src/sys/dev/mpr/mpr.c, branch releng/12.2

MFC r363608:

2020-08-03T23:06:33+00:00

mpr(4), mps(4): Stop checking for failures from malloc(M_WAITOK).

Catch up with stability fixes to mpr/mps in -current since 12

2019-09-26T01:54:24+00:00

was branched.

MFC r341755,r342354,r342355,r342386,r342387,r342388,r342526,r342528,
r342530,r342531,r342533,r342534,r342535,r342536,r342659,r345479,
r345485,r345573,r347237,r349849,r349909:

  r341755:
    Copy and clear the reply descriptor atomically.  This prevents concurrency
    in the interrupt handlers (usually due to timeout/error recovery) from
    seeing and processing the same descriptor twice.
  r342354:
    mps(4), mpr(4): Fix lifetime of command buffer for mp?sas_get_sata_identify
  r342355:
    mps(4), mpr(4): remove SATA ID command cancellation hack
  r342386:
    First step in refactoring and fixing the error recovery and task management
    code in the mpr and mps drivers.  Eliminate duplicated code and fix some
    comments.
  r342387:
    Commands for user-initated device resets should come from the high-priority
    allocator.  Prior to this change, they would leak from the normal
    allocator.
  r342388:
    Further refactoring for task management commands.  Also fix a related typo
    from the previous commit.
  r342526:
    Update MPI header files version to 2.00.52 from 2.00.48
  r342528:
    Copy back the Sense data at proper location expected by the application
  r342530:
    Added support for NVMe Task Management
  r342531:
    Add Sea,Aero adapter support
  r342533:
    Enable atomic type descriptor support only for Sea & Aero cards
  r342534:
    On Aero/Sea A0 cards retry MPT Fusion registers reads for max three times
  r342535:
    Update copyright information
  r342536:
    Update mpr driver version to v23.00.00.00-fbsd
  r342659:
    Fix whitespace from r342528
  r345479:
    r329522 created problemss with commands that enter the TIMEDOUT state but
    are successfully returned by the card (usually due to an abort being issued
    as part of timeout recovery). Remove what amounts to an insufficient
    KASSERT, and don't overwrite the state value. State should probably be
    re-designed, and that will be done with a future commit.
  r345485:
    Add event table decoding for SAS Broadcast Primitive events.
  r345573:
    Add missing break statements.  Coverity CID 1400446.
  r347237:
    Add missing newline to debug printf.
  r349849:
    Fix bugs in recovery path and improve cm tracking
  r349909:
    More fully implement the state machine.

Sponsored by: Netflix, Inc

Implement a sysctl to dump in-flight I/O state for debugging. The tool to

2018-03-12T05:02:22+00:00

parse it will be committed in a separate action.

Sponsored by:	Netflix

Allow physically non-contiguous chain frames allocation in mps(4)/mpr(4).

2018-02-27T01:48:13+00:00

Chain frames required to satisfy all 2K of declared I/Os of 128KB each take
more then a megabyte of a physical memory, all of which existing code tries
allocate as physically contiguous.  This patch removes that physical
contiguousness requirement, leaving only virtual contiguousness.  I was
thinking about other ways of allocation, but the less granular allocation
becomes, the bigger is the overhead and/or complexity, reaching about 100%
overhead if allocate each frame separately.

The patch also bumps the chain frames hard limit from 2K to 16K.  It is more
than enough for the case of default REQ_FRAMES and MAXPHYS (the drivers will
allocate less than that automatically), while in case of increased MAXPHYS
it will control maximal memory usage.

Sponsored by:	iXsystems, Inc.
Differential Revision:	https://reviews.freebsd.org/D14420

Improve command lifecycle debugging and detection of problems.

2018-02-18T16:41:34+00:00

Sponsored by:	Netflix

Fix non-64-bit platform build by printing bus_addr_t values using %#jx

2018-02-13T16:26:06+00:00

Reviewed by:	slm
Differential Revision:	https://reviews.freebsd.org/D14344

Print out the shared memory queues during initialization

2018-02-11T20:15:47+00:00

Sponsored by:	Netflix

Teach mps(4) and mpr(4) drivers to autotune chain frames.

2018-02-10T00:55:46+00:00

This is a first part of the change.  It makes the drivers to calculate
the required number of chain frames to satisfy worst case scenarios, but
it does not change existing overly strict limits on them.  The next step
will be to rewrite the allocator to not require megabytes of physically
contiguous address space, that may be problematic if done after boot,
after doing which the limits can be removed.  Until that this code can
just correct user set limits, if they are set too high.

Sponsored by:	iXsystems, Inc.
Differential Revision:	https://reviews.freebsd.org/D14261

Cache the value of the request and reply frame size since it's used quite

2018-02-06T21:01:38+00:00

a bit in the normal operation of the driver.  Covert it to represent bytes
instead of 32bit words.  Fix what I believe to be is a bug in this respect
with the Tri-mode cards.

Sponsored by:	Netflix

Fix queue length reporting in mps(4) and mpr(4).

2018-02-06T16:02:25+00:00

Both drivers were found to report CAM bigger queue depth then they really
can handle.  It made them later under high load with many disks return
some of submitted requests back with CAM_REQUEUE_REQ status for later
resubmission.

Reviewed by:	scottl
MFC after:	1 week
Sponsored by:	iXsystems, Inc.
Differential Revision:	https://reviews.freebsd.org/D14215