Certain versions of Sandisk x400 firmware can hang under extremely
heavly load of large I/Os for prolonged periods of time. Newer /
current versions work fine, and should be used where possible. Where
not possible, this quirk ensures that I/O requests are limited to 128k
to avoids the bug, even under extreme load. Since MAXPHYS is 128k,
only users with custom kernels are at risk on the older firmware.
Once all known users of the older firmware have upgraded, this quirk
will be removed.
Sponsored by: Netflix, Inc.
r212160 tightened this from always using MSG_SIMPLE_Q_TAG to always
MSG_ORDERED_Q_TAG. Since it also marked all BIO_FLUSH requests with
BIO_ORDERED, this commit changes nothing immediately, but it returns
BIO_FLUSH callers ability to actually specify ordering they really
need, alike to other request types.
MFC after: 2 weeks
Sponsored by: iXsystems, Inc.
It has been reported that on some systems (with real hardware passed
through to a virtual machine) the WP detection causes USB disk probing
failures.
While here, also fix the selection of the next state in the case
of malloc failure in DA_STATE_PROBE_WP. It was DA_STATE_PROBE_RC
unconditionally even when it should have been DA_STATE_PROBE_RC16.
PR: 225794
Reported by: David Boyd <David.Boyd49@twc.com>
MFC after: 3 weeks
Differential Revision: https://reviews.freebsd.org/D18496
Many async events that we see are called for this specific path. When
calling an async callback for a targetted device, XTP will lock that
specific device's path lock (same as what cam_periph_lock does). For
those AC_ events, assert we have the lock rather than trying to
recusrively take it (which causes panics since it's not recursive).
Add annotations about this and about the fact that AC_SCSI_AEN events
are generated now only in the ata stack (which cannot have a scsi_da
attachment). Leave it in place in case I've overlooked something as
the code is harmless.
This is fallout from my attempts to "fix" locking for softc->flags in
r330796 that's not been triggered often enough to get my attention
until now.
Sponsored by: Netflix
MFC After: 3 days
Differential Revision: https://reviews.freebsd.org/D17837
Add a counter for the LBAs, Ranges and hardware commands so that we
can provide additional color to the statistics we provide to vendors.
Sponsored by: Netflix, Inc
scsi_low was a common set of routines to do the SCSI bus sequencing
for the ncv, nsp and stg drivers. Those have been removed, so it's no
longer needed since nothing else in the tree uses it and nothing
likely ever will (it's for super-low-end 8-bit parallel SCSI cards).
ioctl(2) commands only have meaning in the context of a file descriptor
so translating them in the syscall layer is incorrect.
The new handler users an accessor to retrieve/construct a pointer from
the last member of the passed structure and relies on type punning to
access the other members which require no translation.
Reviewed by: kib (prior version), jhb
Approved by: re (rgrimes)
Obtained from: CheriBSD
Sponsored by: DARPA, AFRL
Differential Review: https://reviews.freebsd.org/D17378
In the probe case for SCSI SMR Host Aware or Most Managed drives, be sure
to free allocated memory.
sys/cam/scsi/scsi_da.c:
In dadone_probezone(), free the data pointer before returning.
MFC after: 3 days
Sponsored by: Spectra Logic
Approved by: re (kib)
The idea was to get the uncontroversial mechanical change out of the way,
then get the meatier functional changes reviewed subsequently. I had not
realized that the immediately adjacent issue was addressed in a different
direction in r334506 (see Warner's guidance in D15592).
Discussion continues, trying to determine if there is a secondary issue
still[1] and how best to fix it. With 12-related activities coming up,
while that is ongoing, just take this back for now.
[1]: Shutdown-time eventhandler events fire normally during panic's reboot
path. Driver callbacks that attempt to issue and wait on interrupt-
completed IO may never complete, hanging the system. This is particularly
obnoxious in the shutdown/panic path, as the debugger cannot be entered
anymore and the hang prevents reboot restoring availability.
(There's nothing CAM-specific about this problem -- any shutdown
event-triggered driver could do something like this during panic. But most
NICs, etc. don't try to send spin-down commands at shutdown. ;-))
Discussed with: imp, markj
No functional change.
Note that this change is careful to set the CCB header xflags after
foo_fill_bar() routines, which generally zero existing flags. An earlier
version of this patch mistakenly set the flag before the fill routines.
Submitted by: Scott Ferris <sferris AT isilon.com>, jhibbits@
Reviewed by: bdrewery@, markj@, and non-committer FreeBSD contributor Anton Rang
Sponsored by: Dell EMC Isilon
It's likely that the header was needed in the past for swi(9).
But now that code does not use swi(9) or any other interfaces defined
in sys/interrupt.h.
MFC after: 1 week
Probing host aware and host managed SMR drives got broken in revision
330796.
The added cam_periph_lock() calls were in areas in dadone() where
the peripheral lock was already held.
Since then, dadone() has been split into separate functions that are
dedicated to each probe state.
The result is that when probing a host aware drive, I ran into a recursive
lock acquisition in dadone_probeatalogdir(). I would have run into the
same problem in dadone_probeataiddir(), and in dadone_probeatasup() and
dadone_probeatazone() in the error paths had the probe continued.
The solution is to take out all of the extra cam_periph_lock() calls. I
also added cam_periph_assert(periph, MA_OWNED) near the top of each of
the dadone_* calls. These make it clear to anyone coming along in the
the future that the lock is held in the probe done functions.
Also add a locking assert in daprobedone(), to make it clear that it must
be called with the periph lock held.
Sponsored by: Spectra Logic
Differential Revision: https://reviews.freebsd.org/D15764
When a disk disappears and the periph is invalidated, any I/Os that
are pending with the controller can cause a crash when they
complete. Move to holding the softc reference count taken in dastart()
until the I/O is complete rather than only until xpt_action()
returns. (This approach was suggested by Ken Merry.)
Sponsored by: Netflix
Submitted by: Chuck Silvers
Differential Revision: https://reviews.freebsd.org/D15435
a 1500 line switch statement. Callers now specify a discrete completion
handler, though they're still welcome to track state via ccb_state.
Sponsored by: Netflix
pointer. It's now unhelpful and misleading for callers to continue to set
it, so bring all callers into conformance. There's no real functional change,
but it makes reading the code a lot less confusing.
Sponsored by: Netflix
When the tape position is inside the Early Warning area, the tape
drive will return a sense key of NO SENSE, and an ASC/ASCQ of
0x00,0x02, which means: End-of-partition/medium detected". If
this was in response to a control command like WRITE FILEMARKS,
we correctly translate this as informational status and return
0 from saerror().
Programmable Early Warning should be handled the same way, but
we weren't handling it that way. As a result, if a PEW status
(sense key of NO SENSE, ASC/ASCQ of 0x00,0x07, "Programmable early
warning detected") came back in response to a WRITE FILEMARKS,
we returned an error.
The impact of this was that if an application was writing to a
sa(4) device, and a PEW area was set (in the Device Configuration
Extension subpage -- mode page 0x10, subpage 1), and a filemark
needed to be written on close, we could wind up returning an error
to the user on close because of a "failure" to write the filemarks.
It actually isn't a failure, but rather just a status report from
the drive, and shouldn't be treated as a failure.
sys/cam/scsi/scsi_sa.c:
For control commands in saerror(), treat asc/ascq 0x00,0x07
the same as 0x00,{0-5} -- not an error. Return 0, since
the command actually did succeed.
Reported by: Dr. Andreas Haakh <andreas@haakh.de>
Tested by: Dr. Andreas Haakh <andreas@haakh.de>
Sponsored by: Spectra Logic
MFC after: 3 days
opt_compat.h is mentioned in nearly 180 files. In-progress network
driver compabibility improvements may add over 100 more so this is
closer to "just about everywhere" than "only some files" per the
guidance in sys/conf/options.
Keep COMPAT_LINUX32 in opt_compat.h as it is confined to a subset of
sys/compat/linux/*.c. A fake _COMPAT_LINUX option ensure opt_compat.h
is created on all architectures.
Move COMPAT_LINUXKPI to opt_dontuse.h as it is only used to control the
set of compiled files.
Reviewed by: kib, cem, jhb, jtl
Sponsored by: DARPA, AFRL
Differential Revision: https://reviews.freebsd.org/D14941
have one pending. Otherwise, we can race and send two, which is
wasteful in close proximity. It can also cause the acaquire/release
count for TUR to be > 1, which is undexpected.
PR: 226510
Differential Review: https://reviews.freebsd.org/D14792
do this right, except when there's no BP and we do a TUR by request.
In that case, we clear the flag, but don't release the reference,
leaking the reference on rare occasion.
PR: 226510
Sponsored by: Netflix
<machine/stdarg.h> is a kernel-only header. The standard header for
userland is <stdarg.h>. Using the standard header in userland avoids
weird build errors when building with external compilers that include
their own stdarg.h header.
Reviewed by: arichardson, brooks, imp
Sponsored by: DARPA / AFRL
Differential Revision: https://reviews.freebsd.org/D14776
The ch(4) driver was missed in change 328918, which changed
cam_periph_acquire() to return an errno instead of cam_status.
As a result, ch(4) failed to attach.
Sponsored by: Spectra Logic
softc for an async unit attention. CAM locks, sometimes, the periph
lock and other times does not. We were taking the lock always and
running into lock recursion issues on a non-recursive lock. Now we
take it selectively. It's not clear why xpt takes the lock selectively
before calling us, though, and that's still under investigation.
Reported by: avg
PR: 226510 (same panic, differnt circumstances)
Sponsored by: Netflix
kern.cam.{,a,n}da.X.invalidate=1 forces *daX to detach by calling
cam_periph_invalidate on the underlying periph. This is for testing
purposes only. Include only with options CAM_TEST_FAILURE and rename
the former [AN]DA_TEST_FAILURE, and fix nda to compile with it set.
We're using it at work to harden geom and the buffer cache to be
resilient in the face of drive failure. Today, it far too often
results in a panic. While much work was done on SIM initiated removal
for the USB thumnb drive removal work, little has been done for periph
initiated removal. This simulates what *daerror() does for some errors
nicely: we get the same panics with it that we do with failing drives.
Sponsored by: Netflix
Differential Revision: https://reviews.freebsd.org/D14581
The ioctl objects contain pointers and require translation and some
refactoring of the infrastructure to work. For now prevent opertion
on garbage values. This is very slightly overbroad in that ENCIOC_INIT
is safe.
Reviewed by: imp, kib
Obtained from: CheriBSD
Sponsored by: DARPA, AFRL
Differential Revision: https://reviews.freebsd.org/D14671
These take a union ccb argument which is full of kernel pointers.
Substantial translation efforts would be required to make this work.
By rejecting the request we avoid processing or returning entierly
wrong data.
Reviewed by: imp, ken, markj, cem
Obtained from: CheriBSD
MFC after: 1 week
Sponsored by: DARPA, AFRL
Differential Revision: https://reviews.freebsd.org/D14654
Make sure the periph lock is held around rmw access to softc data,
espeically flags, including work flags in iosched.
Add asserts for the periph lock where it should be held.
PR: 226510
Sponsored by: Netflix
Differential Review: https://reviews.freebsd.org/D14456
Introduce flags word to describe the capacities of the peripheral.
First bit will describe if the periph driver allows multiple
outstanding TRIMS to be active in a device.
Modify the I/O scheduler so that the nda driver can queue trims
for a while after the first one arrives. We'll queue until we see
a I/O scheduler tick, then we'll schedule as many TRIMs as allowed
by other factors (currently this is slocts in the NVMe controller).
This mariginally helps the read latency issues we see with reads,
but sets the stage for the nda driver to do TRIM collapsing like the
da and ada drivers do today.
Sponsored by: Netflix
There's no compelling reason to return a cam_status type for this
function and doing so only creates confusion with normal C
coding practices. It's technically an API change, but the periph API
isn't widely used. No efffective change to operation.
Reviewed by: imp, mav, ken
Sponsored by: Netflix
Differential Revision: D14063
leaks. We assume each source can be taken / dropped only once and
don't recurse. These are only enabled via DA_TRACK_REFS or
INVARIANTS. There appreas to be a reference leak under extreme load,
and these should help us colaberatively work it out. It also documents
better the reference / holding protocol better.
Reviewed by: ken@, scottl@
Sponsored by: Netflix
Differential Revision: https://reviews.freebsd.org/D14040
