keep compatibility and solve cases where SIM lock can't be dropped, all
xpt_async() calls in addition to xpt_done() calls are queued to completion
threads for async processing in clean environment without SIM lock held.
Instead of single CAM SWI thread, used for commands completion processing
before, use multiple (depending on number of CPUs) threads. Load balanced
between them using "hash" of the device B:T:L address.
HBA drivers that can drop SIM lock during completion processing and have
sufficient number of completion threads to efficiently scale to multiple
CPUs can use new function xpt_done_direct() to avoid extra context switch.
Make ahci(4) driver to use this mechanism depending on hardware setup.
Sponsored by: iXsystems, Inc.
MFC after: 2 months
extra read from PxCI/PxSACT registers. If only NCQ commands are running, we
don't really need PxCI. If only non-NCQ commands are running we don't need
PxSACT. Mixed set may happen only on controllers with FIS-based switching
when port multiplier is attached, and then we have to read both registers.
MFC after: 1 month
every architecture's busdma_machdep.c. It is done by unifying the
bus_dmamap_load_buffer() routines so that they may be called from MI
code. The MD busdma is then given a chance to do any final processing
in the complete() callback.
The cam changes unify the bus_dmamap_load* handling in cam drivers.
The arm and mips implementations are updated to track virtual
addresses for sync(). Previously this was done in a type specific
way. Now it is done in a generic way by recording the list of
virtuals in the map.
Submitted by: jeff (sponsored by EMC/Isilon)
Reviewed by: kan (previous version), scottl,
mjacob (isp(4), no objections for target mode changes)
Discussed with: ian (arm changes)
Tested by: marius (sparc64), mips (jmallet), isci(4) on x86 (jharris),
amd64 (Fabian Keil <freebsd-listen@fabiankeil.de>)
