All resources lack of which may put CTIO into the queue are either
per-channel or potentially per-queue, but none of them are per-LUN.
This is a first step to fix live LUN disabling. Before this change
any CTIOs held in a queue in time of disabling were just leaked.
MFC after: 2 weeks
The isp(4) driver was changing the tag type for REQUEST SENSE
commands to Head of Queue, when the CAM CCB flag
CAM_TAG_ACTION_VALID was NOT set. CAM_TAG_ACTION_VALID is set
when the tag action in the XPT_SCSI_IO is not CAM_TAG_ACTION_NONE
and when the target has tagged queueing turned on.
In most cases when CAM_TAG_ACTION_VALID is not set, it is because
the target is not doing tagged queueing. In those cases, trying to
send a Head of Queue tag may cause problems. Instead, default to
sending a simple tag.
IBM tape drives claim to support tagged queueing in their standard
Inquiry data, but have the DQue bit set in the control mode page
(mode page 10). CAM correctly detects that these drives do not
support tagged queueing, and clears the CAM_TAG_ACTION_VALID flag
on CCBs sent down to the drives.
This caused the isp(4) driver to go down the path of setting the
tag action to a default value, and for Request Sense commands only,
set the tag action to Head of Queue.
If an IBM tape drive does get a Head of Queue tag, it rejects it with
Invalid Message Error (0x49,0x00). (The Qlogic firmware translates that
to a Transport Error, which the driver translates to an Unrecoverable
HBA Error, or CAM_UNREC_HBA_ERROR.) So, by default, it wasn't possible
to get a good response from a REQUEST SENSE to an FC-attached IBM
tape drive with the isp(4) driver.
IBM tape drives (tested on an LTO-5 with G9N1 firmware and a TS1150
with 4470 firmware) also have a bug in that sending a command with a
non-simple tag attribute breaks the tape drive's Command Reference
Number (CRN) accounting and causes it to ignore all subsequent
commands because it and the initiator disagree about the next
expected CRN. The drives do reject the initial command with a head
of queue tag with an Invalid Message Error (0x49,0x00), but after that
they ignore any subsequent commands. IBM confirmed that it is a bug,
and sent me test firmware that fixes the bug. However tape drives in
the field will still exhibit the bug until they are upgraded.
Request Sense is not often sent to targets because most errors are
reported automatically through autosense in Fibre Channel and other
modern transports. ("Modern" meaning post SCSI-2.) So this is not
an error that would crop up frequently. But Request Sense is useful on
tape devices to report status information, aside from error reporting.
This problem is less serious without FC-Tape features turned on,
specifically precise delivery of commands (which enables Command
Reference Numbers), enabled on the target and initiator. Without
FC-Tape features turned on, the target would return an error and
things would continue on.
And it also does not cause problems for targets that do tagged
queueing, because in those cases the isp(4) driver just uses the
tag type that is specified in the CCB, assuming the
CAM_TAG_ACTION_VALID flag is set, and defaults to sending a Simple
tag action if it isn't an ordered or head of queue tag.
sys/dev/isp/isp.c:
In isp_start(), don't try to send Request Sense commands
with the Head of Queue tag attribute if the CCB doesn't
have a valid tag action. The tag action likely isn't valid
because the target doesn't support tagged queueing.
Sponsored by: Spectra Logic
MFC after: 3 days
Replace archaic "busses" with modern form "buses."
Intentionally excluded:
* Old/random drivers I didn't recognize
* Old hardware in general
* Use of "busses" in code as identifiers
No functional change.
http://grammarist.com/spelling/buses-busses/
PR: 216099
Reported by: bltsrc at mail.ru
Sponsored by: Dell EMC Isilon
The sim_vid, hba_vid, and dev_name fields of struct ccb_pathinq are
fixed-length strings. AFAICT the only place they're read is in
sbin/camcontrol/camcontrol.c, which assumes they'll be null-terminated.
However, the kernel doesn't null-terminate them. A bunch of copy-pasted code
uses strncpy to write them, and doesn't guarantee null-termination. For at
least 4 drivers (mpr, mps, ciss, and hyperv), the hba_vid field actually
overflows. You can see the result by doing "camcontrol negotiate da0 -v".
This change null-terminates those fields everywhere they're set in the
kernel. It also shortens a few strings to ensure they'll fit within the
16-character field.
PR: 215474
Reported by: Coverity
CID: 1009997 1010000 1010001 1010002 1010003 1010004 1010005
CID: 1331519 1010006 1215097 1010007 1288967 1010008 1306000
CID: 1211924 1010009 1010010 1010011 1010012 1010013 1010014
CID: 1147190 1010017 1010016 1010018 1216435 1010020 1010021
CID: 1010022 1009666 1018185 1010023 1010025 1010026 1010027
CID: 1010028 1010029 1010030 1010031 1010033 1018186 1018187
CID: 1010035 1010036 1010042 1010041 1010040 1010039
Reviewed by: imp, sephe, slm
MFC after: 4 weeks
Sponsored by: Spectra Logic Corp
Differential Revision: https://reviews.freebsd.org/D9037
Differential Revision: https://reviews.freebsd.org/D9038
FC-Tape provides additional link level error recovery, and is
highly recommended for tape devices. It will only be turned on for
a given target if the target supports it.
Without this setting, we default to whatever FC-Tape setting is in
NVRAM on the card.
This can be overridden by setting the following loader tunable, for
example for isp0:
hint.isp.0.nofctape=1
sys/conf/options:
Add a new kernel config option, ISP_FCTAPE_OFF, that
defaults the FC-Tape configuration to off.
sys/dev/isp/isp_pci.c:
If ISP_FCTAPE_OFF is defined, turn off FC-Tape. Otherwise,
turn it on if the card supports it.
share/man/man4/isp.4:
Add a description of FC-Tape to the isp(4) man page.
Add descriptions of the fctape and nofctape options, as well as the
ISP_FCTAPE_OFF kernel configuration option.
Add the ispfw module and kernel drivers to the suggested
configurations at the top of the man page so that users are less
likely to leave it out. The driver works well with the included
firmware, but may not work at all with whatever firmware the user
has flashed on their card.
MFC after: 3 days
Sponsored by: Spectra Logic
utilizing previously unused arg field of struct ccb_notify_acknowledge.
This makes new QUERY TASK, QUERY TASK SET and QUERY ASYNC EVENT requests
really functional for CAM target mode drivers.
This code was originally implemented 7 years ago, but never really worked
due to trivial error. I think this functionality may be not required.
Initiators supporting optional periodic command status checks detected
those terminated commands and retried them 3 seconds later. But thinking
about less featured initiators and the fact that it is our race makes
virtual ports "unknown" it may be good to have this feature.
It is normal for ZOMBIE ports to be logged out. This status is not really
an error until Gone Device Timeout expires, so make CAM retry after delay.
MFC after: 1 week
Firmware automatically logs in only to local loop ports, and those ports
can be easily identified without extra flag by zero domain and area IDs.
MFC after: 1 week
Since we no longer need additional buffers for request and response IOCBs,
we can increase receive space by 192 bytes, that is enough for fetching 48
more ports. The new limit is 1020 fabric ports per virtual port.
MFC after: 1 month
This should close the race between request arriving on new target mode
virtual port and its scanner thread finally fetch its address for request
routing.
For some reason firmware sends Port Database Changed notifications in case
of explicit login requests from the driver when target port is unavailabe.
Those notifications don't give driver any new information, but only cause
infinite scan loop.
Previously we had to do it synchronously because we could not drop the lock
due to potential scratch memory use conflicts. Previous commits fixed that
collision, so here it goes -- slower and less reliable external requests
are executed asynchronously without spinning in tight loop and with more
safe timeout handling.
Usually IOCBs should be put on queue for asynchronous processing and should
not require additional DMA memory. But there are some cases like aborts and
resets that for external reasons has to be synchronous. Give those cases
separate 2*64 byte DMA area to decouple them from other DMA scratch area
users, using it for asynchronous requests.
This is cosmetics that simplifies identification of new ports on FC switch.
It would be good to use target name from CTL here instead of hostname, but
it is not passed here through CAM now.
MFC after: 2 weeks
This space does not require DMA syncing. It reduces lock scope of the DMA
scratch space. It allows whole DMA scratch space to be used to I/O, so now
we can fetch up to ~1000 ports from SNS.
Due to the last fact, increase maximal number of ports from 256 to 1024.
Before this change virtual ports control IOCBs were executed synchronously
via Execute IOCB mailbox command. It required exclusive use of scratch
space of driver and mailbox registers of the hardware. Because of that
shared resources use this code could not really sleep, having to spin for
completion, blocking any other operation.
This change introduces new asynchronous design, sending the IOCBs directly
on request queue and gracefully waiting for their return on response queue.
Returned IOCBs are identified with unified handle space from r292725.
I am not sure why this was split long ago, but I see no reason for it.
At this point this unification just slightly reduces memory usage, but
as next step I plan to reuse shared handle space for other IOCB types.
- Make scan aborted by event restart immediately and infinitely.
- Improve handling of some loop events from firmware.
- Remove loop down timer, adding its functionality to scanner thread.
- Some more unification and simplification.
Hacks to enable target mode there complicated code, while didn't really
work. And for outdated hardware fixing it is not really interesting.
Initiator mode tested with Qlogic 1080 adapter is still working fine.
While later firmware always registers for RSCN requests, older one does
it only in initiator mode. But in target mode there RSCN can be the only
way to detect gone intiator.
For those chips we are not receiving login events, adding initiators
based on ATIO requests. But there is no port ID in that structure, so
in fabric mode we have to explicitly fetch it from firmware to be able
to do normal scan after that.
This change simplifies and unifies port adding/updating for loop and
fabric scanners. It also fixes problems with scanning restarts due to
concurrent port databases changes. It also fixes many cosmetic issues.
Modern cards in most cases operate abstract port handles, that have no
any relation to real loop IDs. Leave loopid used only where it really
goes about local loop IDs.
While there, fix few more cases where LUNs were still printed in decimal.
For the most of chips (except anscient ones) port handlers have no relation
to port IDs. In such situation old code scanning first 125 handlers was
quite naive. Instead of doing that, send to chip single request to get full
list of port handlers available on specific virtual port and scan only them.
Old code had problems with case of several virtual ports enabled, when port
handlers allocated from global address space could easily go above 125.
This change was successfully tested on 23xx, 24xx and 25xx chips in loop
mode with 4 virtual initiator ports, each seing 50 virtual target ports.
Now on 24xx and above chips it is really possible to simulate several
virtual FC ports with single physical one. For example, it allows to
configure several targets in ctl.conf, assign each of them to separate
virtual port, and let user to control access to them with switch zoning.
I still doubt that all problems are solved there, but at now it passes
at least basic tests.
Now 24xx and above chips support full 8-byte LUN address space.
Older FC chips may support up to 16K LUNs when firmware allows.
Tested in both initiator and target modes for 23xx, 24xx and 25xx.
This change allows to decode respective functions in isp(4) in target mode
and pass them through CAM to CTL. Unfortunately neither CAM nor isp(4)
support returning response info for those task management functions now.
On the other side I just have no initiator to test this functionality.
FC port database code already notifies CAM about all devices. Additional
full scan is just a waste of time, that by definition won't find anything
that is not present in port database.
IDs in initiator mode -- index in port database instead of handlers.
This makes initiator IDs persist across role changes and firmware resets,
when handlers previously assigned by firmware are lost and reused.
Sponsored by: iXsystems, Inc.
Aside from cleaner and more consistent code, this allows ports to be both
target and initiator same time, and easily switch from any role to any.
Sponsored by: iXsystems, Inc.
To avoid conflicts between target and initiator devices in CAM, make
CTL use target ID reported by HBA as its initiator_id in XPT_PATH_INQ.
That target ID is known to never be used for initiator role, so it won't
conflict. For Fibre Channel and FireWire HBAs this specific ID choice
is irrelevant since all target IDs there are virtual. Same time for SPI
HBAs it seems could be even requirement to use same target ID for both
initiator and target roles.
While there are some more things to polish in isp(4) driver, first tests
of using both roles same time on the same port appeared successfull:
# camcontrol devlist -v
scbus0 on isp0 bus 0:
<FREEBSD CTLDISK 0001> at scbus0 target 1 lun 0 (da20,pass21)
<> at scbus0 target 256 lun 0 (ctl0)
<> at scbus0 target -1 lun ffffffff (ctl1)
FreeBSD never had limitation on number of target IDs, and there is no
any other requirement to allocate them densely. Since slots of port
database already populated just sequentially, there is no much need
for another indirection to allocate sequentially too.
It is odd that driver first tries to generate synthetic WWNN based on
WWPN starting from 2, but then refuses to use it. If we don't trust
generated WWNN, we should probably not generate it. Same time this
limitation prevents potentially valid WWNN setting by user.
This makes driver better track reported CAM_SEL_TIMEOUTs to properly
report device reappearance later. This fixes target 0 not reappearing
after initiator mode disabled and then reenabled.
MFC after: 3 days
It (mostly) allows to enable/disable initiator mode in run time.
Target mode control is blocked here to force coordination with CTL.
While there, add separate tunables/sysctls for virtual channels.
Previous implementation was too fragile to initiator parameters changes.
In case of port role change it could not survive different handle assigned
to the same initiator by firmware, even though initiator was logged out.
The new implementation should be more resillient to this kind of problems,
trying to work in any situation and only warn user about suspisious events.
MFC after: 1 week
Sponsored by: iXsystems, Inc.
ct_rxid value on CTIO completion. Try to workaround that using tag_id
from the CCB, pointed by still valid ct_syshandle.
I don't know whether this is valid fix or dirty hack, but considering that
alternative is indefinitely stuck command -- it worth trying.
MFC after: 1 week
The newer boards don't have the response field that indicates
whether the SCSI status byte is present. You have to just look to
see whether it is non-zero.
The code was looking to see whether the sense length was valid
before propagating the SCSI status byte (and sense information) up
the stack. With a status like Reservation Conflict, there is no
sense information, only the SCSI status byte. So it wasn't getting
correctly returned.
isp.c:
In isp_intr(), if we are on a 2400 or 2500 type board and
get a response, look at the actual contents of the
SCSI status value and set the RQSF_GOT_STATUS flag
accordingly so that return any SCSI status value we get. The
RQSF_GOT_SENSE flag will get set later on if there is
actual sense information returned.
Submitted by: ken
MFC after: 1 week
Sponsored by: Spectra Logic
MFSpectraBSD: 1112791 on 2015/01/15
If the user sends an XPT_RESET_DEV CCB, make sure to reset the
Fibre Channel Command Reference Number if we're running on a FC
controller.
We send a SCSI Target Reset when we get this CCB, and as a result
need to reset the CRN to 1 on the next command.
isp_freebsd.c:
In the XPT_RESET_DEV implementation in isp_action(), reset
the CRN if we're on a FC controller.
Submitted by: ken
MFC after: 1 week
Sponsored by: Spectra Logic
MFSpectraBSD: 1112787 on 2015/01/15
Fix SCSI status byte reporting on 4Gb and 8Gb Qlogic boards.
The newer boards don't have the response field that indicates
whether the SCSI status byte is present. You have to just look to
see whether it is non-zero.
The code was looking to see whether the sense length was valid
before propagating the SCSI status byte (and sense information) up
the stack. With a status like Reservation Conflict, there is no
sense information, only the SCSI status byte. So it wasn't getting
correctly returned.
isp.c:
In isp_intr(), if we are on a 2400 or 2500 type board and
get a response, look at the actual contents of the
SCSI status value and set the RQSF_GOT_STATUS flag
accordingly so that return any SCSI status value we get. The
RQSF_GOT_SENSE flag will get set later on if there is
actual sense information returned.
Submitted by: ken
MFC after: 1 week
Sponsored by: Spectra Logic
MFSpectraBSD: 1112791 on 2015/01/15
and not automatically come back if they were gone for a short
period of time.
The isp(4) driver has a 30 second gone device timer that gets
activated whenever a device goes away. If the device comes back
before the timer expires, we don't send a notification to CAM that
it has gone away. If, however, there is a command sent to the
device while it is gone and before it comes back, the isp(4) driver
sends the command back with CAM_SEL_TIMEOUT status.
CAM responds to the CAM_SEL_TIMEOUT status by removing the device.
In the case where a device comes back within the 30 second gone
device timer window, though, we weren't telling CAM the device
came back.
So, fix this by tracking whether we have told CAM the device is
gone, and if we have, send a rescan if it comes back within the 30
second window.
ispvar.h:
In the fcportdb_t structure, add a new bitfield,
reported_gone. This gets set whenever we return a command
with CAM_SEL_TIMEOUT status on a Fibre Channel device.
isp_freebsd.c:
In isp_done(), if we're sending CAM_SEL_TIMEOUT for for a
command sent to a FC device, set the reported_gone bit.
In isp_async(), in the ISPASYNC_DEV_STAYED case, rescan the
device in question if it is mapped to a target ID and has
been reported gone.
In isp_make_here(), take a port database entry argument,
and clear the reported_gone bit when we send a rescan to
CAM.
In isp_make_gone(), take a port database entry as an
argument, and set the reported_gone bit when we send an
async event telling CAM consumers that the device is gone.
Sponsored by: Spectra Logic
MFC after: 1 week
The Command Reference Number is used for precise delivery of
commands, and is part of the FC-Tape functionality set. (This is
only enabled for devices that support precise delivery of commands.)
It is an 8-bit unsigned number that increments from 1 to 255. The
commands sent by the initiator must be processed by the target in
CRN order if the CRN is non-zero.
There are certain scenarios where the Command Reference Number
sequence needs to be reset. When the target is power cycled, for
instance, the initiator needs to reset the CRN to 1. The initiator
will know this because it will see a LIP (when directly connected)
or get a logout/login event (when connected to a switch).
The isp(4) driver was not resetting the CRN when a target
went away and came back. When it saw the target again after a
power cycle, it would continue the CRN sequence where it left off.
The target would ignore the command because the CRN sequence is
supposed to be reset to 1 after a power cycle or other similar
event.
The symptom that the user would see is that there would be lots of
aborted INQUIRY commands after a tape library was power cycled, and
the library would fail to probe. The INQUIRY commands were being
ignored by the tape drive due to the CRN issue mentioned above.
isp_freebsd.c:
Add a new function, isp_fcp_reset_crn(). This will reset
all of the CRNs for a given port, or the CRNs for all LUNs
on a target.
Reset the CRNs for all targets on a port when we get a LIP,
loop reset, or loop down event.
Reset the CRN for a particular target when it arrives, is changed
or departs. This is less precise behavior than the
clearing behavior specified in the FCP-4 spec (which says
that it should be reset for PRLI, PRLO, PLOGI and LOGO),
but this is the level of information we have here. If this
is insufficient, then we will need to add more precise
notification from the lower level isp(4) code.
isp_freebsd.h:
Add a prototype for isp_fcp_reset_crn().
Sponsored by: Spectra Logic
MFC after: 1 week
Records with target_mode == 1 are allocated from the end of portdb, so it
seems logical to start search from the end not traverse whole array.
MFC after: 1 month
In the current implementation, the isp_kthread() threads never exit.
The target threads do have an exit mode from isp_attach(), but it is
not invoked from isp_detach().
Ensure isp_detach() notifies threads started for each channel, such
that they exit before their parent device softc detaches, and thus
before the module does. Otherwise, a page fault panic occurs later in:
sysctl_kern_proc
sysctl_out_proc
kern_proc_out
fill_kinfo_proc
fill_kinfo_thread
strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
For isp_kthread() (and isp(4) target threads), td->td_wmesg references
now-unmapped memory after the module has been unloaded. These threads
are typically msleep()ing at the time of unload, but they could also
attempt to execute now-unmapped code segments.
MFC after: 1 month
Sponsored by: Spectra Logic
MFSpectraBSD: r1070921 on 2014/06/22 13:01:17
Delaying isp_reqodx update, we should be ready to update it every time
we read it. Otherwise requests using several indexes may be requeued
ndefinitely without ever updating the variable.
MFC after: 3 days
- Process ATIO queue only if interrupt status tells so;
- Do not update queue out pointers after each processed command, do it
only once at the end of the loop.
every time. The purpose of that register is unlikely output queue overflow
detection, so read it only when its last known (and probably stale now)
value signals overflow.
This reduces CPU load and lock congestion and rises bottleneck in CTL
while doing target mode via two 8Gbps ports from 100K to 120K IOPS.
mostly by adjustments to debugging printf() format specifiers. For high
numbered LUNs, also switch to printing them in hex as per SAM-5.
MFC after: 2 weeks
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
- Remove two excessive and slow register reads from isp_intr(). Instead
of rereading value every time, assume that registers contain what we have
written there.
- Avoid sequential search through 4096 array elements when looking for
command tag. Use hash of lists to store active tags separately from free
ones and so greatly speedup the searches.
Reviewed by: mjacob
driver.
This tells consumers up the stack the maximum I/O size that the
controller can handle.
The I/O size is bounded by the number of scatter/gather segments
the controller can handle and the page size. For an amd64 system,
it works out to around 5MB.
Reviewed by: mjacob
MFC after: 3 days
Sponsored by: Spectra Logic
command register. The lazy BAR allocation code in FreeBSD sometimes
disables this bit when it detects a range conflict, and will re-enable
it on demand when a driver allocates the BAR. Thus, the bit is no longer
a reliable indication of capability, and should not be checked. This
results in the elimination of a lot of code from drivers, and also gives
the opportunity to simplify a lot of drivers to use a helper API to set
the busmaster enable bit.
This changes fixes some recent reports of disk controllers and their
associated drives/enclosures disappearing during boot.
Submitted by: jhb
Reviewed by: jfv, marius, achadd, achim
MFC after: 1 day
a mailbox command and which registers to copy back in when
the command completes, the bits being set need to not only
specify what bits you want to add from the default from the
table but also what bits you want *subtract* (mask) from the
default from the table.
A failing ISP2200 command pointed this out.
Much appreciation to: marius, who persisted and narrowed down what
the failure delta was, and shamed me into actually fixing it.
MFC after: 1 week
