2005-01-05 22:34:37 +00:00
|
|
|
/*-
|
Rewrite of the CAM error recovery code.
Some of the major changes include:
- The SCSI error handling portion of cam_periph_error() has
been broken out into a number of subfunctions to better
modularize the code that handles the hierarchy of SCSI errors.
As a result, the code is now much easier to read.
- String handling and error printing has been significantly
revamped. We now use sbufs to do string formatting instead
of using printfs (for the kernel) and snprintf/strncat (for
userland) as before.
There is a new catchall error printing routine,
cam_error_print() and its string-based counterpart,
cam_error_string() that allow the kernel and userland
applications to pass in a CCB and have errors printed out
properly, whether or not they're SCSI errors. Among other
things, this helped eliminate a fair amount of duplicate code
in camcontrol.
We now print out more information than before, including
the CAM status and SCSI status and the error recovery action
taken to remedy the problem.
- sbufs are now available in userland, via libsbuf. This
change was necessary since most of the error printing code
is shared between libcam and the kernel.
- A new transfer settings interface is included in this checkin.
This code is #ifdef'ed out, and is primarily intended to aid
discussion with HBA driver authors on the final form the
interface should take. There is example code in the ahc(4)
driver that implements the HBA driver side of the new
interface. The new transfer settings code won't be enabled
until we're ready to switch all HBA drivers over to the new
interface.
src/Makefile.inc1,
lib/Makefile: Add libsbuf. It must be built before libcam,
since libcam uses sbuf routines.
libcam/Makefile: libcam now depends on libsbuf.
libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the
sbuf sources from sys/kern.
bsd.libnames.mk: Add LIBSBUF.
camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically
linked, we can't depend on the dynamic linker
to pull in libsbuf.
camcontrol.c: Use cam_error_print() instead of checking for
CAM_SCSI_STATUS_ERROR on every failed CCB.
sbuf.9: Change the prototypes for sbuf_cat() and
sbuf_cpy() so that the source string is now a
const char *. This is more in line wth the
standard system string functions, and helps
eliminate warnings when dealing with a const
source buffer.
Fix a typo.
cam.c: Add description strings for the various CAM
error status values, as well as routines to
look up those strings.
Add new cam_error_string() and
cam_error_print() routines for userland and
the kernel.
cam.h: Add a new CAM flag, CAM_RETRY_SELTO.
Add enumerated types for the various options
available with cam_error_print() and
cam_error_string().
cam_ccb.h: Add new transfer negotiation structures/types.
Change inq_len in the ccb_getdev structure to
be "reserved". This field has never been
filled in, and will be removed when we next
bump the CAM version.
cam_debug.h: Fix typo.
cam_periph.c: Modularize cam_periph_error(). The SCSI error
handling part of cam_periph_error() is now
in camperiphscsistatuserror() and
camperiphscsisenseerror().
In cam_periph_lock(), increase the reference
count on the periph while we wait for our lock
attempt to succeed so that the periph won't go
away while we're sleeping.
cam_xpt.c: Add new transfer negotiation code. (ifdefed
out)
Add a new function, xpt_path_string(). This
is a string/sbuf analog to xpt_print_path().
scsi_all.c: Revamp string handing and error printing code.
We now use sbufs for much of the string
formatting code. More of that code is shared
between userland the kernel.
scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly
useful in the first place.
Add a new error action, SS_REQSENSE. (Send a
request sense and then retry the command.)
This is useful when the controller hasn't
performed autosense for some reason.
Change the default actions around a bit.
scsi_cd.c,
scsi_da.c,
scsi_pt.c,
scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection
timeouts shouldn't be covered by a sense flag.
scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO.
Get rid of the last vestiges of a read/write
interface.
libkern/bsearch.c,
sys/libkern.h,
conf/files: Add bsearch.c, which is needed for some of the
new table lookup routines.
aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if
CAM_NEW_TRAN_CODE is defined.
sbuf.h,
subr_sbuf.c: Add the appropriate #ifdefs so sbufs can
compile and run in userland.
Change sbuf_printf() to use vsnprintf()
instead of kvprintf(), which is only available
in the kernel.
Change the source string for sbuf_cpy() and
sbuf_cat() to be a const char *.
Add __BEGIN_DECLS and __END_DECLS around
function prototypes since they're now exported
to userland.
kdump/mkioctls: Include stdio.h before cam.h since cam.h now
includes a function with a FILE * argument.
Submitted by: gibbs (mostly)
Reviewed by: jdp, marcel (libsbuf makefile changes)
Reviewed by: des (sbuf changes)
Reviewed by: ken
2001-03-27 05:45:52 +00:00
|
|
|
* Copyright (c) 1997, 1998, 2000 Justin T. Gibbs.
|
Add a prioritization field to the devstat_add_entry() call so that
peripheral drivers can determine where in the devstat(9) list they are
inserted.
This requires recompilation of libdevstat, systat, vmstat, rpc.rstatd, and
any ports that depend on the devstat code, since the size of the devstat
structure has changed. The devstat version number has been incremented as
well to reflect the change.
This sorts devices in the devstat list in "more interesting" to "less
interesting" order. So, for instance, da devices are now more important
than floppy drives, and so will appear before floppy drives in the default
output from systat, iostat, vmstat, etc.
The order of devices is, for now, kept in a central table in devicestat.h.
If individual drivers were able to make a meaningful decision on what
priority they should be at attach time, we could consider splitting the
priority information out into the various drivers. For now, though, they
have no way of knowing that, so it's easier to put them in an easy to find
table.
Also, move the checkversion() call in vmstat(8) to a more logical place.
Thanks to Bruce and David O'Brien for suggestions, for reviewing this, and
for putting up with the long time it has taken me to commit it. Bruce did
object somewhat to the central priority table (he would rather the
priorities be distributed in each driver), so his objection is duly noted
here.
Reviewed by: bde, obrien
1999-02-10 00:04:13 +00:00
|
|
|
* Copyright (c) 1997, 1998, 1999 Kenneth D. Merry.
|
1998-09-15 06:36:34 +00:00
|
|
|
* All rights reserved.
|
|
|
|
|
*
|
|
|
|
|
* Redistribution and use in source and binary forms, with or without
|
|
|
|
|
* modification, are permitted provided that the following conditions
|
|
|
|
|
* are met:
|
|
|
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
|
|
|
* notice, this list of conditions, and the following disclaimer,
|
|
|
|
|
* without modification, immediately at the beginning of the file.
|
|
|
|
|
* 2. The name of the author may not be used to endorse or promote products
|
|
|
|
|
* derived from this software without specific prior written permission.
|
|
|
|
|
*
|
|
|
|
|
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
|
|
|
|
|
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
|
|
|
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
|
|
|
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR
|
|
|
|
|
* ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
|
|
|
|
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
|
|
|
|
|
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
|
|
|
|
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
|
|
|
|
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
|
|
|
|
|
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
|
|
|
|
* SUCH DAMAGE.
|
|
|
|
|
*/
|
|
|
|
|
|
2003-06-10 18:14:05 +00:00
|
|
|
#include <sys/cdefs.h>
|
|
|
|
|
__FBSDID("$FreeBSD$");
|
|
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
#include "opt_kdtrace.h"
|
|
|
|
|
|
1998-09-15 06:36:34 +00:00
|
|
|
#include <sys/param.h>
|
|
|
|
|
#include <sys/systm.h>
|
|
|
|
|
#include <sys/kernel.h>
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
#include <sys/conf.h>
|
1998-09-15 06:36:34 +00:00
|
|
|
#include <sys/types.h>
|
2000-05-05 09:59:14 +00:00
|
|
|
#include <sys/bio.h>
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
#include <sys/bus.h>
|
1998-09-15 06:36:34 +00:00
|
|
|
#include <sys/devicestat.h>
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
#include <sys/errno.h>
|
|
|
|
|
#include <sys/fcntl.h>
|
|
|
|
|
#include <sys/malloc.h>
|
2001-09-26 20:13:16 +00:00
|
|
|
#include <sys/proc.h>
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
#include <sys/poll.h>
|
|
|
|
|
#include <sys/selinfo.h>
|
|
|
|
|
#include <sys/sdt.h>
|
Plumb device physical path reporting from CAM devices, through GEOM and
DEVFS, and make it accessible via the diskinfo utility.
Extend GEOM's generic attribute query mechanism into generic disk consumers.
sys/geom/geom_disk.c:
sys/geom/geom_disk.h:
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Allow disk providers to implement a new method which can override
the default BIO_GETATTR response, d_getattr(struct bio *). This
function returns -1 if not handled, otherwise it returns 0 or an
errno to be passed to g_io_deliver().
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Don't copy the serial number to dp->d_ident anymore, as the CAM XPT
is now responsible for returning this information via
d_getattr()->(a)dagetattr()->xpt_getatr().
sys/geom/geom_dev.c:
- Implement a new ioctl, DIOCGPHYSPATH, which returns the GEOM
attribute "GEOM::physpath", if possible. If the attribute request
returns a zero-length string, ENOENT is returned.
usr.sbin/diskinfo/diskinfo.c:
- If the DIOCGPHYSPATH ioctl is successful, report physical path
data when diskinfo is executed with the '-v' option.
Submitted by: will
Reviewed by: gibbs
Sponsored by: Spectra Logic Corporation
Add generic attribute change notification support to GEOM.
sys/sys/geom/geom.h:
Add a new attrchanged method field to both g_class
and g_geom.
sys/sys/geom/geom.h:
sys/geom/geom_event.c:
- Provide the g_attr_changed() function that providers
can use to advertise attribute changes.
- Perform delivery of attribute change notifications
from a thread context via the standard GEOM event
mechanism.
sys/geom/geom_subr.c:
Inherit the attrchanged method from class to geom (class instance).
sys/geom/geom_disk.c:
Provide disk_attr_changed() to provide g_attr_changed() access
to consumers of the disk API.
sys/cam/scsi/scsi_pass.c:
sys/cam/scsi/scsi_da.c:
sys/geom/geom_dev.c:
sys/geom/geom_disk.c:
Use attribute changed events to track updates to physical path
information.
sys/cam/scsi/scsi_da.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, and
the updated buffer type references our physical path
attribute, emit a GEOM attribute changed event via the
disk_attr_changed() API.
sys/cam/scsi/scsi_pass.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, update
the physical patch devfs alias for this pass instance.
Submitted by: gibbs
Sponsored by: Spectra Logic Corporation
2011-06-14 17:10:32 +00:00
|
|
|
#include <sys/taskqueue.h>
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
#include <vm/uma.h>
|
|
|
|
|
#include <vm/vm.h>
|
|
|
|
|
#include <vm/vm_extern.h>
|
|
|
|
|
|
|
|
|
|
#include <machine/bus.h>
|
1998-09-15 06:36:34 +00:00
|
|
|
|
|
|
|
|
#include <cam/cam.h>
|
|
|
|
|
#include <cam/cam_ccb.h>
|
|
|
|
|
#include <cam/cam_periph.h>
|
Rewrite of the CAM error recovery code.
Some of the major changes include:
- The SCSI error handling portion of cam_periph_error() has
been broken out into a number of subfunctions to better
modularize the code that handles the hierarchy of SCSI errors.
As a result, the code is now much easier to read.
- String handling and error printing has been significantly
revamped. We now use sbufs to do string formatting instead
of using printfs (for the kernel) and snprintf/strncat (for
userland) as before.
There is a new catchall error printing routine,
cam_error_print() and its string-based counterpart,
cam_error_string() that allow the kernel and userland
applications to pass in a CCB and have errors printed out
properly, whether or not they're SCSI errors. Among other
things, this helped eliminate a fair amount of duplicate code
in camcontrol.
We now print out more information than before, including
the CAM status and SCSI status and the error recovery action
taken to remedy the problem.
- sbufs are now available in userland, via libsbuf. This
change was necessary since most of the error printing code
is shared between libcam and the kernel.
- A new transfer settings interface is included in this checkin.
This code is #ifdef'ed out, and is primarily intended to aid
discussion with HBA driver authors on the final form the
interface should take. There is example code in the ahc(4)
driver that implements the HBA driver side of the new
interface. The new transfer settings code won't be enabled
until we're ready to switch all HBA drivers over to the new
interface.
src/Makefile.inc1,
lib/Makefile: Add libsbuf. It must be built before libcam,
since libcam uses sbuf routines.
libcam/Makefile: libcam now depends on libsbuf.
libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the
sbuf sources from sys/kern.
bsd.libnames.mk: Add LIBSBUF.
camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically
linked, we can't depend on the dynamic linker
to pull in libsbuf.
camcontrol.c: Use cam_error_print() instead of checking for
CAM_SCSI_STATUS_ERROR on every failed CCB.
sbuf.9: Change the prototypes for sbuf_cat() and
sbuf_cpy() so that the source string is now a
const char *. This is more in line wth the
standard system string functions, and helps
eliminate warnings when dealing with a const
source buffer.
Fix a typo.
cam.c: Add description strings for the various CAM
error status values, as well as routines to
look up those strings.
Add new cam_error_string() and
cam_error_print() routines for userland and
the kernel.
cam.h: Add a new CAM flag, CAM_RETRY_SELTO.
Add enumerated types for the various options
available with cam_error_print() and
cam_error_string().
cam_ccb.h: Add new transfer negotiation structures/types.
Change inq_len in the ccb_getdev structure to
be "reserved". This field has never been
filled in, and will be removed when we next
bump the CAM version.
cam_debug.h: Fix typo.
cam_periph.c: Modularize cam_periph_error(). The SCSI error
handling part of cam_periph_error() is now
in camperiphscsistatuserror() and
camperiphscsisenseerror().
In cam_periph_lock(), increase the reference
count on the periph while we wait for our lock
attempt to succeed so that the periph won't go
away while we're sleeping.
cam_xpt.c: Add new transfer negotiation code. (ifdefed
out)
Add a new function, xpt_path_string(). This
is a string/sbuf analog to xpt_print_path().
scsi_all.c: Revamp string handing and error printing code.
We now use sbufs for much of the string
formatting code. More of that code is shared
between userland the kernel.
scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly
useful in the first place.
Add a new error action, SS_REQSENSE. (Send a
request sense and then retry the command.)
This is useful when the controller hasn't
performed autosense for some reason.
Change the default actions around a bit.
scsi_cd.c,
scsi_da.c,
scsi_pt.c,
scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection
timeouts shouldn't be covered by a sense flag.
scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO.
Get rid of the last vestiges of a read/write
interface.
libkern/bsearch.c,
sys/libkern.h,
conf/files: Add bsearch.c, which is needed for some of the
new table lookup routines.
aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if
CAM_NEW_TRAN_CODE is defined.
sbuf.h,
subr_sbuf.c: Add the appropriate #ifdefs so sbufs can
compile and run in userland.
Change sbuf_printf() to use vsnprintf()
instead of kvprintf(), which is only available
in the kernel.
Change the source string for sbuf_cpy() and
sbuf_cat() to be a const char *.
Add __BEGIN_DECLS and __END_DECLS around
function prototypes since they're now exported
to userland.
kdump/mkioctls: Include stdio.h before cam.h since cam.h now
includes a function with a FILE * argument.
Submitted by: gibbs (mostly)
Reviewed by: jdp, marcel (libsbuf makefile changes)
Reviewed by: des (sbuf changes)
Reviewed by: ken
2001-03-27 05:45:52 +00:00
|
|
|
#include <cam/cam_queue.h>
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
#include <cam/cam_xpt.h>
|
1998-09-15 06:36:34 +00:00
|
|
|
#include <cam/cam_xpt_periph.h>
|
|
|
|
|
#include <cam/cam_debug.h>
|
2013-06-17 08:57:09 +00:00
|
|
|
#include <cam/cam_compat.h>
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
#include <cam/cam_xpt_periph.h>
|
1998-09-15 06:36:34 +00:00
|
|
|
|
|
|
|
|
#include <cam/scsi/scsi_all.h>
|
|
|
|
|
#include <cam/scsi/scsi_pass.h>
|
|
|
|
|
|
|
|
|
|
typedef enum {
|
|
|
|
|
PASS_FLAG_OPEN = 0x01,
|
|
|
|
|
PASS_FLAG_LOCKED = 0x02,
|
Fix several reference counting and object lifetime issues between
the pass(4) and enc(4) drivers and devfs.
The pass(4) driver uses the destroy_dev_sched() routine to
schedule its device node for destruction in a separate thread
context. It does this because the passcleanup() routine can get
called indirectly from the passclose() routine, and that would
cause a deadlock if the close routine tried to destroy its own
device node.
In any case, once a particular passthrough driver number, e.g.
pass3, is destroyed, CAM considers that unit number (3 in this
case) available for reuse.
The problem is that devfs may not be done cleaning up the previous
instance of pass3, and will panic if isn't done cleaning up the
previous instance.
The solution is to get a callback from devfs when the device node
is removed, and make sure we hold a reference to the peripheral
until that happens.
Testing exposed some other cases where we have reference counting
issues, and those were also fixed in the pass(4) driver.
cam_periph.c: In camperiphfree(), reorder some of the operations.
The peripheral destructor needs to be called before
the peripheral is removed from the peripheral is
removed from the list. This is because once we
remove the peripheral from the list, and drop the
topology lock, the peripheral number may be reused.
But if the destructor hasn't been called yet, there
may still be resources hanging around (like devfs
nodes) that haven't been fully cleaned up.
cam_xpt.c: Add an argument to xpt_remove_periph() to indicate
whether the topology lock is already held.
scsi_enc.c: Acquire an extra reference to the peripheral during
registration, and release it once we get a callback
from devfs indicating that the device node is gone.
Call destroy_dev_sched_cb() in enc_oninvalidate()
instead of calling destroy_dev() in the cleanup
routine.
scsi_pass.c: Add reference counting to handle peripheral and
devfs object lifetime issues.
Add a reference to the peripheral and the devfs
node in the peripheral registration.
Don't attempt to add a physical path alias if the
peripheral has been marked invalid.
Release the devfs reference once the initial
physical path alias taskqueue run has completed.
Schedule devfs node destruction in the
passoninvalidate(), and release our peripheral
reference in a new routine, passdevgonecb() once
the devfs node is gone. This allows the peripheral
to fully go away, and the peripheral destructor,
passcleanup(), will get called.
MFC after: 3 days
Sponsored by: Spectra Logic
2012-06-20 17:08:00 +00:00
|
|
|
PASS_FLAG_INVALID = 0x04,
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
PASS_FLAG_INITIAL_PHYSPATH = 0x08,
|
|
|
|
|
PASS_FLAG_ZONE_INPROG = 0x10,
|
|
|
|
|
PASS_FLAG_ZONE_VALID = 0x20,
|
|
|
|
|
PASS_FLAG_UNMAPPED_CAPABLE = 0x40,
|
|
|
|
|
PASS_FLAG_ABANDONED_REF_SET = 0x80
|
1998-09-15 06:36:34 +00:00
|
|
|
} pass_flags;
|
|
|
|
|
|
|
|
|
|
typedef enum {
|
|
|
|
|
PASS_STATE_NORMAL
|
|
|
|
|
} pass_state;
|
|
|
|
|
|
|
|
|
|
typedef enum {
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
PASS_CCB_BUFFER_IO,
|
|
|
|
|
PASS_CCB_QUEUED_IO
|
1998-09-15 06:36:34 +00:00
|
|
|
} pass_ccb_types;
|
|
|
|
|
|
|
|
|
|
#define ccb_type ppriv_field0
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
#define ccb_ioreq ppriv_ptr1
|
1998-09-15 06:36:34 +00:00
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
/*
|
|
|
|
|
* The maximum number of memory segments we preallocate.
|
|
|
|
|
*/
|
|
|
|
|
#define PASS_MAX_SEGS 16
|
|
|
|
|
|
|
|
|
|
typedef enum {
|
|
|
|
|
PASS_IO_NONE = 0x00,
|
|
|
|
|
PASS_IO_USER_SEG_MALLOC = 0x01,
|
|
|
|
|
PASS_IO_KERN_SEG_MALLOC = 0x02,
|
|
|
|
|
PASS_IO_ABANDONED = 0x04
|
|
|
|
|
} pass_io_flags;
|
|
|
|
|
|
|
|
|
|
struct pass_io_req {
|
|
|
|
|
union ccb ccb;
|
|
|
|
|
union ccb *alloced_ccb;
|
|
|
|
|
union ccb *user_ccb_ptr;
|
|
|
|
|
camq_entry user_periph_links;
|
|
|
|
|
ccb_ppriv_area user_periph_priv;
|
|
|
|
|
struct cam_periph_map_info mapinfo;
|
|
|
|
|
pass_io_flags flags;
|
|
|
|
|
ccb_flags data_flags;
|
|
|
|
|
int num_user_segs;
|
|
|
|
|
bus_dma_segment_t user_segs[PASS_MAX_SEGS];
|
|
|
|
|
int num_kern_segs;
|
|
|
|
|
bus_dma_segment_t kern_segs[PASS_MAX_SEGS];
|
|
|
|
|
bus_dma_segment_t *user_segptr;
|
|
|
|
|
bus_dma_segment_t *kern_segptr;
|
|
|
|
|
int num_bufs;
|
|
|
|
|
uint32_t dirs[CAM_PERIPH_MAXMAPS];
|
|
|
|
|
uint32_t lengths[CAM_PERIPH_MAXMAPS];
|
|
|
|
|
uint8_t *user_bufs[CAM_PERIPH_MAXMAPS];
|
|
|
|
|
uint8_t *kern_bufs[CAM_PERIPH_MAXMAPS];
|
|
|
|
|
struct bintime start_time;
|
|
|
|
|
TAILQ_ENTRY(pass_io_req) links;
|
1998-09-15 06:36:34 +00:00
|
|
|
};
|
|
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
struct pass_softc {
|
|
|
|
|
pass_state state;
|
|
|
|
|
pass_flags flags;
|
|
|
|
|
u_int8_t pd_type;
|
|
|
|
|
union ccb saved_ccb;
|
|
|
|
|
int open_count;
|
|
|
|
|
u_int maxio;
|
|
|
|
|
struct devstat *device_stats;
|
|
|
|
|
struct cdev *dev;
|
|
|
|
|
struct cdev *alias_dev;
|
|
|
|
|
struct task add_physpath_task;
|
|
|
|
|
struct task shutdown_kqueue_task;
|
|
|
|
|
struct selinfo read_select;
|
|
|
|
|
TAILQ_HEAD(, pass_io_req) incoming_queue;
|
|
|
|
|
TAILQ_HEAD(, pass_io_req) active_queue;
|
|
|
|
|
TAILQ_HEAD(, pass_io_req) abandoned_queue;
|
|
|
|
|
TAILQ_HEAD(, pass_io_req) done_queue;
|
|
|
|
|
struct cam_periph *periph;
|
|
|
|
|
char zone_name[12];
|
|
|
|
|
char io_zone_name[12];
|
|
|
|
|
uma_zone_t pass_zone;
|
|
|
|
|
uma_zone_t pass_io_zone;
|
|
|
|
|
size_t io_zone_size;
|
|
|
|
|
};
|
1998-09-15 06:36:34 +00:00
|
|
|
|
|
|
|
|
static d_open_t passopen;
|
|
|
|
|
static d_close_t passclose;
|
|
|
|
|
static d_ioctl_t passioctl;
|
2013-06-17 08:57:09 +00:00
|
|
|
static d_ioctl_t passdoioctl;
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
static d_poll_t passpoll;
|
|
|
|
|
static d_kqfilter_t passkqfilter;
|
|
|
|
|
static void passreadfiltdetach(struct knote *kn);
|
|
|
|
|
static int passreadfilt(struct knote *kn, long hint);
|
1998-09-15 06:36:34 +00:00
|
|
|
|
|
|
|
|
static periph_init_t passinit;
|
|
|
|
|
static periph_ctor_t passregister;
|
Fix a problem with the way we handled device invalidation when attaching
to a device failed.
In theory, the same steps that happen when we get an AC_LOST_DEVICE async
notification should have been taken when a driver fails to attach. In
practice, that wasn't the case.
This only affected the da, cd and ch drivers, but the fix affects all
peripheral drivers.
There were several possible problems:
- In the da driver, we didn't remove the peripheral's softc from the da
driver's linked list of softcs. Once the peripheral and softc got
removed, we'd get a kernel panic the next time the timeout routine
called dasendorderedtag().
- In the da, cd and possibly ch drivers, we didn't remove the
peripheral's devstat structure from the devstat queue. Once the
peripheral and softc were removed, this could cause a panic if anyone
tried to access device statistics. (one component of the linked list
wouldn't exist anymore)
- In the cd driver, we didn't take the peripheral off the changer run
queue if it was scheduled to run. In practice, it's highly unlikely,
and maybe impossible that the peripheral would have been on the
changer run queue at that stage of the probe process.
The fix is:
- Add a new peripheral callback function (the "oninvalidate" function)
that is called the first time cam_periph_invalidate() is called for a
peripheral.
- Create new foooninvalidate() routines for each peripheral driver. This
routine is always called at splsoftcam(), and contains all the stuff
that used to be in the AC_LOST_DEVICE case of the async callback
handler.
- Move the devstat cleanup call to the destructor/cleanup routines, since
some of the drivers do I/O in their close routines.
- Make sure that when we're flushing the buffer queue, we traverse it at
splbio().
- Add a check for the invalid flag in the pt driver's open routine.
Reviewed by: gibbs
1998-10-22 22:16:56 +00:00
|
|
|
static periph_oninv_t passoninvalidate;
|
1998-09-15 06:36:34 +00:00
|
|
|
static periph_dtor_t passcleanup;
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
static periph_start_t passstart;
|
|
|
|
|
static void pass_shutdown_kqueue(void *context, int pending);
|
|
|
|
|
static void pass_add_physpath(void *context, int pending);
|
1998-09-15 06:36:34 +00:00
|
|
|
static void passasync(void *callback_arg, u_int32_t code,
|
|
|
|
|
struct cam_path *path, void *arg);
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
static void passdone(struct cam_periph *periph,
|
|
|
|
|
union ccb *done_ccb);
|
|
|
|
|
static int passcreatezone(struct cam_periph *periph);
|
|
|
|
|
static void passiocleanup(struct pass_softc *softc,
|
|
|
|
|
struct pass_io_req *io_req);
|
|
|
|
|
static int passcopysglist(struct cam_periph *periph,
|
|
|
|
|
struct pass_io_req *io_req,
|
|
|
|
|
ccb_flags direction);
|
|
|
|
|
static int passmemsetup(struct cam_periph *periph,
|
|
|
|
|
struct pass_io_req *io_req);
|
|
|
|
|
static int passmemdone(struct cam_periph *periph,
|
|
|
|
|
struct pass_io_req *io_req);
|
1998-09-15 06:36:34 +00:00
|
|
|
static int passerror(union ccb *ccb, u_int32_t cam_flags,
|
|
|
|
|
u_int32_t sense_flags);
|
|
|
|
|
static int passsendccb(struct cam_periph *periph, union ccb *ccb,
|
|
|
|
|
union ccb *inccb);
|
|
|
|
|
|
|
|
|
|
static struct periph_driver passdriver =
|
|
|
|
|
{
|
|
|
|
|
passinit, "pass",
|
|
|
|
|
TAILQ_HEAD_INITIALIZER(passdriver.units), /* generation */ 0
|
|
|
|
|
};
|
|
|
|
|
|
2001-02-07 07:05:59 +00:00
|
|
|
PERIPHDRIVER_DECLARE(pass, passdriver);
|
1998-09-15 06:36:34 +00:00
|
|
|
|
1999-05-30 16:53:49 +00:00
|
|
|
static struct cdevsw pass_cdevsw = {
|
2004-02-21 21:10:55 +00:00
|
|
|
.d_version = D_VERSION,
|
Work around a race condition in devfs by changing the way closes
are handled in most CAM peripheral drivers that are not handled by
GEOM's disk class.
The usual character driver open and close semantics are that the
driver gets N open calls, but only one close, when the last caller
closes the device.
CAM peripheral drivers expect that behavior to be honored to the
letter, and the CAM peripheral driver code (specifically
cam_periph_release_locked_busses()) panics if it is done incorrectly.
Since devfs has to drop its locks while it calls a driver's close
routine, and it does not have a way to delay or prevent open calls
while it is calling the close routine, there is a race.
The sequence of events, simplified a bit, is:
- devfs acquires a lock
- devfs checks the reference count, and if it is 1, continues to close.
- devfs releases the lock
- 2nd process open call on the device happens here
- devfs calls the driver's close routine
- devfs acquires a lock
- devfs decrements the reference count
- devfs releases the lock
- 2nd process close call on the device happens here
At the second close, we get a panic in
cam_periph_release_locked_busses(), complaining that peripheral
has been released when the reference count is already 0. This is
because we have gotten two closes in a row, which should not
happen.
The fix is to add the D_TRACKCLOSE flag to the driver's cdevsw, so
that we get a close() call for each open(). That does happen
reliably, so we can make sure that our reference counts are
correct.
Note that the sa(4) and pt(4) drivers only allow one context
through the open routine. So these drivers aren't exposed to the
same race condition.
scsi_ch.c,
scsi_enc.c,
scsi_enc_internal.h,
scsi_pass.c,
scsi_sg.c:
For these drivers, change the open() routine to
increment the reference count for every open, and
just decrement the reference count in the close.
Call cam_periph_release_locked() in some scenarios
to avoid additional lock and unlock calls.
scsi_pt.c: Call cam_periph_release_locked() in some scenarios
to avoid additional lock and unlock calls.
MFC after: 3 days
2012-05-27 06:11:09 +00:00
|
|
|
.d_flags = D_TRACKCLOSE,
|
2003-03-03 12:15:54 +00:00
|
|
|
.d_open = passopen,
|
|
|
|
|
.d_close = passclose,
|
|
|
|
|
.d_ioctl = passioctl,
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
.d_poll = passpoll,
|
|
|
|
|
.d_kqfilter = passkqfilter,
|
2003-03-03 12:15:54 +00:00
|
|
|
.d_name = "pass",
|
1998-09-15 06:36:34 +00:00
|
|
|
};
|
|
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
static struct filterops passread_filtops = {
|
|
|
|
|
.f_isfd = 1,
|
|
|
|
|
.f_detach = passreadfiltdetach,
|
|
|
|
|
.f_event = passreadfilt
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
static MALLOC_DEFINE(M_SCSIPASS, "scsi_pass", "scsi passthrough buffers");
|
|
|
|
|
|
1998-09-15 06:36:34 +00:00
|
|
|
static void
|
|
|
|
|
passinit(void)
|
|
|
|
|
{
|
|
|
|
|
cam_status status;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Install a global async callback. This callback will
|
|
|
|
|
* receive async callbacks like "new device found".
|
|
|
|
|
*/
|
2007-05-16 16:54:23 +00:00
|
|
|
status = xpt_register_async(AC_FOUND_DEVICE, passasync, NULL, NULL);
|
1998-09-15 06:36:34 +00:00
|
|
|
|
|
|
|
|
if (status != CAM_REQ_CMP) {
|
|
|
|
|
printf("pass: Failed to attach master async callback "
|
|
|
|
|
"due to status 0x%x!\n", status);
|
|
|
|
|
}
|
Fix several reference counting and object lifetime issues between
the pass(4) and enc(4) drivers and devfs.
The pass(4) driver uses the destroy_dev_sched() routine to
schedule its device node for destruction in a separate thread
context. It does this because the passcleanup() routine can get
called indirectly from the passclose() routine, and that would
cause a deadlock if the close routine tried to destroy its own
device node.
In any case, once a particular passthrough driver number, e.g.
pass3, is destroyed, CAM considers that unit number (3 in this
case) available for reuse.
The problem is that devfs may not be done cleaning up the previous
instance of pass3, and will panic if isn't done cleaning up the
previous instance.
The solution is to get a callback from devfs when the device node
is removed, and make sure we hold a reference to the peripheral
until that happens.
Testing exposed some other cases where we have reference counting
issues, and those were also fixed in the pass(4) driver.
cam_periph.c: In camperiphfree(), reorder some of the operations.
The peripheral destructor needs to be called before
the peripheral is removed from the peripheral is
removed from the list. This is because once we
remove the peripheral from the list, and drop the
topology lock, the peripheral number may be reused.
But if the destructor hasn't been called yet, there
may still be resources hanging around (like devfs
nodes) that haven't been fully cleaned up.
cam_xpt.c: Add an argument to xpt_remove_periph() to indicate
whether the topology lock is already held.
scsi_enc.c: Acquire an extra reference to the peripheral during
registration, and release it once we get a callback
from devfs indicating that the device node is gone.
Call destroy_dev_sched_cb() in enc_oninvalidate()
instead of calling destroy_dev() in the cleanup
routine.
scsi_pass.c: Add reference counting to handle peripheral and
devfs object lifetime issues.
Add a reference to the peripheral and the devfs
node in the peripheral registration.
Don't attempt to add a physical path alias if the
peripheral has been marked invalid.
Release the devfs reference once the initial
physical path alias taskqueue run has completed.
Schedule devfs node destruction in the
passoninvalidate(), and release our peripheral
reference in a new routine, passdevgonecb() once
the devfs node is gone. This allows the peripheral
to fully go away, and the peripheral destructor,
passcleanup(), will get called.
MFC after: 3 days
Sponsored by: Spectra Logic
2012-06-20 17:08:00 +00:00
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
static void
|
|
|
|
|
passrejectios(struct cam_periph *periph)
|
|
|
|
|
{
|
|
|
|
|
struct pass_io_req *io_req, *io_req2;
|
|
|
|
|
struct pass_softc *softc;
|
|
|
|
|
|
|
|
|
|
softc = (struct pass_softc *)periph->softc;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* The user can no longer get status for I/O on the done queue, so
|
|
|
|
|
* clean up all outstanding I/O on the done queue.
|
|
|
|
|
*/
|
|
|
|
|
TAILQ_FOREACH_SAFE(io_req, &softc->done_queue, links, io_req2) {
|
|
|
|
|
TAILQ_REMOVE(&softc->done_queue, io_req, links);
|
|
|
|
|
passiocleanup(softc, io_req);
|
|
|
|
|
uma_zfree(softc->pass_zone, io_req);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* The underlying device is gone, so we can't issue these I/Os.
|
|
|
|
|
* The devfs node has been shut down, so we can't return status to
|
|
|
|
|
* the user. Free any I/O left on the incoming queue.
|
|
|
|
|
*/
|
|
|
|
|
TAILQ_FOREACH_SAFE(io_req, &softc->incoming_queue, links, io_req2) {
|
|
|
|
|
TAILQ_REMOVE(&softc->incoming_queue, io_req, links);
|
|
|
|
|
passiocleanup(softc, io_req);
|
|
|
|
|
uma_zfree(softc->pass_zone, io_req);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Normally we would put I/Os on the abandoned queue and acquire a
|
|
|
|
|
* reference when we saw the final close. But, the device went
|
|
|
|
|
* away and devfs may have moved everything off to deadfs by the
|
|
|
|
|
* time the I/O done callback is called; as a result, we won't see
|
|
|
|
|
* any more closes. So, if we have any active I/Os, we need to put
|
|
|
|
|
* them on the abandoned queue. When the abandoned queue is empty,
|
|
|
|
|
* we'll release the remaining reference (see below) to the peripheral.
|
|
|
|
|
*/
|
|
|
|
|
TAILQ_FOREACH_SAFE(io_req, &softc->active_queue, links, io_req2) {
|
|
|
|
|
TAILQ_REMOVE(&softc->active_queue, io_req, links);
|
|
|
|
|
io_req->flags |= PASS_IO_ABANDONED;
|
|
|
|
|
TAILQ_INSERT_TAIL(&softc->abandoned_queue, io_req, links);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* If we put any I/O on the abandoned queue, acquire a reference.
|
|
|
|
|
*/
|
|
|
|
|
if ((!TAILQ_EMPTY(&softc->abandoned_queue))
|
|
|
|
|
&& ((softc->flags & PASS_FLAG_ABANDONED_REF_SET) == 0)) {
|
|
|
|
|
cam_periph_doacquire(periph);
|
|
|
|
|
softc->flags |= PASS_FLAG_ABANDONED_REF_SET;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
Fix several reference counting and object lifetime issues between
the pass(4) and enc(4) drivers and devfs.
The pass(4) driver uses the destroy_dev_sched() routine to
schedule its device node for destruction in a separate thread
context. It does this because the passcleanup() routine can get
called indirectly from the passclose() routine, and that would
cause a deadlock if the close routine tried to destroy its own
device node.
In any case, once a particular passthrough driver number, e.g.
pass3, is destroyed, CAM considers that unit number (3 in this
case) available for reuse.
The problem is that devfs may not be done cleaning up the previous
instance of pass3, and will panic if isn't done cleaning up the
previous instance.
The solution is to get a callback from devfs when the device node
is removed, and make sure we hold a reference to the peripheral
until that happens.
Testing exposed some other cases where we have reference counting
issues, and those were also fixed in the pass(4) driver.
cam_periph.c: In camperiphfree(), reorder some of the operations.
The peripheral destructor needs to be called before
the peripheral is removed from the peripheral is
removed from the list. This is because once we
remove the peripheral from the list, and drop the
topology lock, the peripheral number may be reused.
But if the destructor hasn't been called yet, there
may still be resources hanging around (like devfs
nodes) that haven't been fully cleaned up.
cam_xpt.c: Add an argument to xpt_remove_periph() to indicate
whether the topology lock is already held.
scsi_enc.c: Acquire an extra reference to the peripheral during
registration, and release it once we get a callback
from devfs indicating that the device node is gone.
Call destroy_dev_sched_cb() in enc_oninvalidate()
instead of calling destroy_dev() in the cleanup
routine.
scsi_pass.c: Add reference counting to handle peripheral and
devfs object lifetime issues.
Add a reference to the peripheral and the devfs
node in the peripheral registration.
Don't attempt to add a physical path alias if the
peripheral has been marked invalid.
Release the devfs reference once the initial
physical path alias taskqueue run has completed.
Schedule devfs node destruction in the
passoninvalidate(), and release our peripheral
reference in a new routine, passdevgonecb() once
the devfs node is gone. This allows the peripheral
to fully go away, and the peripheral destructor,
passcleanup(), will get called.
MFC after: 3 days
Sponsored by: Spectra Logic
2012-06-20 17:08:00 +00:00
|
|
|
static void
|
|
|
|
|
passdevgonecb(void *arg)
|
|
|
|
|
{
|
|
|
|
|
struct cam_periph *periph;
|
Merge CAM locking changes from the projects/camlock branch to radically
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
2013-10-21 12:00:26 +00:00
|
|
|
struct mtx *mtx;
|
Fix a device departure bug for the the pass(4), enc(4), sg(4) and ch(4)
drivers.
The bug occurrs when a userland process has the driver instance
open and the underlying device goes away. We get the devfs
callback that the device node has been destroyed, but not all of
the closes necessary to fully decrement the reference count on the
CAM peripheral.
The reason is that once devfs calls back and says the device has
been destroyed, it is moved off to deadfs, and devfs guarantees
that there will be no more open or close calls. So the solution
is to keep track of how many outstanding open calls there are on
the device, and just release that many references when we get the
callback from devfs.
scsi_pass.c,
scsi_enc.c,
scsi_enc_internal.h: Add an open count to the softc in these
drivers. Increment it on open and
decrement it on close.
When we get a devfs callback to say that
the device node has gone away, decrement
the peripheral reference count by the
number of still outstanding opens.
Make sure we don't access the peripheral
with cam_periph_unlock() after what might
be the final call to
cam_periph_release_locked(). The
peripheral might have been freed, and we
will be dereferencing freed memory.
scsi_ch.c,
scsi_sg.c: For the ch(4) and sg(4) drivers, add the
same changes described above, and in
addition, fix another bug that was
previously fixed in the pass(4) and enc(4)
drivers.
These drivers were calling destroy_dev()
from their cleanup routine, but that could
cause a deadlock because the cleanup
routine could be indirectly called from
the driver's close routine. This would
cause a deadlock, because the device node
is being held open by the active close
call, and can't be destroyed.
Sponsored by: Spectra Logic Corporation
MFC after: 1 week
2012-12-08 04:03:04 +00:00
|
|
|
struct pass_softc *softc;
|
|
|
|
|
int i;
|
Fix several reference counting and object lifetime issues between
the pass(4) and enc(4) drivers and devfs.
The pass(4) driver uses the destroy_dev_sched() routine to
schedule its device node for destruction in a separate thread
context. It does this because the passcleanup() routine can get
called indirectly from the passclose() routine, and that would
cause a deadlock if the close routine tried to destroy its own
device node.
In any case, once a particular passthrough driver number, e.g.
pass3, is destroyed, CAM considers that unit number (3 in this
case) available for reuse.
The problem is that devfs may not be done cleaning up the previous
instance of pass3, and will panic if isn't done cleaning up the
previous instance.
The solution is to get a callback from devfs when the device node
is removed, and make sure we hold a reference to the peripheral
until that happens.
Testing exposed some other cases where we have reference counting
issues, and those were also fixed in the pass(4) driver.
cam_periph.c: In camperiphfree(), reorder some of the operations.
The peripheral destructor needs to be called before
the peripheral is removed from the peripheral is
removed from the list. This is because once we
remove the peripheral from the list, and drop the
topology lock, the peripheral number may be reused.
But if the destructor hasn't been called yet, there
may still be resources hanging around (like devfs
nodes) that haven't been fully cleaned up.
cam_xpt.c: Add an argument to xpt_remove_periph() to indicate
whether the topology lock is already held.
scsi_enc.c: Acquire an extra reference to the peripheral during
registration, and release it once we get a callback
from devfs indicating that the device node is gone.
Call destroy_dev_sched_cb() in enc_oninvalidate()
instead of calling destroy_dev() in the cleanup
routine.
scsi_pass.c: Add reference counting to handle peripheral and
devfs object lifetime issues.
Add a reference to the peripheral and the devfs
node in the peripheral registration.
Don't attempt to add a physical path alias if the
peripheral has been marked invalid.
Release the devfs reference once the initial
physical path alias taskqueue run has completed.
Schedule devfs node destruction in the
passoninvalidate(), and release our peripheral
reference in a new routine, passdevgonecb() once
the devfs node is gone. This allows the peripheral
to fully go away, and the peripheral destructor,
passcleanup(), will get called.
MFC after: 3 days
Sponsored by: Spectra Logic
2012-06-20 17:08:00 +00:00
|
|
|
|
|
|
|
|
periph = (struct cam_periph *)arg;
|
Merge CAM locking changes from the projects/camlock branch to radically
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
2013-10-21 12:00:26 +00:00
|
|
|
mtx = cam_periph_mtx(periph);
|
|
|
|
|
mtx_lock(mtx);
|
Fix a device departure bug for the the pass(4), enc(4), sg(4) and ch(4)
drivers.
The bug occurrs when a userland process has the driver instance
open and the underlying device goes away. We get the devfs
callback that the device node has been destroyed, but not all of
the closes necessary to fully decrement the reference count on the
CAM peripheral.
The reason is that once devfs calls back and says the device has
been destroyed, it is moved off to deadfs, and devfs guarantees
that there will be no more open or close calls. So the solution
is to keep track of how many outstanding open calls there are on
the device, and just release that many references when we get the
callback from devfs.
scsi_pass.c,
scsi_enc.c,
scsi_enc_internal.h: Add an open count to the softc in these
drivers. Increment it on open and
decrement it on close.
When we get a devfs callback to say that
the device node has gone away, decrement
the peripheral reference count by the
number of still outstanding opens.
Make sure we don't access the peripheral
with cam_periph_unlock() after what might
be the final call to
cam_periph_release_locked(). The
peripheral might have been freed, and we
will be dereferencing freed memory.
scsi_ch.c,
scsi_sg.c: For the ch(4) and sg(4) drivers, add the
same changes described above, and in
addition, fix another bug that was
previously fixed in the pass(4) and enc(4)
drivers.
These drivers were calling destroy_dev()
from their cleanup routine, but that could
cause a deadlock because the cleanup
routine could be indirectly called from
the driver's close routine. This would
cause a deadlock, because the device node
is being held open by the active close
call, and can't be destroyed.
Sponsored by: Spectra Logic Corporation
MFC after: 1 week
2012-12-08 04:03:04 +00:00
|
|
|
|
Merge CAM locking changes from the projects/camlock branch to radically
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
2013-10-21 12:00:26 +00:00
|
|
|
softc = (struct pass_softc *)periph->softc;
|
Fix a device departure bug for the the pass(4), enc(4), sg(4) and ch(4)
drivers.
The bug occurrs when a userland process has the driver instance
open and the underlying device goes away. We get the devfs
callback that the device node has been destroyed, but not all of
the closes necessary to fully decrement the reference count on the
CAM peripheral.
The reason is that once devfs calls back and says the device has
been destroyed, it is moved off to deadfs, and devfs guarantees
that there will be no more open or close calls. So the solution
is to keep track of how many outstanding open calls there are on
the device, and just release that many references when we get the
callback from devfs.
scsi_pass.c,
scsi_enc.c,
scsi_enc_internal.h: Add an open count to the softc in these
drivers. Increment it on open and
decrement it on close.
When we get a devfs callback to say that
the device node has gone away, decrement
the peripheral reference count by the
number of still outstanding opens.
Make sure we don't access the peripheral
with cam_periph_unlock() after what might
be the final call to
cam_periph_release_locked(). The
peripheral might have been freed, and we
will be dereferencing freed memory.
scsi_ch.c,
scsi_sg.c: For the ch(4) and sg(4) drivers, add the
same changes described above, and in
addition, fix another bug that was
previously fixed in the pass(4) and enc(4)
drivers.
These drivers were calling destroy_dev()
from their cleanup routine, but that could
cause a deadlock because the cleanup
routine could be indirectly called from
the driver's close routine. This would
cause a deadlock, because the device node
is being held open by the active close
call, and can't be destroyed.
Sponsored by: Spectra Logic Corporation
MFC after: 1 week
2012-12-08 04:03:04 +00:00
|
|
|
KASSERT(softc->open_count >= 0, ("Negative open count %d",
|
|
|
|
|
softc->open_count));
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* When we get this callback, we will get no more close calls from
|
|
|
|
|
* devfs. So if we have any dangling opens, we need to release the
|
|
|
|
|
* reference held for that particular context.
|
|
|
|
|
*/
|
|
|
|
|
for (i = 0; i < softc->open_count; i++)
|
|
|
|
|
cam_periph_release_locked(periph);
|
|
|
|
|
|
|
|
|
|
softc->open_count = 0;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Release the reference held for the device node, it is gone now.
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
* Accordingly, inform all queued I/Os of their fate.
|
Fix a device departure bug for the the pass(4), enc(4), sg(4) and ch(4)
drivers.
The bug occurrs when a userland process has the driver instance
open and the underlying device goes away. We get the devfs
callback that the device node has been destroyed, but not all of
the closes necessary to fully decrement the reference count on the
CAM peripheral.
The reason is that once devfs calls back and says the device has
been destroyed, it is moved off to deadfs, and devfs guarantees
that there will be no more open or close calls. So the solution
is to keep track of how many outstanding open calls there are on
the device, and just release that many references when we get the
callback from devfs.
scsi_pass.c,
scsi_enc.c,
scsi_enc_internal.h: Add an open count to the softc in these
drivers. Increment it on open and
decrement it on close.
When we get a devfs callback to say that
the device node has gone away, decrement
the peripheral reference count by the
number of still outstanding opens.
Make sure we don't access the peripheral
with cam_periph_unlock() after what might
be the final call to
cam_periph_release_locked(). The
peripheral might have been freed, and we
will be dereferencing freed memory.
scsi_ch.c,
scsi_sg.c: For the ch(4) and sg(4) drivers, add the
same changes described above, and in
addition, fix another bug that was
previously fixed in the pass(4) and enc(4)
drivers.
These drivers were calling destroy_dev()
from their cleanup routine, but that could
cause a deadlock because the cleanup
routine could be indirectly called from
the driver's close routine. This would
cause a deadlock, because the device node
is being held open by the active close
call, and can't be destroyed.
Sponsored by: Spectra Logic Corporation
MFC after: 1 week
2012-12-08 04:03:04 +00:00
|
|
|
*/
|
|
|
|
|
cam_periph_release_locked(periph);
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
passrejectios(periph);
|
Fix several reference counting and object lifetime issues between
the pass(4) and enc(4) drivers and devfs.
The pass(4) driver uses the destroy_dev_sched() routine to
schedule its device node for destruction in a separate thread
context. It does this because the passcleanup() routine can get
called indirectly from the passclose() routine, and that would
cause a deadlock if the close routine tried to destroy its own
device node.
In any case, once a particular passthrough driver number, e.g.
pass3, is destroyed, CAM considers that unit number (3 in this
case) available for reuse.
The problem is that devfs may not be done cleaning up the previous
instance of pass3, and will panic if isn't done cleaning up the
previous instance.
The solution is to get a callback from devfs when the device node
is removed, and make sure we hold a reference to the peripheral
until that happens.
Testing exposed some other cases where we have reference counting
issues, and those were also fixed in the pass(4) driver.
cam_periph.c: In camperiphfree(), reorder some of the operations.
The peripheral destructor needs to be called before
the peripheral is removed from the peripheral is
removed from the list. This is because once we
remove the peripheral from the list, and drop the
topology lock, the peripheral number may be reused.
But if the destructor hasn't been called yet, there
may still be resources hanging around (like devfs
nodes) that haven't been fully cleaned up.
cam_xpt.c: Add an argument to xpt_remove_periph() to indicate
whether the topology lock is already held.
scsi_enc.c: Acquire an extra reference to the peripheral during
registration, and release it once we get a callback
from devfs indicating that the device node is gone.
Call destroy_dev_sched_cb() in enc_oninvalidate()
instead of calling destroy_dev() in the cleanup
routine.
scsi_pass.c: Add reference counting to handle peripheral and
devfs object lifetime issues.
Add a reference to the peripheral and the devfs
node in the peripheral registration.
Don't attempt to add a physical path alias if the
peripheral has been marked invalid.
Release the devfs reference once the initial
physical path alias taskqueue run has completed.
Schedule devfs node destruction in the
passoninvalidate(), and release our peripheral
reference in a new routine, passdevgonecb() once
the devfs node is gone. This allows the peripheral
to fully go away, and the peripheral destructor,
passcleanup(), will get called.
MFC after: 3 days
Sponsored by: Spectra Logic
2012-06-20 17:08:00 +00:00
|
|
|
|
Fix a device departure bug for the the pass(4), enc(4), sg(4) and ch(4)
drivers.
The bug occurrs when a userland process has the driver instance
open and the underlying device goes away. We get the devfs
callback that the device node has been destroyed, but not all of
the closes necessary to fully decrement the reference count on the
CAM peripheral.
The reason is that once devfs calls back and says the device has
been destroyed, it is moved off to deadfs, and devfs guarantees
that there will be no more open or close calls. So the solution
is to keep track of how many outstanding open calls there are on
the device, and just release that many references when we get the
callback from devfs.
scsi_pass.c,
scsi_enc.c,
scsi_enc_internal.h: Add an open count to the softc in these
drivers. Increment it on open and
decrement it on close.
When we get a devfs callback to say that
the device node has gone away, decrement
the peripheral reference count by the
number of still outstanding opens.
Make sure we don't access the peripheral
with cam_periph_unlock() after what might
be the final call to
cam_periph_release_locked(). The
peripheral might have been freed, and we
will be dereferencing freed memory.
scsi_ch.c,
scsi_sg.c: For the ch(4) and sg(4) drivers, add the
same changes described above, and in
addition, fix another bug that was
previously fixed in the pass(4) and enc(4)
drivers.
These drivers were calling destroy_dev()
from their cleanup routine, but that could
cause a deadlock because the cleanup
routine could be indirectly called from
the driver's close routine. This would
cause a deadlock, because the device node
is being held open by the active close
call, and can't be destroyed.
Sponsored by: Spectra Logic Corporation
MFC after: 1 week
2012-12-08 04:03:04 +00:00
|
|
|
/*
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
* We reference the SIM lock directly here, instead of using
|
Fix a device departure bug for the the pass(4), enc(4), sg(4) and ch(4)
drivers.
The bug occurrs when a userland process has the driver instance
open and the underlying device goes away. We get the devfs
callback that the device node has been destroyed, but not all of
the closes necessary to fully decrement the reference count on the
CAM peripheral.
The reason is that once devfs calls back and says the device has
been destroyed, it is moved off to deadfs, and devfs guarantees
that there will be no more open or close calls. So the solution
is to keep track of how many outstanding open calls there are on
the device, and just release that many references when we get the
callback from devfs.
scsi_pass.c,
scsi_enc.c,
scsi_enc_internal.h: Add an open count to the softc in these
drivers. Increment it on open and
decrement it on close.
When we get a devfs callback to say that
the device node has gone away, decrement
the peripheral reference count by the
number of still outstanding opens.
Make sure we don't access the peripheral
with cam_periph_unlock() after what might
be the final call to
cam_periph_release_locked(). The
peripheral might have been freed, and we
will be dereferencing freed memory.
scsi_ch.c,
scsi_sg.c: For the ch(4) and sg(4) drivers, add the
same changes described above, and in
addition, fix another bug that was
previously fixed in the pass(4) and enc(4)
drivers.
These drivers were calling destroy_dev()
from their cleanup routine, but that could
cause a deadlock because the cleanup
routine could be indirectly called from
the driver's close routine. This would
cause a deadlock, because the device node
is being held open by the active close
call, and can't be destroyed.
Sponsored by: Spectra Logic Corporation
MFC after: 1 week
2012-12-08 04:03:04 +00:00
|
|
|
* cam_periph_unlock(). The reason is that the final call to
|
|
|
|
|
* cam_periph_release_locked() above could result in the periph
|
|
|
|
|
* getting freed. If that is the case, dereferencing the periph
|
|
|
|
|
* with a cam_periph_unlock() call would cause a page fault.
|
|
|
|
|
*/
|
Merge CAM locking changes from the projects/camlock branch to radically
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
2013-10-21 12:00:26 +00:00
|
|
|
mtx_unlock(mtx);
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* We have to remove our kqueue context from a thread because it
|
|
|
|
|
* may sleep. It would be nice if we could get a callback from
|
|
|
|
|
* kqueue when it is done cleaning up resources.
|
|
|
|
|
*/
|
|
|
|
|
taskqueue_enqueue(taskqueue_thread, &softc->shutdown_kqueue_task);
|
1998-09-15 06:36:34 +00:00
|
|
|
}
|
|
|
|
|
|
Fix a problem with the way we handled device invalidation when attaching
to a device failed.
In theory, the same steps that happen when we get an AC_LOST_DEVICE async
notification should have been taken when a driver fails to attach. In
practice, that wasn't the case.
This only affected the da, cd and ch drivers, but the fix affects all
peripheral drivers.
There were several possible problems:
- In the da driver, we didn't remove the peripheral's softc from the da
driver's linked list of softcs. Once the peripheral and softc got
removed, we'd get a kernel panic the next time the timeout routine
called dasendorderedtag().
- In the da, cd and possibly ch drivers, we didn't remove the
peripheral's devstat structure from the devstat queue. Once the
peripheral and softc were removed, this could cause a panic if anyone
tried to access device statistics. (one component of the linked list
wouldn't exist anymore)
- In the cd driver, we didn't take the peripheral off the changer run
queue if it was scheduled to run. In practice, it's highly unlikely,
and maybe impossible that the peripheral would have been on the
changer run queue at that stage of the probe process.
The fix is:
- Add a new peripheral callback function (the "oninvalidate" function)
that is called the first time cam_periph_invalidate() is called for a
peripheral.
- Create new foooninvalidate() routines for each peripheral driver. This
routine is always called at splsoftcam(), and contains all the stuff
that used to be in the AC_LOST_DEVICE case of the async callback
handler.
- Move the devstat cleanup call to the destructor/cleanup routines, since
some of the drivers do I/O in their close routines.
- Make sure that when we're flushing the buffer queue, we traverse it at
splbio().
- Add a check for the invalid flag in the pt driver's open routine.
Reviewed by: gibbs
1998-10-22 22:16:56 +00:00
|
|
|
static void
|
|
|
|
|
passoninvalidate(struct cam_periph *periph)
|
|
|
|
|
{
|
|
|
|
|
struct pass_softc *softc;
|
|
|
|
|
|
|
|
|
|
softc = (struct pass_softc *)periph->softc;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* De-register any async callbacks.
|
|
|
|
|
*/
|
2007-05-16 16:54:23 +00:00
|
|
|
xpt_register_async(0, passasync, periph, periph->path);
|
Fix a problem with the way we handled device invalidation when attaching
to a device failed.
In theory, the same steps that happen when we get an AC_LOST_DEVICE async
notification should have been taken when a driver fails to attach. In
practice, that wasn't the case.
This only affected the da, cd and ch drivers, but the fix affects all
peripheral drivers.
There were several possible problems:
- In the da driver, we didn't remove the peripheral's softc from the da
driver's linked list of softcs. Once the peripheral and softc got
removed, we'd get a kernel panic the next time the timeout routine
called dasendorderedtag().
- In the da, cd and possibly ch drivers, we didn't remove the
peripheral's devstat structure from the devstat queue. Once the
peripheral and softc were removed, this could cause a panic if anyone
tried to access device statistics. (one component of the linked list
wouldn't exist anymore)
- In the cd driver, we didn't take the peripheral off the changer run
queue if it was scheduled to run. In practice, it's highly unlikely,
and maybe impossible that the peripheral would have been on the
changer run queue at that stage of the probe process.
The fix is:
- Add a new peripheral callback function (the "oninvalidate" function)
that is called the first time cam_periph_invalidate() is called for a
peripheral.
- Create new foooninvalidate() routines for each peripheral driver. This
routine is always called at splsoftcam(), and contains all the stuff
that used to be in the AC_LOST_DEVICE case of the async callback
handler.
- Move the devstat cleanup call to the destructor/cleanup routines, since
some of the drivers do I/O in their close routines.
- Make sure that when we're flushing the buffer queue, we traverse it at
splbio().
- Add a check for the invalid flag in the pt driver's open routine.
Reviewed by: gibbs
1998-10-22 22:16:56 +00:00
|
|
|
|
|
|
|
|
softc->flags |= PASS_FLAG_INVALID;
|
|
|
|
|
|
Fix several reference counting and object lifetime issues between
the pass(4) and enc(4) drivers and devfs.
The pass(4) driver uses the destroy_dev_sched() routine to
schedule its device node for destruction in a separate thread
context. It does this because the passcleanup() routine can get
called indirectly from the passclose() routine, and that would
cause a deadlock if the close routine tried to destroy its own
device node.
In any case, once a particular passthrough driver number, e.g.
pass3, is destroyed, CAM considers that unit number (3 in this
case) available for reuse.
The problem is that devfs may not be done cleaning up the previous
instance of pass3, and will panic if isn't done cleaning up the
previous instance.
The solution is to get a callback from devfs when the device node
is removed, and make sure we hold a reference to the peripheral
until that happens.
Testing exposed some other cases where we have reference counting
issues, and those were also fixed in the pass(4) driver.
cam_periph.c: In camperiphfree(), reorder some of the operations.
The peripheral destructor needs to be called before
the peripheral is removed from the peripheral is
removed from the list. This is because once we
remove the peripheral from the list, and drop the
topology lock, the peripheral number may be reused.
But if the destructor hasn't been called yet, there
may still be resources hanging around (like devfs
nodes) that haven't been fully cleaned up.
cam_xpt.c: Add an argument to xpt_remove_periph() to indicate
whether the topology lock is already held.
scsi_enc.c: Acquire an extra reference to the peripheral during
registration, and release it once we get a callback
from devfs indicating that the device node is gone.
Call destroy_dev_sched_cb() in enc_oninvalidate()
instead of calling destroy_dev() in the cleanup
routine.
scsi_pass.c: Add reference counting to handle peripheral and
devfs object lifetime issues.
Add a reference to the peripheral and the devfs
node in the peripheral registration.
Don't attempt to add a physical path alias if the
peripheral has been marked invalid.
Release the devfs reference once the initial
physical path alias taskqueue run has completed.
Schedule devfs node destruction in the
passoninvalidate(), and release our peripheral
reference in a new routine, passdevgonecb() once
the devfs node is gone. This allows the peripheral
to fully go away, and the peripheral destructor,
passcleanup(), will get called.
MFC after: 3 days
Sponsored by: Spectra Logic
2012-06-20 17:08:00 +00:00
|
|
|
/*
|
|
|
|
|
* Tell devfs this device has gone away, and ask for a callback
|
|
|
|
|
* when it has cleaned up its state.
|
|
|
|
|
*/
|
|
|
|
|
destroy_dev_sched_cb(softc->dev, passdevgonecb, periph);
|
Fix a problem with the way we handled device invalidation when attaching
to a device failed.
In theory, the same steps that happen when we get an AC_LOST_DEVICE async
notification should have been taken when a driver fails to attach. In
practice, that wasn't the case.
This only affected the da, cd and ch drivers, but the fix affects all
peripheral drivers.
There were several possible problems:
- In the da driver, we didn't remove the peripheral's softc from the da
driver's linked list of softcs. Once the peripheral and softc got
removed, we'd get a kernel panic the next time the timeout routine
called dasendorderedtag().
- In the da, cd and possibly ch drivers, we didn't remove the
peripheral's devstat structure from the devstat queue. Once the
peripheral and softc were removed, this could cause a panic if anyone
tried to access device statistics. (one component of the linked list
wouldn't exist anymore)
- In the cd driver, we didn't take the peripheral off the changer run
queue if it was scheduled to run. In practice, it's highly unlikely,
and maybe impossible that the peripheral would have been on the
changer run queue at that stage of the probe process.
The fix is:
- Add a new peripheral callback function (the "oninvalidate" function)
that is called the first time cam_periph_invalidate() is called for a
peripheral.
- Create new foooninvalidate() routines for each peripheral driver. This
routine is always called at splsoftcam(), and contains all the stuff
that used to be in the AC_LOST_DEVICE case of the async callback
handler.
- Move the devstat cleanup call to the destructor/cleanup routines, since
some of the drivers do I/O in their close routines.
- Make sure that when we're flushing the buffer queue, we traverse it at
splbio().
- Add a check for the invalid flag in the pt driver's open routine.
Reviewed by: gibbs
1998-10-22 22:16:56 +00:00
|
|
|
}
|
|
|
|
|
|
1998-09-15 06:36:34 +00:00
|
|
|
static void
|
|
|
|
|
passcleanup(struct cam_periph *periph)
|
|
|
|
|
{
|
Fix a problem with the way we handled device invalidation when attaching
to a device failed.
In theory, the same steps that happen when we get an AC_LOST_DEVICE async
notification should have been taken when a driver fails to attach. In
practice, that wasn't the case.
This only affected the da, cd and ch drivers, but the fix affects all
peripheral drivers.
There were several possible problems:
- In the da driver, we didn't remove the peripheral's softc from the da
driver's linked list of softcs. Once the peripheral and softc got
removed, we'd get a kernel panic the next time the timeout routine
called dasendorderedtag().
- In the da, cd and possibly ch drivers, we didn't remove the
peripheral's devstat structure from the devstat queue. Once the
peripheral and softc were removed, this could cause a panic if anyone
tried to access device statistics. (one component of the linked list
wouldn't exist anymore)
- In the cd driver, we didn't take the peripheral off the changer run
queue if it was scheduled to run. In practice, it's highly unlikely,
and maybe impossible that the peripheral would have been on the
changer run queue at that stage of the probe process.
The fix is:
- Add a new peripheral callback function (the "oninvalidate" function)
that is called the first time cam_periph_invalidate() is called for a
peripheral.
- Create new foooninvalidate() routines for each peripheral driver. This
routine is always called at splsoftcam(), and contains all the stuff
that used to be in the AC_LOST_DEVICE case of the async callback
handler.
- Move the devstat cleanup call to the destructor/cleanup routines, since
some of the drivers do I/O in their close routines.
- Make sure that when we're flushing the buffer queue, we traverse it at
splbio().
- Add a check for the invalid flag in the pt driver's open routine.
Reviewed by: gibbs
1998-10-22 22:16:56 +00:00
|
|
|
struct pass_softc *softc;
|
|
|
|
|
|
|
|
|
|
softc = (struct pass_softc *)periph->softc;
|
|
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
cam_periph_assert(periph, MA_OWNED);
|
|
|
|
|
KASSERT(TAILQ_EMPTY(&softc->active_queue),
|
|
|
|
|
("%s called when there are commands on the active queue!\n",
|
|
|
|
|
__func__));
|
|
|
|
|
KASSERT(TAILQ_EMPTY(&softc->abandoned_queue),
|
|
|
|
|
("%s called when there are commands on the abandoned queue!\n",
|
|
|
|
|
__func__));
|
|
|
|
|
KASSERT(TAILQ_EMPTY(&softc->incoming_queue),
|
|
|
|
|
("%s called when there are commands on the incoming queue!\n",
|
|
|
|
|
__func__));
|
|
|
|
|
KASSERT(TAILQ_EMPTY(&softc->done_queue),
|
|
|
|
|
("%s called when there are commands on the done queue!\n",
|
|
|
|
|
__func__));
|
|
|
|
|
|
2003-03-08 21:44:46 +00:00
|
|
|
devstat_remove_entry(softc->device_stats);
|
Plumb device physical path reporting from CAM devices, through GEOM and
DEVFS, and make it accessible via the diskinfo utility.
Extend GEOM's generic attribute query mechanism into generic disk consumers.
sys/geom/geom_disk.c:
sys/geom/geom_disk.h:
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Allow disk providers to implement a new method which can override
the default BIO_GETATTR response, d_getattr(struct bio *). This
function returns -1 if not handled, otherwise it returns 0 or an
errno to be passed to g_io_deliver().
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Don't copy the serial number to dp->d_ident anymore, as the CAM XPT
is now responsible for returning this information via
d_getattr()->(a)dagetattr()->xpt_getatr().
sys/geom/geom_dev.c:
- Implement a new ioctl, DIOCGPHYSPATH, which returns the GEOM
attribute "GEOM::physpath", if possible. If the attribute request
returns a zero-length string, ENOENT is returned.
usr.sbin/diskinfo/diskinfo.c:
- If the DIOCGPHYSPATH ioctl is successful, report physical path
data when diskinfo is executed with the '-v' option.
Submitted by: will
Reviewed by: gibbs
Sponsored by: Spectra Logic Corporation
Add generic attribute change notification support to GEOM.
sys/sys/geom/geom.h:
Add a new attrchanged method field to both g_class
and g_geom.
sys/sys/geom/geom.h:
sys/geom/geom_event.c:
- Provide the g_attr_changed() function that providers
can use to advertise attribute changes.
- Perform delivery of attribute change notifications
from a thread context via the standard GEOM event
mechanism.
sys/geom/geom_subr.c:
Inherit the attrchanged method from class to geom (class instance).
sys/geom/geom_disk.c:
Provide disk_attr_changed() to provide g_attr_changed() access
to consumers of the disk API.
sys/cam/scsi/scsi_pass.c:
sys/cam/scsi/scsi_da.c:
sys/geom/geom_dev.c:
sys/geom/geom_disk.c:
Use attribute changed events to track updates to physical path
information.
sys/cam/scsi/scsi_da.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, and
the updated buffer type references our physical path
attribute, emit a GEOM attribute changed event via the
disk_attr_changed() API.
sys/cam/scsi/scsi_pass.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, update
the physical patch devfs alias for this pass instance.
Submitted by: gibbs
Sponsored by: Spectra Logic Corporation
2011-06-14 17:10:32 +00:00
|
|
|
|
2009-01-10 17:22:49 +00:00
|
|
|
cam_periph_unlock(periph);
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* We call taskqueue_drain() for the physpath task to make sure it
|
|
|
|
|
* is complete. We drop the lock because this can potentially
|
|
|
|
|
* sleep. XXX KDM that is bad. Need a way to get a callback when
|
|
|
|
|
* a taskqueue is drained.
|
|
|
|
|
*
|
|
|
|
|
* Note that we don't drain the kqueue shutdown task queue. This
|
|
|
|
|
* is because we hold a reference on the periph for kqueue, and
|
|
|
|
|
* release that reference from the kqueue shutdown task queue. So
|
|
|
|
|
* we cannot come into this routine unless we've released that
|
|
|
|
|
* reference. Also, because that could be the last reference, we
|
|
|
|
|
* could be called from the cam_periph_release() call in
|
|
|
|
|
* pass_shutdown_kqueue(). In that case, the taskqueue_drain()
|
|
|
|
|
* would deadlock. It would be preferable if we had a way to
|
|
|
|
|
* get a callback when a taskqueue is done.
|
|
|
|
|
*/
|
Plumb device physical path reporting from CAM devices, through GEOM and
DEVFS, and make it accessible via the diskinfo utility.
Extend GEOM's generic attribute query mechanism into generic disk consumers.
sys/geom/geom_disk.c:
sys/geom/geom_disk.h:
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Allow disk providers to implement a new method which can override
the default BIO_GETATTR response, d_getattr(struct bio *). This
function returns -1 if not handled, otherwise it returns 0 or an
errno to be passed to g_io_deliver().
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Don't copy the serial number to dp->d_ident anymore, as the CAM XPT
is now responsible for returning this information via
d_getattr()->(a)dagetattr()->xpt_getatr().
sys/geom/geom_dev.c:
- Implement a new ioctl, DIOCGPHYSPATH, which returns the GEOM
attribute "GEOM::physpath", if possible. If the attribute request
returns a zero-length string, ENOENT is returned.
usr.sbin/diskinfo/diskinfo.c:
- If the DIOCGPHYSPATH ioctl is successful, report physical path
data when diskinfo is executed with the '-v' option.
Submitted by: will
Reviewed by: gibbs
Sponsored by: Spectra Logic Corporation
Add generic attribute change notification support to GEOM.
sys/sys/geom/geom.h:
Add a new attrchanged method field to both g_class
and g_geom.
sys/sys/geom/geom.h:
sys/geom/geom_event.c:
- Provide the g_attr_changed() function that providers
can use to advertise attribute changes.
- Perform delivery of attribute change notifications
from a thread context via the standard GEOM event
mechanism.
sys/geom/geom_subr.c:
Inherit the attrchanged method from class to geom (class instance).
sys/geom/geom_disk.c:
Provide disk_attr_changed() to provide g_attr_changed() access
to consumers of the disk API.
sys/cam/scsi/scsi_pass.c:
sys/cam/scsi/scsi_da.c:
sys/geom/geom_dev.c:
sys/geom/geom_disk.c:
Use attribute changed events to track updates to physical path
information.
sys/cam/scsi/scsi_da.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, and
the updated buffer type references our physical path
attribute, emit a GEOM attribute changed event via the
disk_attr_changed() API.
sys/cam/scsi/scsi_pass.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, update
the physical patch devfs alias for this pass instance.
Submitted by: gibbs
Sponsored by: Spectra Logic Corporation
2011-06-14 17:10:32 +00:00
|
|
|
taskqueue_drain(taskqueue_thread, &softc->add_physpath_task);
|
|
|
|
|
|
2009-01-10 17:22:49 +00:00
|
|
|
cam_periph_lock(periph);
|
Plumb device physical path reporting from CAM devices, through GEOM and
DEVFS, and make it accessible via the diskinfo utility.
Extend GEOM's generic attribute query mechanism into generic disk consumers.
sys/geom/geom_disk.c:
sys/geom/geom_disk.h:
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Allow disk providers to implement a new method which can override
the default BIO_GETATTR response, d_getattr(struct bio *). This
function returns -1 if not handled, otherwise it returns 0 or an
errno to be passed to g_io_deliver().
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Don't copy the serial number to dp->d_ident anymore, as the CAM XPT
is now responsible for returning this information via
d_getattr()->(a)dagetattr()->xpt_getatr().
sys/geom/geom_dev.c:
- Implement a new ioctl, DIOCGPHYSPATH, which returns the GEOM
attribute "GEOM::physpath", if possible. If the attribute request
returns a zero-length string, ENOENT is returned.
usr.sbin/diskinfo/diskinfo.c:
- If the DIOCGPHYSPATH ioctl is successful, report physical path
data when diskinfo is executed with the '-v' option.
Submitted by: will
Reviewed by: gibbs
Sponsored by: Spectra Logic Corporation
Add generic attribute change notification support to GEOM.
sys/sys/geom/geom.h:
Add a new attrchanged method field to both g_class
and g_geom.
sys/sys/geom/geom.h:
sys/geom/geom_event.c:
- Provide the g_attr_changed() function that providers
can use to advertise attribute changes.
- Perform delivery of attribute change notifications
from a thread context via the standard GEOM event
mechanism.
sys/geom/geom_subr.c:
Inherit the attrchanged method from class to geom (class instance).
sys/geom/geom_disk.c:
Provide disk_attr_changed() to provide g_attr_changed() access
to consumers of the disk API.
sys/cam/scsi/scsi_pass.c:
sys/cam/scsi/scsi_da.c:
sys/geom/geom_dev.c:
sys/geom/geom_disk.c:
Use attribute changed events to track updates to physical path
information.
sys/cam/scsi/scsi_da.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, and
the updated buffer type references our physical path
attribute, emit a GEOM attribute changed event via the
disk_attr_changed() API.
sys/cam/scsi/scsi_pass.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, update
the physical patch devfs alias for this pass instance.
Submitted by: gibbs
Sponsored by: Spectra Logic Corporation
2011-06-14 17:10:32 +00:00
|
|
|
|
Fix a problem with the way we handled device invalidation when attaching
to a device failed.
In theory, the same steps that happen when we get an AC_LOST_DEVICE async
notification should have been taken when a driver fails to attach. In
practice, that wasn't the case.
This only affected the da, cd and ch drivers, but the fix affects all
peripheral drivers.
There were several possible problems:
- In the da driver, we didn't remove the peripheral's softc from the da
driver's linked list of softcs. Once the peripheral and softc got
removed, we'd get a kernel panic the next time the timeout routine
called dasendorderedtag().
- In the da, cd and possibly ch drivers, we didn't remove the
peripheral's devstat structure from the devstat queue. Once the
peripheral and softc were removed, this could cause a panic if anyone
tried to access device statistics. (one component of the linked list
wouldn't exist anymore)
- In the cd driver, we didn't take the peripheral off the changer run
queue if it was scheduled to run. In practice, it's highly unlikely,
and maybe impossible that the peripheral would have been on the
changer run queue at that stage of the probe process.
The fix is:
- Add a new peripheral callback function (the "oninvalidate" function)
that is called the first time cam_periph_invalidate() is called for a
peripheral.
- Create new foooninvalidate() routines for each peripheral driver. This
routine is always called at splsoftcam(), and contains all the stuff
that used to be in the AC_LOST_DEVICE case of the async callback
handler.
- Move the devstat cleanup call to the destructor/cleanup routines, since
some of the drivers do I/O in their close routines.
- Make sure that when we're flushing the buffer queue, we traverse it at
splbio().
- Add a check for the invalid flag in the pt driver's open routine.
Reviewed by: gibbs
1998-10-22 22:16:56 +00:00
|
|
|
free(softc, M_DEVBUF);
|
1998-09-15 06:36:34 +00:00
|
|
|
}
|
|
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
static void
|
|
|
|
|
pass_shutdown_kqueue(void *context, int pending)
|
|
|
|
|
{
|
|
|
|
|
struct cam_periph *periph;
|
|
|
|
|
struct pass_softc *softc;
|
|
|
|
|
|
|
|
|
|
periph = context;
|
|
|
|
|
softc = periph->softc;
|
|
|
|
|
|
|
|
|
|
knlist_clear(&softc->read_select.si_note, /*is_locked*/ 0);
|
|
|
|
|
knlist_destroy(&softc->read_select.si_note);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Release the reference we held for kqueue.
|
|
|
|
|
*/
|
|
|
|
|
cam_periph_release(periph);
|
|
|
|
|
}
|
|
|
|
|
|
Plumb device physical path reporting from CAM devices, through GEOM and
DEVFS, and make it accessible via the diskinfo utility.
Extend GEOM's generic attribute query mechanism into generic disk consumers.
sys/geom/geom_disk.c:
sys/geom/geom_disk.h:
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Allow disk providers to implement a new method which can override
the default BIO_GETATTR response, d_getattr(struct bio *). This
function returns -1 if not handled, otherwise it returns 0 or an
errno to be passed to g_io_deliver().
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Don't copy the serial number to dp->d_ident anymore, as the CAM XPT
is now responsible for returning this information via
d_getattr()->(a)dagetattr()->xpt_getatr().
sys/geom/geom_dev.c:
- Implement a new ioctl, DIOCGPHYSPATH, which returns the GEOM
attribute "GEOM::physpath", if possible. If the attribute request
returns a zero-length string, ENOENT is returned.
usr.sbin/diskinfo/diskinfo.c:
- If the DIOCGPHYSPATH ioctl is successful, report physical path
data when diskinfo is executed with the '-v' option.
Submitted by: will
Reviewed by: gibbs
Sponsored by: Spectra Logic Corporation
Add generic attribute change notification support to GEOM.
sys/sys/geom/geom.h:
Add a new attrchanged method field to both g_class
and g_geom.
sys/sys/geom/geom.h:
sys/geom/geom_event.c:
- Provide the g_attr_changed() function that providers
can use to advertise attribute changes.
- Perform delivery of attribute change notifications
from a thread context via the standard GEOM event
mechanism.
sys/geom/geom_subr.c:
Inherit the attrchanged method from class to geom (class instance).
sys/geom/geom_disk.c:
Provide disk_attr_changed() to provide g_attr_changed() access
to consumers of the disk API.
sys/cam/scsi/scsi_pass.c:
sys/cam/scsi/scsi_da.c:
sys/geom/geom_dev.c:
sys/geom/geom_disk.c:
Use attribute changed events to track updates to physical path
information.
sys/cam/scsi/scsi_da.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, and
the updated buffer type references our physical path
attribute, emit a GEOM attribute changed event via the
disk_attr_changed() API.
sys/cam/scsi/scsi_pass.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, update
the physical patch devfs alias for this pass instance.
Submitted by: gibbs
Sponsored by: Spectra Logic Corporation
2011-06-14 17:10:32 +00:00
|
|
|
static void
|
|
|
|
|
pass_add_physpath(void *context, int pending)
|
|
|
|
|
{
|
|
|
|
|
struct cam_periph *periph;
|
|
|
|
|
struct pass_softc *softc;
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
struct mtx *mtx;
|
Plumb device physical path reporting from CAM devices, through GEOM and
DEVFS, and make it accessible via the diskinfo utility.
Extend GEOM's generic attribute query mechanism into generic disk consumers.
sys/geom/geom_disk.c:
sys/geom/geom_disk.h:
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Allow disk providers to implement a new method which can override
the default BIO_GETATTR response, d_getattr(struct bio *). This
function returns -1 if not handled, otherwise it returns 0 or an
errno to be passed to g_io_deliver().
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Don't copy the serial number to dp->d_ident anymore, as the CAM XPT
is now responsible for returning this information via
d_getattr()->(a)dagetattr()->xpt_getatr().
sys/geom/geom_dev.c:
- Implement a new ioctl, DIOCGPHYSPATH, which returns the GEOM
attribute "GEOM::physpath", if possible. If the attribute request
returns a zero-length string, ENOENT is returned.
usr.sbin/diskinfo/diskinfo.c:
- If the DIOCGPHYSPATH ioctl is successful, report physical path
data when diskinfo is executed with the '-v' option.
Submitted by: will
Reviewed by: gibbs
Sponsored by: Spectra Logic Corporation
Add generic attribute change notification support to GEOM.
sys/sys/geom/geom.h:
Add a new attrchanged method field to both g_class
and g_geom.
sys/sys/geom/geom.h:
sys/geom/geom_event.c:
- Provide the g_attr_changed() function that providers
can use to advertise attribute changes.
- Perform delivery of attribute change notifications
from a thread context via the standard GEOM event
mechanism.
sys/geom/geom_subr.c:
Inherit the attrchanged method from class to geom (class instance).
sys/geom/geom_disk.c:
Provide disk_attr_changed() to provide g_attr_changed() access
to consumers of the disk API.
sys/cam/scsi/scsi_pass.c:
sys/cam/scsi/scsi_da.c:
sys/geom/geom_dev.c:
sys/geom/geom_disk.c:
Use attribute changed events to track updates to physical path
information.
sys/cam/scsi/scsi_da.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, and
the updated buffer type references our physical path
attribute, emit a GEOM attribute changed event via the
disk_attr_changed() API.
sys/cam/scsi/scsi_pass.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, update
the physical patch devfs alias for this pass instance.
Submitted by: gibbs
Sponsored by: Spectra Logic Corporation
2011-06-14 17:10:32 +00:00
|
|
|
char *physpath;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* If we have one, create a devfs alias for our
|
|
|
|
|
* physical path.
|
|
|
|
|
*/
|
|
|
|
|
periph = context;
|
|
|
|
|
softc = periph->softc;
|
2012-10-12 17:18:24 +00:00
|
|
|
physpath = malloc(MAXPATHLEN, M_DEVBUF, M_WAITOK);
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
mtx = cam_periph_mtx(periph);
|
|
|
|
|
mtx_lock(mtx);
|
|
|
|
|
|
|
|
|
|
if (periph->flags & CAM_PERIPH_INVALID)
|
2012-10-12 17:18:24 +00:00
|
|
|
goto out;
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
|
Plumb device physical path reporting from CAM devices, through GEOM and
DEVFS, and make it accessible via the diskinfo utility.
Extend GEOM's generic attribute query mechanism into generic disk consumers.
sys/geom/geom_disk.c:
sys/geom/geom_disk.h:
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Allow disk providers to implement a new method which can override
the default BIO_GETATTR response, d_getattr(struct bio *). This
function returns -1 if not handled, otherwise it returns 0 or an
errno to be passed to g_io_deliver().
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Don't copy the serial number to dp->d_ident anymore, as the CAM XPT
is now responsible for returning this information via
d_getattr()->(a)dagetattr()->xpt_getatr().
sys/geom/geom_dev.c:
- Implement a new ioctl, DIOCGPHYSPATH, which returns the GEOM
attribute "GEOM::physpath", if possible. If the attribute request
returns a zero-length string, ENOENT is returned.
usr.sbin/diskinfo/diskinfo.c:
- If the DIOCGPHYSPATH ioctl is successful, report physical path
data when diskinfo is executed with the '-v' option.
Submitted by: will
Reviewed by: gibbs
Sponsored by: Spectra Logic Corporation
Add generic attribute change notification support to GEOM.
sys/sys/geom/geom.h:
Add a new attrchanged method field to both g_class
and g_geom.
sys/sys/geom/geom.h:
sys/geom/geom_event.c:
- Provide the g_attr_changed() function that providers
can use to advertise attribute changes.
- Perform delivery of attribute change notifications
from a thread context via the standard GEOM event
mechanism.
sys/geom/geom_subr.c:
Inherit the attrchanged method from class to geom (class instance).
sys/geom/geom_disk.c:
Provide disk_attr_changed() to provide g_attr_changed() access
to consumers of the disk API.
sys/cam/scsi/scsi_pass.c:
sys/cam/scsi/scsi_da.c:
sys/geom/geom_dev.c:
sys/geom/geom_disk.c:
Use attribute changed events to track updates to physical path
information.
sys/cam/scsi/scsi_da.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, and
the updated buffer type references our physical path
attribute, emit a GEOM attribute changed event via the
disk_attr_changed() API.
sys/cam/scsi/scsi_pass.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, update
the physical patch devfs alias for this pass instance.
Submitted by: gibbs
Sponsored by: Spectra Logic Corporation
2011-06-14 17:10:32 +00:00
|
|
|
if (xpt_getattr(physpath, MAXPATHLEN,
|
|
|
|
|
"GEOM::physpath", periph->path) == 0
|
|
|
|
|
&& strlen(physpath) != 0) {
|
|
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
mtx_unlock(mtx);
|
Plumb device physical path reporting from CAM devices, through GEOM and
DEVFS, and make it accessible via the diskinfo utility.
Extend GEOM's generic attribute query mechanism into generic disk consumers.
sys/geom/geom_disk.c:
sys/geom/geom_disk.h:
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Allow disk providers to implement a new method which can override
the default BIO_GETATTR response, d_getattr(struct bio *). This
function returns -1 if not handled, otherwise it returns 0 or an
errno to be passed to g_io_deliver().
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Don't copy the serial number to dp->d_ident anymore, as the CAM XPT
is now responsible for returning this information via
d_getattr()->(a)dagetattr()->xpt_getatr().
sys/geom/geom_dev.c:
- Implement a new ioctl, DIOCGPHYSPATH, which returns the GEOM
attribute "GEOM::physpath", if possible. If the attribute request
returns a zero-length string, ENOENT is returned.
usr.sbin/diskinfo/diskinfo.c:
- If the DIOCGPHYSPATH ioctl is successful, report physical path
data when diskinfo is executed with the '-v' option.
Submitted by: will
Reviewed by: gibbs
Sponsored by: Spectra Logic Corporation
Add generic attribute change notification support to GEOM.
sys/sys/geom/geom.h:
Add a new attrchanged method field to both g_class
and g_geom.
sys/sys/geom/geom.h:
sys/geom/geom_event.c:
- Provide the g_attr_changed() function that providers
can use to advertise attribute changes.
- Perform delivery of attribute change notifications
from a thread context via the standard GEOM event
mechanism.
sys/geom/geom_subr.c:
Inherit the attrchanged method from class to geom (class instance).
sys/geom/geom_disk.c:
Provide disk_attr_changed() to provide g_attr_changed() access
to consumers of the disk API.
sys/cam/scsi/scsi_pass.c:
sys/cam/scsi/scsi_da.c:
sys/geom/geom_dev.c:
sys/geom/geom_disk.c:
Use attribute changed events to track updates to physical path
information.
sys/cam/scsi/scsi_da.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, and
the updated buffer type references our physical path
attribute, emit a GEOM attribute changed event via the
disk_attr_changed() API.
sys/cam/scsi/scsi_pass.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, update
the physical patch devfs alias for this pass instance.
Submitted by: gibbs
Sponsored by: Spectra Logic Corporation
2011-06-14 17:10:32 +00:00
|
|
|
make_dev_physpath_alias(MAKEDEV_WAITOK, &softc->alias_dev,
|
|
|
|
|
softc->dev, softc->alias_dev, physpath);
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
mtx_lock(mtx);
|
Plumb device physical path reporting from CAM devices, through GEOM and
DEVFS, and make it accessible via the diskinfo utility.
Extend GEOM's generic attribute query mechanism into generic disk consumers.
sys/geom/geom_disk.c:
sys/geom/geom_disk.h:
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Allow disk providers to implement a new method which can override
the default BIO_GETATTR response, d_getattr(struct bio *). This
function returns -1 if not handled, otherwise it returns 0 or an
errno to be passed to g_io_deliver().
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Don't copy the serial number to dp->d_ident anymore, as the CAM XPT
is now responsible for returning this information via
d_getattr()->(a)dagetattr()->xpt_getatr().
sys/geom/geom_dev.c:
- Implement a new ioctl, DIOCGPHYSPATH, which returns the GEOM
attribute "GEOM::physpath", if possible. If the attribute request
returns a zero-length string, ENOENT is returned.
usr.sbin/diskinfo/diskinfo.c:
- If the DIOCGPHYSPATH ioctl is successful, report physical path
data when diskinfo is executed with the '-v' option.
Submitted by: will
Reviewed by: gibbs
Sponsored by: Spectra Logic Corporation
Add generic attribute change notification support to GEOM.
sys/sys/geom/geom.h:
Add a new attrchanged method field to both g_class
and g_geom.
sys/sys/geom/geom.h:
sys/geom/geom_event.c:
- Provide the g_attr_changed() function that providers
can use to advertise attribute changes.
- Perform delivery of attribute change notifications
from a thread context via the standard GEOM event
mechanism.
sys/geom/geom_subr.c:
Inherit the attrchanged method from class to geom (class instance).
sys/geom/geom_disk.c:
Provide disk_attr_changed() to provide g_attr_changed() access
to consumers of the disk API.
sys/cam/scsi/scsi_pass.c:
sys/cam/scsi/scsi_da.c:
sys/geom/geom_dev.c:
sys/geom/geom_disk.c:
Use attribute changed events to track updates to physical path
information.
sys/cam/scsi/scsi_da.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, and
the updated buffer type references our physical path
attribute, emit a GEOM attribute changed event via the
disk_attr_changed() API.
sys/cam/scsi/scsi_pass.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, update
the physical patch devfs alias for this pass instance.
Submitted by: gibbs
Sponsored by: Spectra Logic Corporation
2011-06-14 17:10:32 +00:00
|
|
|
}
|
Fix several reference counting and object lifetime issues between
the pass(4) and enc(4) drivers and devfs.
The pass(4) driver uses the destroy_dev_sched() routine to
schedule its device node for destruction in a separate thread
context. It does this because the passcleanup() routine can get
called indirectly from the passclose() routine, and that would
cause a deadlock if the close routine tried to destroy its own
device node.
In any case, once a particular passthrough driver number, e.g.
pass3, is destroyed, CAM considers that unit number (3 in this
case) available for reuse.
The problem is that devfs may not be done cleaning up the previous
instance of pass3, and will panic if isn't done cleaning up the
previous instance.
The solution is to get a callback from devfs when the device node
is removed, and make sure we hold a reference to the peripheral
until that happens.
Testing exposed some other cases where we have reference counting
issues, and those were also fixed in the pass(4) driver.
cam_periph.c: In camperiphfree(), reorder some of the operations.
The peripheral destructor needs to be called before
the peripheral is removed from the peripheral is
removed from the list. This is because once we
remove the peripheral from the list, and drop the
topology lock, the peripheral number may be reused.
But if the destructor hasn't been called yet, there
may still be resources hanging around (like devfs
nodes) that haven't been fully cleaned up.
cam_xpt.c: Add an argument to xpt_remove_periph() to indicate
whether the topology lock is already held.
scsi_enc.c: Acquire an extra reference to the peripheral during
registration, and release it once we get a callback
from devfs indicating that the device node is gone.
Call destroy_dev_sched_cb() in enc_oninvalidate()
instead of calling destroy_dev() in the cleanup
routine.
scsi_pass.c: Add reference counting to handle peripheral and
devfs object lifetime issues.
Add a reference to the peripheral and the devfs
node in the peripheral registration.
Don't attempt to add a physical path alias if the
peripheral has been marked invalid.
Release the devfs reference once the initial
physical path alias taskqueue run has completed.
Schedule devfs node destruction in the
passoninvalidate(), and release our peripheral
reference in a new routine, passdevgonecb() once
the devfs node is gone. This allows the peripheral
to fully go away, and the peripheral destructor,
passcleanup(), will get called.
MFC after: 3 days
Sponsored by: Spectra Logic
2012-06-20 17:08:00 +00:00
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
out:
|
Fix several reference counting and object lifetime issues between
the pass(4) and enc(4) drivers and devfs.
The pass(4) driver uses the destroy_dev_sched() routine to
schedule its device node for destruction in a separate thread
context. It does this because the passcleanup() routine can get
called indirectly from the passclose() routine, and that would
cause a deadlock if the close routine tried to destroy its own
device node.
In any case, once a particular passthrough driver number, e.g.
pass3, is destroyed, CAM considers that unit number (3 in this
case) available for reuse.
The problem is that devfs may not be done cleaning up the previous
instance of pass3, and will panic if isn't done cleaning up the
previous instance.
The solution is to get a callback from devfs when the device node
is removed, and make sure we hold a reference to the peripheral
until that happens.
Testing exposed some other cases where we have reference counting
issues, and those were also fixed in the pass(4) driver.
cam_periph.c: In camperiphfree(), reorder some of the operations.
The peripheral destructor needs to be called before
the peripheral is removed from the peripheral is
removed from the list. This is because once we
remove the peripheral from the list, and drop the
topology lock, the peripheral number may be reused.
But if the destructor hasn't been called yet, there
may still be resources hanging around (like devfs
nodes) that haven't been fully cleaned up.
cam_xpt.c: Add an argument to xpt_remove_periph() to indicate
whether the topology lock is already held.
scsi_enc.c: Acquire an extra reference to the peripheral during
registration, and release it once we get a callback
from devfs indicating that the device node is gone.
Call destroy_dev_sched_cb() in enc_oninvalidate()
instead of calling destroy_dev() in the cleanup
routine.
scsi_pass.c: Add reference counting to handle peripheral and
devfs object lifetime issues.
Add a reference to the peripheral and the devfs
node in the peripheral registration.
Don't attempt to add a physical path alias if the
peripheral has been marked invalid.
Release the devfs reference once the initial
physical path alias taskqueue run has completed.
Schedule devfs node destruction in the
passoninvalidate(), and release our peripheral
reference in a new routine, passdevgonecb() once
the devfs node is gone. This allows the peripheral
to fully go away, and the peripheral destructor,
passcleanup(), will get called.
MFC after: 3 days
Sponsored by: Spectra Logic
2012-06-20 17:08:00 +00:00
|
|
|
/*
|
|
|
|
|
* Now that we've made our alias, we no longer have to have a
|
|
|
|
|
* reference to the device.
|
|
|
|
|
*/
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
if ((softc->flags & PASS_FLAG_INITIAL_PHYSPATH) == 0)
|
Fix several reference counting and object lifetime issues between
the pass(4) and enc(4) drivers and devfs.
The pass(4) driver uses the destroy_dev_sched() routine to
schedule its device node for destruction in a separate thread
context. It does this because the passcleanup() routine can get
called indirectly from the passclose() routine, and that would
cause a deadlock if the close routine tried to destroy its own
device node.
In any case, once a particular passthrough driver number, e.g.
pass3, is destroyed, CAM considers that unit number (3 in this
case) available for reuse.
The problem is that devfs may not be done cleaning up the previous
instance of pass3, and will panic if isn't done cleaning up the
previous instance.
The solution is to get a callback from devfs when the device node
is removed, and make sure we hold a reference to the peripheral
until that happens.
Testing exposed some other cases where we have reference counting
issues, and those were also fixed in the pass(4) driver.
cam_periph.c: In camperiphfree(), reorder some of the operations.
The peripheral destructor needs to be called before
the peripheral is removed from the peripheral is
removed from the list. This is because once we
remove the peripheral from the list, and drop the
topology lock, the peripheral number may be reused.
But if the destructor hasn't been called yet, there
may still be resources hanging around (like devfs
nodes) that haven't been fully cleaned up.
cam_xpt.c: Add an argument to xpt_remove_periph() to indicate
whether the topology lock is already held.
scsi_enc.c: Acquire an extra reference to the peripheral during
registration, and release it once we get a callback
from devfs indicating that the device node is gone.
Call destroy_dev_sched_cb() in enc_oninvalidate()
instead of calling destroy_dev() in the cleanup
routine.
scsi_pass.c: Add reference counting to handle peripheral and
devfs object lifetime issues.
Add a reference to the peripheral and the devfs
node in the peripheral registration.
Don't attempt to add a physical path alias if the
peripheral has been marked invalid.
Release the devfs reference once the initial
physical path alias taskqueue run has completed.
Schedule devfs node destruction in the
passoninvalidate(), and release our peripheral
reference in a new routine, passdevgonecb() once
the devfs node is gone. This allows the peripheral
to fully go away, and the peripheral destructor,
passcleanup(), will get called.
MFC after: 3 days
Sponsored by: Spectra Logic
2012-06-20 17:08:00 +00:00
|
|
|
softc->flags |= PASS_FLAG_INITIAL_PHYSPATH;
|
2012-10-12 17:18:24 +00:00
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
/*
|
|
|
|
|
* We always acquire a reference to the periph before queueing this
|
|
|
|
|
* task queue function, so it won't go away before we run.
|
|
|
|
|
*/
|
|
|
|
|
while (pending-- > 0)
|
|
|
|
|
cam_periph_release_locked(periph);
|
|
|
|
|
mtx_unlock(mtx);
|
|
|
|
|
|
2012-10-12 17:18:24 +00:00
|
|
|
free(physpath, M_DEVBUF);
|
Plumb device physical path reporting from CAM devices, through GEOM and
DEVFS, and make it accessible via the diskinfo utility.
Extend GEOM's generic attribute query mechanism into generic disk consumers.
sys/geom/geom_disk.c:
sys/geom/geom_disk.h:
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Allow disk providers to implement a new method which can override
the default BIO_GETATTR response, d_getattr(struct bio *). This
function returns -1 if not handled, otherwise it returns 0 or an
errno to be passed to g_io_deliver().
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Don't copy the serial number to dp->d_ident anymore, as the CAM XPT
is now responsible for returning this information via
d_getattr()->(a)dagetattr()->xpt_getatr().
sys/geom/geom_dev.c:
- Implement a new ioctl, DIOCGPHYSPATH, which returns the GEOM
attribute "GEOM::physpath", if possible. If the attribute request
returns a zero-length string, ENOENT is returned.
usr.sbin/diskinfo/diskinfo.c:
- If the DIOCGPHYSPATH ioctl is successful, report physical path
data when diskinfo is executed with the '-v' option.
Submitted by: will
Reviewed by: gibbs
Sponsored by: Spectra Logic Corporation
Add generic attribute change notification support to GEOM.
sys/sys/geom/geom.h:
Add a new attrchanged method field to both g_class
and g_geom.
sys/sys/geom/geom.h:
sys/geom/geom_event.c:
- Provide the g_attr_changed() function that providers
can use to advertise attribute changes.
- Perform delivery of attribute change notifications
from a thread context via the standard GEOM event
mechanism.
sys/geom/geom_subr.c:
Inherit the attrchanged method from class to geom (class instance).
sys/geom/geom_disk.c:
Provide disk_attr_changed() to provide g_attr_changed() access
to consumers of the disk API.
sys/cam/scsi/scsi_pass.c:
sys/cam/scsi/scsi_da.c:
sys/geom/geom_dev.c:
sys/geom/geom_disk.c:
Use attribute changed events to track updates to physical path
information.
sys/cam/scsi/scsi_da.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, and
the updated buffer type references our physical path
attribute, emit a GEOM attribute changed event via the
disk_attr_changed() API.
sys/cam/scsi/scsi_pass.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, update
the physical patch devfs alias for this pass instance.
Submitted by: gibbs
Sponsored by: Spectra Logic Corporation
2011-06-14 17:10:32 +00:00
|
|
|
}
|
|
|
|
|
|
1998-09-15 06:36:34 +00:00
|
|
|
static void
|
|
|
|
|
passasync(void *callback_arg, u_int32_t code,
|
|
|
|
|
struct cam_path *path, void *arg)
|
|
|
|
|
{
|
|
|
|
|
struct cam_periph *periph;
|
|
|
|
|
|
|
|
|
|
periph = (struct cam_periph *)callback_arg;
|
|
|
|
|
|
|
|
|
|
switch (code) {
|
|
|
|
|
case AC_FOUND_DEVICE:
|
|
|
|
|
{
|
|
|
|
|
struct ccb_getdev *cgd;
|
|
|
|
|
cam_status status;
|
|
|
|
|
|
|
|
|
|
cgd = (struct ccb_getdev *)arg;
|
2001-07-04 05:22:42 +00:00
|
|
|
if (cgd == NULL)
|
|
|
|
|
break;
|
1998-09-15 06:36:34 +00:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Allocate a peripheral instance for
|
|
|
|
|
* this device and start the probe
|
|
|
|
|
* process.
|
|
|
|
|
*/
|
Fix a problem with the way we handled device invalidation when attaching
to a device failed.
In theory, the same steps that happen when we get an AC_LOST_DEVICE async
notification should have been taken when a driver fails to attach. In
practice, that wasn't the case.
This only affected the da, cd and ch drivers, but the fix affects all
peripheral drivers.
There were several possible problems:
- In the da driver, we didn't remove the peripheral's softc from the da
driver's linked list of softcs. Once the peripheral and softc got
removed, we'd get a kernel panic the next time the timeout routine
called dasendorderedtag().
- In the da, cd and possibly ch drivers, we didn't remove the
peripheral's devstat structure from the devstat queue. Once the
peripheral and softc were removed, this could cause a panic if anyone
tried to access device statistics. (one component of the linked list
wouldn't exist anymore)
- In the cd driver, we didn't take the peripheral off the changer run
queue if it was scheduled to run. In practice, it's highly unlikely,
and maybe impossible that the peripheral would have been on the
changer run queue at that stage of the probe process.
The fix is:
- Add a new peripheral callback function (the "oninvalidate" function)
that is called the first time cam_periph_invalidate() is called for a
peripheral.
- Create new foooninvalidate() routines for each peripheral driver. This
routine is always called at splsoftcam(), and contains all the stuff
that used to be in the AC_LOST_DEVICE case of the async callback
handler.
- Move the devstat cleanup call to the destructor/cleanup routines, since
some of the drivers do I/O in their close routines.
- Make sure that when we're flushing the buffer queue, we traverse it at
splbio().
- Add a check for the invalid flag in the pt driver's open routine.
Reviewed by: gibbs
1998-10-22 22:16:56 +00:00
|
|
|
status = cam_periph_alloc(passregister, passoninvalidate,
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
passcleanup, passstart, "pass",
|
Merge CAM locking changes from the projects/camlock branch to radically
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
2013-10-21 12:00:26 +00:00
|
|
|
CAM_PERIPH_BIO, path,
|
Fix a problem with the way we handled device invalidation when attaching
to a device failed.
In theory, the same steps that happen when we get an AC_LOST_DEVICE async
notification should have been taken when a driver fails to attach. In
practice, that wasn't the case.
This only affected the da, cd and ch drivers, but the fix affects all
peripheral drivers.
There were several possible problems:
- In the da driver, we didn't remove the peripheral's softc from the da
driver's linked list of softcs. Once the peripheral and softc got
removed, we'd get a kernel panic the next time the timeout routine
called dasendorderedtag().
- In the da, cd and possibly ch drivers, we didn't remove the
peripheral's devstat structure from the devstat queue. Once the
peripheral and softc were removed, this could cause a panic if anyone
tried to access device statistics. (one component of the linked list
wouldn't exist anymore)
- In the cd driver, we didn't take the peripheral off the changer run
queue if it was scheduled to run. In practice, it's highly unlikely,
and maybe impossible that the peripheral would have been on the
changer run queue at that stage of the probe process.
The fix is:
- Add a new peripheral callback function (the "oninvalidate" function)
that is called the first time cam_periph_invalidate() is called for a
peripheral.
- Create new foooninvalidate() routines for each peripheral driver. This
routine is always called at splsoftcam(), and contains all the stuff
that used to be in the AC_LOST_DEVICE case of the async callback
handler.
- Move the devstat cleanup call to the destructor/cleanup routines, since
some of the drivers do I/O in their close routines.
- Make sure that when we're flushing the buffer queue, we traverse it at
splbio().
- Add a check for the invalid flag in the pt driver's open routine.
Reviewed by: gibbs
1998-10-22 22:16:56 +00:00
|
|
|
passasync, AC_FOUND_DEVICE, cgd);
|
1998-09-15 06:36:34 +00:00
|
|
|
|
|
|
|
|
if (status != CAM_REQ_CMP
|
Rewrite of the CAM error recovery code.
Some of the major changes include:
- The SCSI error handling portion of cam_periph_error() has
been broken out into a number of subfunctions to better
modularize the code that handles the hierarchy of SCSI errors.
As a result, the code is now much easier to read.
- String handling and error printing has been significantly
revamped. We now use sbufs to do string formatting instead
of using printfs (for the kernel) and snprintf/strncat (for
userland) as before.
There is a new catchall error printing routine,
cam_error_print() and its string-based counterpart,
cam_error_string() that allow the kernel and userland
applications to pass in a CCB and have errors printed out
properly, whether or not they're SCSI errors. Among other
things, this helped eliminate a fair amount of duplicate code
in camcontrol.
We now print out more information than before, including
the CAM status and SCSI status and the error recovery action
taken to remedy the problem.
- sbufs are now available in userland, via libsbuf. This
change was necessary since most of the error printing code
is shared between libcam and the kernel.
- A new transfer settings interface is included in this checkin.
This code is #ifdef'ed out, and is primarily intended to aid
discussion with HBA driver authors on the final form the
interface should take. There is example code in the ahc(4)
driver that implements the HBA driver side of the new
interface. The new transfer settings code won't be enabled
until we're ready to switch all HBA drivers over to the new
interface.
src/Makefile.inc1,
lib/Makefile: Add libsbuf. It must be built before libcam,
since libcam uses sbuf routines.
libcam/Makefile: libcam now depends on libsbuf.
libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the
sbuf sources from sys/kern.
bsd.libnames.mk: Add LIBSBUF.
camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically
linked, we can't depend on the dynamic linker
to pull in libsbuf.
camcontrol.c: Use cam_error_print() instead of checking for
CAM_SCSI_STATUS_ERROR on every failed CCB.
sbuf.9: Change the prototypes for sbuf_cat() and
sbuf_cpy() so that the source string is now a
const char *. This is more in line wth the
standard system string functions, and helps
eliminate warnings when dealing with a const
source buffer.
Fix a typo.
cam.c: Add description strings for the various CAM
error status values, as well as routines to
look up those strings.
Add new cam_error_string() and
cam_error_print() routines for userland and
the kernel.
cam.h: Add a new CAM flag, CAM_RETRY_SELTO.
Add enumerated types for the various options
available with cam_error_print() and
cam_error_string().
cam_ccb.h: Add new transfer negotiation structures/types.
Change inq_len in the ccb_getdev structure to
be "reserved". This field has never been
filled in, and will be removed when we next
bump the CAM version.
cam_debug.h: Fix typo.
cam_periph.c: Modularize cam_periph_error(). The SCSI error
handling part of cam_periph_error() is now
in camperiphscsistatuserror() and
camperiphscsisenseerror().
In cam_periph_lock(), increase the reference
count on the periph while we wait for our lock
attempt to succeed so that the periph won't go
away while we're sleeping.
cam_xpt.c: Add new transfer negotiation code. (ifdefed
out)
Add a new function, xpt_path_string(). This
is a string/sbuf analog to xpt_print_path().
scsi_all.c: Revamp string handing and error printing code.
We now use sbufs for much of the string
formatting code. More of that code is shared
between userland the kernel.
scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly
useful in the first place.
Add a new error action, SS_REQSENSE. (Send a
request sense and then retry the command.)
This is useful when the controller hasn't
performed autosense for some reason.
Change the default actions around a bit.
scsi_cd.c,
scsi_da.c,
scsi_pt.c,
scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection
timeouts shouldn't be covered by a sense flag.
scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO.
Get rid of the last vestiges of a read/write
interface.
libkern/bsearch.c,
sys/libkern.h,
conf/files: Add bsearch.c, which is needed for some of the
new table lookup routines.
aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if
CAM_NEW_TRAN_CODE is defined.
sbuf.h,
subr_sbuf.c: Add the appropriate #ifdefs so sbufs can
compile and run in userland.
Change sbuf_printf() to use vsnprintf()
instead of kvprintf(), which is only available
in the kernel.
Change the source string for sbuf_cpy() and
sbuf_cat() to be a const char *.
Add __BEGIN_DECLS and __END_DECLS around
function prototypes since they're now exported
to userland.
kdump/mkioctls: Include stdio.h before cam.h since cam.h now
includes a function with a FILE * argument.
Submitted by: gibbs (mostly)
Reviewed by: jdp, marcel (libsbuf makefile changes)
Reviewed by: des (sbuf changes)
Reviewed by: ken
2001-03-27 05:45:52 +00:00
|
|
|
&& status != CAM_REQ_INPROG) {
|
|
|
|
|
const struct cam_status_entry *entry;
|
|
|
|
|
|
|
|
|
|
entry = cam_fetch_status_entry(status);
|
|
|
|
|
|
1998-09-15 06:36:34 +00:00
|
|
|
printf("passasync: Unable to attach new device "
|
Rewrite of the CAM error recovery code.
Some of the major changes include:
- The SCSI error handling portion of cam_periph_error() has
been broken out into a number of subfunctions to better
modularize the code that handles the hierarchy of SCSI errors.
As a result, the code is now much easier to read.
- String handling and error printing has been significantly
revamped. We now use sbufs to do string formatting instead
of using printfs (for the kernel) and snprintf/strncat (for
userland) as before.
There is a new catchall error printing routine,
cam_error_print() and its string-based counterpart,
cam_error_string() that allow the kernel and userland
applications to pass in a CCB and have errors printed out
properly, whether or not they're SCSI errors. Among other
things, this helped eliminate a fair amount of duplicate code
in camcontrol.
We now print out more information than before, including
the CAM status and SCSI status and the error recovery action
taken to remedy the problem.
- sbufs are now available in userland, via libsbuf. This
change was necessary since most of the error printing code
is shared between libcam and the kernel.
- A new transfer settings interface is included in this checkin.
This code is #ifdef'ed out, and is primarily intended to aid
discussion with HBA driver authors on the final form the
interface should take. There is example code in the ahc(4)
driver that implements the HBA driver side of the new
interface. The new transfer settings code won't be enabled
until we're ready to switch all HBA drivers over to the new
interface.
src/Makefile.inc1,
lib/Makefile: Add libsbuf. It must be built before libcam,
since libcam uses sbuf routines.
libcam/Makefile: libcam now depends on libsbuf.
libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the
sbuf sources from sys/kern.
bsd.libnames.mk: Add LIBSBUF.
camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically
linked, we can't depend on the dynamic linker
to pull in libsbuf.
camcontrol.c: Use cam_error_print() instead of checking for
CAM_SCSI_STATUS_ERROR on every failed CCB.
sbuf.9: Change the prototypes for sbuf_cat() and
sbuf_cpy() so that the source string is now a
const char *. This is more in line wth the
standard system string functions, and helps
eliminate warnings when dealing with a const
source buffer.
Fix a typo.
cam.c: Add description strings for the various CAM
error status values, as well as routines to
look up those strings.
Add new cam_error_string() and
cam_error_print() routines for userland and
the kernel.
cam.h: Add a new CAM flag, CAM_RETRY_SELTO.
Add enumerated types for the various options
available with cam_error_print() and
cam_error_string().
cam_ccb.h: Add new transfer negotiation structures/types.
Change inq_len in the ccb_getdev structure to
be "reserved". This field has never been
filled in, and will be removed when we next
bump the CAM version.
cam_debug.h: Fix typo.
cam_periph.c: Modularize cam_periph_error(). The SCSI error
handling part of cam_periph_error() is now
in camperiphscsistatuserror() and
camperiphscsisenseerror().
In cam_periph_lock(), increase the reference
count on the periph while we wait for our lock
attempt to succeed so that the periph won't go
away while we're sleeping.
cam_xpt.c: Add new transfer negotiation code. (ifdefed
out)
Add a new function, xpt_path_string(). This
is a string/sbuf analog to xpt_print_path().
scsi_all.c: Revamp string handing and error printing code.
We now use sbufs for much of the string
formatting code. More of that code is shared
between userland the kernel.
scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly
useful in the first place.
Add a new error action, SS_REQSENSE. (Send a
request sense and then retry the command.)
This is useful when the controller hasn't
performed autosense for some reason.
Change the default actions around a bit.
scsi_cd.c,
scsi_da.c,
scsi_pt.c,
scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection
timeouts shouldn't be covered by a sense flag.
scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO.
Get rid of the last vestiges of a read/write
interface.
libkern/bsearch.c,
sys/libkern.h,
conf/files: Add bsearch.c, which is needed for some of the
new table lookup routines.
aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if
CAM_NEW_TRAN_CODE is defined.
sbuf.h,
subr_sbuf.c: Add the appropriate #ifdefs so sbufs can
compile and run in userland.
Change sbuf_printf() to use vsnprintf()
instead of kvprintf(), which is only available
in the kernel.
Change the source string for sbuf_cpy() and
sbuf_cat() to be a const char *.
Add __BEGIN_DECLS and __END_DECLS around
function prototypes since they're now exported
to userland.
kdump/mkioctls: Include stdio.h before cam.h since cam.h now
includes a function with a FILE * argument.
Submitted by: gibbs (mostly)
Reviewed by: jdp, marcel (libsbuf makefile changes)
Reviewed by: des (sbuf changes)
Reviewed by: ken
2001-03-27 05:45:52 +00:00
|
|
|
"due to status %#x: %s\n", status, entry ?
|
|
|
|
|
entry->status_text : "Unknown");
|
|
|
|
|
}
|
1998-09-15 06:36:34 +00:00
|
|
|
|
|
|
|
|
break;
|
|
|
|
|
}
|
Plumb device physical path reporting from CAM devices, through GEOM and
DEVFS, and make it accessible via the diskinfo utility.
Extend GEOM's generic attribute query mechanism into generic disk consumers.
sys/geom/geom_disk.c:
sys/geom/geom_disk.h:
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Allow disk providers to implement a new method which can override
the default BIO_GETATTR response, d_getattr(struct bio *). This
function returns -1 if not handled, otherwise it returns 0 or an
errno to be passed to g_io_deliver().
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Don't copy the serial number to dp->d_ident anymore, as the CAM XPT
is now responsible for returning this information via
d_getattr()->(a)dagetattr()->xpt_getatr().
sys/geom/geom_dev.c:
- Implement a new ioctl, DIOCGPHYSPATH, which returns the GEOM
attribute "GEOM::physpath", if possible. If the attribute request
returns a zero-length string, ENOENT is returned.
usr.sbin/diskinfo/diskinfo.c:
- If the DIOCGPHYSPATH ioctl is successful, report physical path
data when diskinfo is executed with the '-v' option.
Submitted by: will
Reviewed by: gibbs
Sponsored by: Spectra Logic Corporation
Add generic attribute change notification support to GEOM.
sys/sys/geom/geom.h:
Add a new attrchanged method field to both g_class
and g_geom.
sys/sys/geom/geom.h:
sys/geom/geom_event.c:
- Provide the g_attr_changed() function that providers
can use to advertise attribute changes.
- Perform delivery of attribute change notifications
from a thread context via the standard GEOM event
mechanism.
sys/geom/geom_subr.c:
Inherit the attrchanged method from class to geom (class instance).
sys/geom/geom_disk.c:
Provide disk_attr_changed() to provide g_attr_changed() access
to consumers of the disk API.
sys/cam/scsi/scsi_pass.c:
sys/cam/scsi/scsi_da.c:
sys/geom/geom_dev.c:
sys/geom/geom_disk.c:
Use attribute changed events to track updates to physical path
information.
sys/cam/scsi/scsi_da.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, and
the updated buffer type references our physical path
attribute, emit a GEOM attribute changed event via the
disk_attr_changed() API.
sys/cam/scsi/scsi_pass.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, update
the physical patch devfs alias for this pass instance.
Submitted by: gibbs
Sponsored by: Spectra Logic Corporation
2011-06-14 17:10:32 +00:00
|
|
|
case AC_ADVINFO_CHANGED:
|
|
|
|
|
{
|
|
|
|
|
uintptr_t buftype;
|
|
|
|
|
|
|
|
|
|
buftype = (uintptr_t)arg;
|
|
|
|
|
if (buftype == CDAI_TYPE_PHYS_PATH) {
|
|
|
|
|
struct pass_softc *softc;
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
cam_status status;
|
Plumb device physical path reporting from CAM devices, through GEOM and
DEVFS, and make it accessible via the diskinfo utility.
Extend GEOM's generic attribute query mechanism into generic disk consumers.
sys/geom/geom_disk.c:
sys/geom/geom_disk.h:
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Allow disk providers to implement a new method which can override
the default BIO_GETATTR response, d_getattr(struct bio *). This
function returns -1 if not handled, otherwise it returns 0 or an
errno to be passed to g_io_deliver().
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Don't copy the serial number to dp->d_ident anymore, as the CAM XPT
is now responsible for returning this information via
d_getattr()->(a)dagetattr()->xpt_getatr().
sys/geom/geom_dev.c:
- Implement a new ioctl, DIOCGPHYSPATH, which returns the GEOM
attribute "GEOM::physpath", if possible. If the attribute request
returns a zero-length string, ENOENT is returned.
usr.sbin/diskinfo/diskinfo.c:
- If the DIOCGPHYSPATH ioctl is successful, report physical path
data when diskinfo is executed with the '-v' option.
Submitted by: will
Reviewed by: gibbs
Sponsored by: Spectra Logic Corporation
Add generic attribute change notification support to GEOM.
sys/sys/geom/geom.h:
Add a new attrchanged method field to both g_class
and g_geom.
sys/sys/geom/geom.h:
sys/geom/geom_event.c:
- Provide the g_attr_changed() function that providers
can use to advertise attribute changes.
- Perform delivery of attribute change notifications
from a thread context via the standard GEOM event
mechanism.
sys/geom/geom_subr.c:
Inherit the attrchanged method from class to geom (class instance).
sys/geom/geom_disk.c:
Provide disk_attr_changed() to provide g_attr_changed() access
to consumers of the disk API.
sys/cam/scsi/scsi_pass.c:
sys/cam/scsi/scsi_da.c:
sys/geom/geom_dev.c:
sys/geom/geom_disk.c:
Use attribute changed events to track updates to physical path
information.
sys/cam/scsi/scsi_da.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, and
the updated buffer type references our physical path
attribute, emit a GEOM attribute changed event via the
disk_attr_changed() API.
sys/cam/scsi/scsi_pass.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, update
the physical patch devfs alias for this pass instance.
Submitted by: gibbs
Sponsored by: Spectra Logic Corporation
2011-06-14 17:10:32 +00:00
|
|
|
|
|
|
|
|
softc = (struct pass_softc *)periph->softc;
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
/*
|
|
|
|
|
* Acquire a reference to the periph before we
|
|
|
|
|
* start the taskqueue, so that we don't run into
|
|
|
|
|
* a situation where the periph goes away before
|
|
|
|
|
* the task queue has a chance to run.
|
|
|
|
|
*/
|
|
|
|
|
status = cam_periph_acquire(periph);
|
|
|
|
|
if (status != CAM_REQ_CMP)
|
|
|
|
|
break;
|
|
|
|
|
|
Plumb device physical path reporting from CAM devices, through GEOM and
DEVFS, and make it accessible via the diskinfo utility.
Extend GEOM's generic attribute query mechanism into generic disk consumers.
sys/geom/geom_disk.c:
sys/geom/geom_disk.h:
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Allow disk providers to implement a new method which can override
the default BIO_GETATTR response, d_getattr(struct bio *). This
function returns -1 if not handled, otherwise it returns 0 or an
errno to be passed to g_io_deliver().
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Don't copy the serial number to dp->d_ident anymore, as the CAM XPT
is now responsible for returning this information via
d_getattr()->(a)dagetattr()->xpt_getatr().
sys/geom/geom_dev.c:
- Implement a new ioctl, DIOCGPHYSPATH, which returns the GEOM
attribute "GEOM::physpath", if possible. If the attribute request
returns a zero-length string, ENOENT is returned.
usr.sbin/diskinfo/diskinfo.c:
- If the DIOCGPHYSPATH ioctl is successful, report physical path
data when diskinfo is executed with the '-v' option.
Submitted by: will
Reviewed by: gibbs
Sponsored by: Spectra Logic Corporation
Add generic attribute change notification support to GEOM.
sys/sys/geom/geom.h:
Add a new attrchanged method field to both g_class
and g_geom.
sys/sys/geom/geom.h:
sys/geom/geom_event.c:
- Provide the g_attr_changed() function that providers
can use to advertise attribute changes.
- Perform delivery of attribute change notifications
from a thread context via the standard GEOM event
mechanism.
sys/geom/geom_subr.c:
Inherit the attrchanged method from class to geom (class instance).
sys/geom/geom_disk.c:
Provide disk_attr_changed() to provide g_attr_changed() access
to consumers of the disk API.
sys/cam/scsi/scsi_pass.c:
sys/cam/scsi/scsi_da.c:
sys/geom/geom_dev.c:
sys/geom/geom_disk.c:
Use attribute changed events to track updates to physical path
information.
sys/cam/scsi/scsi_da.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, and
the updated buffer type references our physical path
attribute, emit a GEOM attribute changed event via the
disk_attr_changed() API.
sys/cam/scsi/scsi_pass.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, update
the physical patch devfs alias for this pass instance.
Submitted by: gibbs
Sponsored by: Spectra Logic Corporation
2011-06-14 17:10:32 +00:00
|
|
|
taskqueue_enqueue(taskqueue_thread,
|
|
|
|
|
&softc->add_physpath_task);
|
|
|
|
|
}
|
|
|
|
|
break;
|
|
|
|
|
}
|
1998-09-15 06:36:34 +00:00
|
|
|
default:
|
1999-05-22 22:00:24 +00:00
|
|
|
cam_periph_async(periph, code, path, arg);
|
1998-09-15 06:36:34 +00:00
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static cam_status
|
|
|
|
|
passregister(struct cam_periph *periph, void *arg)
|
|
|
|
|
{
|
|
|
|
|
struct pass_softc *softc;
|
|
|
|
|
struct ccb_getdev *cgd;
|
2011-04-14 21:25:32 +00:00
|
|
|
struct ccb_pathinq cpi;
|
2016-01-07 20:22:55 +00:00
|
|
|
struct make_dev_args args;
|
|
|
|
|
int error, no_tags;
|
1998-09-15 06:36:34 +00:00
|
|
|
|
|
|
|
|
cgd = (struct ccb_getdev *)arg;
|
|
|
|
|
if (cgd == NULL) {
|
Fix several reference counting and object lifetime issues between
the pass(4) and enc(4) drivers and devfs.
The pass(4) driver uses the destroy_dev_sched() routine to
schedule its device node for destruction in a separate thread
context. It does this because the passcleanup() routine can get
called indirectly from the passclose() routine, and that would
cause a deadlock if the close routine tried to destroy its own
device node.
In any case, once a particular passthrough driver number, e.g.
pass3, is destroyed, CAM considers that unit number (3 in this
case) available for reuse.
The problem is that devfs may not be done cleaning up the previous
instance of pass3, and will panic if isn't done cleaning up the
previous instance.
The solution is to get a callback from devfs when the device node
is removed, and make sure we hold a reference to the peripheral
until that happens.
Testing exposed some other cases where we have reference counting
issues, and those were also fixed in the pass(4) driver.
cam_periph.c: In camperiphfree(), reorder some of the operations.
The peripheral destructor needs to be called before
the peripheral is removed from the peripheral is
removed from the list. This is because once we
remove the peripheral from the list, and drop the
topology lock, the peripheral number may be reused.
But if the destructor hasn't been called yet, there
may still be resources hanging around (like devfs
nodes) that haven't been fully cleaned up.
cam_xpt.c: Add an argument to xpt_remove_periph() to indicate
whether the topology lock is already held.
scsi_enc.c: Acquire an extra reference to the peripheral during
registration, and release it once we get a callback
from devfs indicating that the device node is gone.
Call destroy_dev_sched_cb() in enc_oninvalidate()
instead of calling destroy_dev() in the cleanup
routine.
scsi_pass.c: Add reference counting to handle peripheral and
devfs object lifetime issues.
Add a reference to the peripheral and the devfs
node in the peripheral registration.
Don't attempt to add a physical path alias if the
peripheral has been marked invalid.
Release the devfs reference once the initial
physical path alias taskqueue run has completed.
Schedule devfs node destruction in the
passoninvalidate(), and release our peripheral
reference in a new routine, passdevgonecb() once
the devfs node is gone. This allows the peripheral
to fully go away, and the peripheral destructor,
passcleanup(), will get called.
MFC after: 3 days
Sponsored by: Spectra Logic
2012-06-20 17:08:00 +00:00
|
|
|
printf("%s: no getdev CCB, can't register device\n", __func__);
|
1998-09-15 06:36:34 +00:00
|
|
|
return(CAM_REQ_CMP_ERR);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
softc = (struct pass_softc *)malloc(sizeof(*softc),
|
|
|
|
|
M_DEVBUF, M_NOWAIT);
|
|
|
|
|
|
|
|
|
|
if (softc == NULL) {
|
Fix several reference counting and object lifetime issues between
the pass(4) and enc(4) drivers and devfs.
The pass(4) driver uses the destroy_dev_sched() routine to
schedule its device node for destruction in a separate thread
context. It does this because the passcleanup() routine can get
called indirectly from the passclose() routine, and that would
cause a deadlock if the close routine tried to destroy its own
device node.
In any case, once a particular passthrough driver number, e.g.
pass3, is destroyed, CAM considers that unit number (3 in this
case) available for reuse.
The problem is that devfs may not be done cleaning up the previous
instance of pass3, and will panic if isn't done cleaning up the
previous instance.
The solution is to get a callback from devfs when the device node
is removed, and make sure we hold a reference to the peripheral
until that happens.
Testing exposed some other cases where we have reference counting
issues, and those were also fixed in the pass(4) driver.
cam_periph.c: In camperiphfree(), reorder some of the operations.
The peripheral destructor needs to be called before
the peripheral is removed from the peripheral is
removed from the list. This is because once we
remove the peripheral from the list, and drop the
topology lock, the peripheral number may be reused.
But if the destructor hasn't been called yet, there
may still be resources hanging around (like devfs
nodes) that haven't been fully cleaned up.
cam_xpt.c: Add an argument to xpt_remove_periph() to indicate
whether the topology lock is already held.
scsi_enc.c: Acquire an extra reference to the peripheral during
registration, and release it once we get a callback
from devfs indicating that the device node is gone.
Call destroy_dev_sched_cb() in enc_oninvalidate()
instead of calling destroy_dev() in the cleanup
routine.
scsi_pass.c: Add reference counting to handle peripheral and
devfs object lifetime issues.
Add a reference to the peripheral and the devfs
node in the peripheral registration.
Don't attempt to add a physical path alias if the
peripheral has been marked invalid.
Release the devfs reference once the initial
physical path alias taskqueue run has completed.
Schedule devfs node destruction in the
passoninvalidate(), and release our peripheral
reference in a new routine, passdevgonecb() once
the devfs node is gone. This allows the peripheral
to fully go away, and the peripheral destructor,
passcleanup(), will get called.
MFC after: 3 days
Sponsored by: Spectra Logic
2012-06-20 17:08:00 +00:00
|
|
|
printf("%s: Unable to probe new device. "
|
|
|
|
|
"Unable to allocate softc\n", __func__);
|
1998-09-15 06:36:34 +00:00
|
|
|
return(CAM_REQ_CMP_ERR);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
bzero(softc, sizeof(*softc));
|
|
|
|
|
softc->state = PASS_STATE_NORMAL;
|
2011-04-14 21:25:32 +00:00
|
|
|
if (cgd->protocol == PROTO_SCSI || cgd->protocol == PROTO_ATAPI)
|
|
|
|
|
softc->pd_type = SID_TYPE(&cgd->inq_data);
|
|
|
|
|
else if (cgd->protocol == PROTO_SATAPM)
|
|
|
|
|
softc->pd_type = T_ENCLOSURE;
|
|
|
|
|
else
|
|
|
|
|
softc->pd_type = T_DIRECT;
|
1998-09-15 06:36:34 +00:00
|
|
|
|
|
|
|
|
periph->softc = softc;
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
softc->periph = periph;
|
|
|
|
|
TAILQ_INIT(&softc->incoming_queue);
|
|
|
|
|
TAILQ_INIT(&softc->active_queue);
|
|
|
|
|
TAILQ_INIT(&softc->abandoned_queue);
|
|
|
|
|
TAILQ_INIT(&softc->done_queue);
|
|
|
|
|
snprintf(softc->zone_name, sizeof(softc->zone_name), "%s%d",
|
|
|
|
|
periph->periph_name, periph->unit_number);
|
|
|
|
|
snprintf(softc->io_zone_name, sizeof(softc->io_zone_name), "%s%dIO",
|
|
|
|
|
periph->periph_name, periph->unit_number);
|
|
|
|
|
softc->io_zone_size = MAXPHYS;
|
|
|
|
|
knlist_init_mtx(&softc->read_select.si_note, cam_periph_mtx(periph));
|
Rewrite of the CAM error recovery code.
Some of the major changes include:
- The SCSI error handling portion of cam_periph_error() has
been broken out into a number of subfunctions to better
modularize the code that handles the hierarchy of SCSI errors.
As a result, the code is now much easier to read.
- String handling and error printing has been significantly
revamped. We now use sbufs to do string formatting instead
of using printfs (for the kernel) and snprintf/strncat (for
userland) as before.
There is a new catchall error printing routine,
cam_error_print() and its string-based counterpart,
cam_error_string() that allow the kernel and userland
applications to pass in a CCB and have errors printed out
properly, whether or not they're SCSI errors. Among other
things, this helped eliminate a fair amount of duplicate code
in camcontrol.
We now print out more information than before, including
the CAM status and SCSI status and the error recovery action
taken to remedy the problem.
- sbufs are now available in userland, via libsbuf. This
change was necessary since most of the error printing code
is shared between libcam and the kernel.
- A new transfer settings interface is included in this checkin.
This code is #ifdef'ed out, and is primarily intended to aid
discussion with HBA driver authors on the final form the
interface should take. There is example code in the ahc(4)
driver that implements the HBA driver side of the new
interface. The new transfer settings code won't be enabled
until we're ready to switch all HBA drivers over to the new
interface.
src/Makefile.inc1,
lib/Makefile: Add libsbuf. It must be built before libcam,
since libcam uses sbuf routines.
libcam/Makefile: libcam now depends on libsbuf.
libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the
sbuf sources from sys/kern.
bsd.libnames.mk: Add LIBSBUF.
camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically
linked, we can't depend on the dynamic linker
to pull in libsbuf.
camcontrol.c: Use cam_error_print() instead of checking for
CAM_SCSI_STATUS_ERROR on every failed CCB.
sbuf.9: Change the prototypes for sbuf_cat() and
sbuf_cpy() so that the source string is now a
const char *. This is more in line wth the
standard system string functions, and helps
eliminate warnings when dealing with a const
source buffer.
Fix a typo.
cam.c: Add description strings for the various CAM
error status values, as well as routines to
look up those strings.
Add new cam_error_string() and
cam_error_print() routines for userland and
the kernel.
cam.h: Add a new CAM flag, CAM_RETRY_SELTO.
Add enumerated types for the various options
available with cam_error_print() and
cam_error_string().
cam_ccb.h: Add new transfer negotiation structures/types.
Change inq_len in the ccb_getdev structure to
be "reserved". This field has never been
filled in, and will be removed when we next
bump the CAM version.
cam_debug.h: Fix typo.
cam_periph.c: Modularize cam_periph_error(). The SCSI error
handling part of cam_periph_error() is now
in camperiphscsistatuserror() and
camperiphscsisenseerror().
In cam_periph_lock(), increase the reference
count on the periph while we wait for our lock
attempt to succeed so that the periph won't go
away while we're sleeping.
cam_xpt.c: Add new transfer negotiation code. (ifdefed
out)
Add a new function, xpt_path_string(). This
is a string/sbuf analog to xpt_print_path().
scsi_all.c: Revamp string handing and error printing code.
We now use sbufs for much of the string
formatting code. More of that code is shared
between userland the kernel.
scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly
useful in the first place.
Add a new error action, SS_REQSENSE. (Send a
request sense and then retry the command.)
This is useful when the controller hasn't
performed autosense for some reason.
Change the default actions around a bit.
scsi_cd.c,
scsi_da.c,
scsi_pt.c,
scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection
timeouts shouldn't be covered by a sense flag.
scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO.
Get rid of the last vestiges of a read/write
interface.
libkern/bsearch.c,
sys/libkern.h,
conf/files: Add bsearch.c, which is needed for some of the
new table lookup routines.
aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if
CAM_NEW_TRAN_CODE is defined.
sbuf.h,
subr_sbuf.c: Add the appropriate #ifdefs so sbufs can
compile and run in userland.
Change sbuf_printf() to use vsnprintf()
instead of kvprintf(), which is only available
in the kernel.
Change the source string for sbuf_cpy() and
sbuf_cat() to be a const char *.
Add __BEGIN_DECLS and __END_DECLS around
function prototypes since they're now exported
to userland.
kdump/mkioctls: Include stdio.h before cam.h since cam.h now
includes a function with a FILE * argument.
Submitted by: gibbs (mostly)
Reviewed by: jdp, marcel (libsbuf makefile changes)
Reviewed by: des (sbuf changes)
Reviewed by: ken
2001-03-27 05:45:52 +00:00
|
|
|
|
2011-04-14 21:25:32 +00:00
|
|
|
bzero(&cpi, sizeof(cpi));
|
|
|
|
|
xpt_setup_ccb(&cpi.ccb_h, periph->path, CAM_PRIORITY_NORMAL);
|
|
|
|
|
cpi.ccb_h.func_code = XPT_PATH_INQ;
|
|
|
|
|
xpt_action((union ccb *)&cpi);
|
|
|
|
|
|
2015-09-30 13:31:37 +00:00
|
|
|
if (cpi.maxio == 0)
|
|
|
|
|
softc->maxio = DFLTPHYS; /* traditional default */
|
|
|
|
|
else if (cpi.maxio > MAXPHYS)
|
|
|
|
|
softc->maxio = MAXPHYS; /* for safety */
|
|
|
|
|
else
|
|
|
|
|
softc->maxio = cpi.maxio; /* real value */
|
|
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
if (cpi.hba_misc & PIM_UNMAPPED)
|
|
|
|
|
softc->flags |= PASS_FLAG_UNMAPPED_CAPABLE;
|
|
|
|
|
|
1998-09-15 06:36:34 +00:00
|
|
|
/*
|
|
|
|
|
* We pass in 0 for a blocksize, since we don't
|
|
|
|
|
* know what the blocksize of this device is, if
|
|
|
|
|
* it even has a blocksize.
|
|
|
|
|
*/
|
2013-04-04 19:07:37 +00:00
|
|
|
cam_periph_unlock(periph);
|
Rewrite of the CAM error recovery code.
Some of the major changes include:
- The SCSI error handling portion of cam_periph_error() has
been broken out into a number of subfunctions to better
modularize the code that handles the hierarchy of SCSI errors.
As a result, the code is now much easier to read.
- String handling and error printing has been significantly
revamped. We now use sbufs to do string formatting instead
of using printfs (for the kernel) and snprintf/strncat (for
userland) as before.
There is a new catchall error printing routine,
cam_error_print() and its string-based counterpart,
cam_error_string() that allow the kernel and userland
applications to pass in a CCB and have errors printed out
properly, whether or not they're SCSI errors. Among other
things, this helped eliminate a fair amount of duplicate code
in camcontrol.
We now print out more information than before, including
the CAM status and SCSI status and the error recovery action
taken to remedy the problem.
- sbufs are now available in userland, via libsbuf. This
change was necessary since most of the error printing code
is shared between libcam and the kernel.
- A new transfer settings interface is included in this checkin.
This code is #ifdef'ed out, and is primarily intended to aid
discussion with HBA driver authors on the final form the
interface should take. There is example code in the ahc(4)
driver that implements the HBA driver side of the new
interface. The new transfer settings code won't be enabled
until we're ready to switch all HBA drivers over to the new
interface.
src/Makefile.inc1,
lib/Makefile: Add libsbuf. It must be built before libcam,
since libcam uses sbuf routines.
libcam/Makefile: libcam now depends on libsbuf.
libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the
sbuf sources from sys/kern.
bsd.libnames.mk: Add LIBSBUF.
camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically
linked, we can't depend on the dynamic linker
to pull in libsbuf.
camcontrol.c: Use cam_error_print() instead of checking for
CAM_SCSI_STATUS_ERROR on every failed CCB.
sbuf.9: Change the prototypes for sbuf_cat() and
sbuf_cpy() so that the source string is now a
const char *. This is more in line wth the
standard system string functions, and helps
eliminate warnings when dealing with a const
source buffer.
Fix a typo.
cam.c: Add description strings for the various CAM
error status values, as well as routines to
look up those strings.
Add new cam_error_string() and
cam_error_print() routines for userland and
the kernel.
cam.h: Add a new CAM flag, CAM_RETRY_SELTO.
Add enumerated types for the various options
available with cam_error_print() and
cam_error_string().
cam_ccb.h: Add new transfer negotiation structures/types.
Change inq_len in the ccb_getdev structure to
be "reserved". This field has never been
filled in, and will be removed when we next
bump the CAM version.
cam_debug.h: Fix typo.
cam_periph.c: Modularize cam_periph_error(). The SCSI error
handling part of cam_periph_error() is now
in camperiphscsistatuserror() and
camperiphscsisenseerror().
In cam_periph_lock(), increase the reference
count on the periph while we wait for our lock
attempt to succeed so that the periph won't go
away while we're sleeping.
cam_xpt.c: Add new transfer negotiation code. (ifdefed
out)
Add a new function, xpt_path_string(). This
is a string/sbuf analog to xpt_print_path().
scsi_all.c: Revamp string handing and error printing code.
We now use sbufs for much of the string
formatting code. More of that code is shared
between userland the kernel.
scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly
useful in the first place.
Add a new error action, SS_REQSENSE. (Send a
request sense and then retry the command.)
This is useful when the controller hasn't
performed autosense for some reason.
Change the default actions around a bit.
scsi_cd.c,
scsi_da.c,
scsi_pt.c,
scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection
timeouts shouldn't be covered by a sense flag.
scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO.
Get rid of the last vestiges of a read/write
interface.
libkern/bsearch.c,
sys/libkern.h,
conf/files: Add bsearch.c, which is needed for some of the
new table lookup routines.
aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if
CAM_NEW_TRAN_CODE is defined.
sbuf.h,
subr_sbuf.c: Add the appropriate #ifdefs so sbufs can
compile and run in userland.
Change sbuf_printf() to use vsnprintf()
instead of kvprintf(), which is only available
in the kernel.
Change the source string for sbuf_cpy() and
sbuf_cat() to be a const char *.
Add __BEGIN_DECLS and __END_DECLS around
function prototypes since they're now exported
to userland.
kdump/mkioctls: Include stdio.h before cam.h since cam.h now
includes a function with a FILE * argument.
Submitted by: gibbs (mostly)
Reviewed by: jdp, marcel (libsbuf makefile changes)
Reviewed by: des (sbuf changes)
Reviewed by: ken
2001-03-27 05:45:52 +00:00
|
|
|
no_tags = (cgd->inq_data.flags & SID_CmdQue) == 0;
|
2005-01-22 07:21:25 +00:00
|
|
|
softc->device_stats = devstat_new_entry("pass",
|
2008-09-26 14:19:52 +00:00
|
|
|
periph->unit_number, 0,
|
Rewrite of the CAM error recovery code.
Some of the major changes include:
- The SCSI error handling portion of cam_periph_error() has
been broken out into a number of subfunctions to better
modularize the code that handles the hierarchy of SCSI errors.
As a result, the code is now much easier to read.
- String handling and error printing has been significantly
revamped. We now use sbufs to do string formatting instead
of using printfs (for the kernel) and snprintf/strncat (for
userland) as before.
There is a new catchall error printing routine,
cam_error_print() and its string-based counterpart,
cam_error_string() that allow the kernel and userland
applications to pass in a CCB and have errors printed out
properly, whether or not they're SCSI errors. Among other
things, this helped eliminate a fair amount of duplicate code
in camcontrol.
We now print out more information than before, including
the CAM status and SCSI status and the error recovery action
taken to remedy the problem.
- sbufs are now available in userland, via libsbuf. This
change was necessary since most of the error printing code
is shared between libcam and the kernel.
- A new transfer settings interface is included in this checkin.
This code is #ifdef'ed out, and is primarily intended to aid
discussion with HBA driver authors on the final form the
interface should take. There is example code in the ahc(4)
driver that implements the HBA driver side of the new
interface. The new transfer settings code won't be enabled
until we're ready to switch all HBA drivers over to the new
interface.
src/Makefile.inc1,
lib/Makefile: Add libsbuf. It must be built before libcam,
since libcam uses sbuf routines.
libcam/Makefile: libcam now depends on libsbuf.
libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the
sbuf sources from sys/kern.
bsd.libnames.mk: Add LIBSBUF.
camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically
linked, we can't depend on the dynamic linker
to pull in libsbuf.
camcontrol.c: Use cam_error_print() instead of checking for
CAM_SCSI_STATUS_ERROR on every failed CCB.
sbuf.9: Change the prototypes for sbuf_cat() and
sbuf_cpy() so that the source string is now a
const char *. This is more in line wth the
standard system string functions, and helps
eliminate warnings when dealing with a const
source buffer.
Fix a typo.
cam.c: Add description strings for the various CAM
error status values, as well as routines to
look up those strings.
Add new cam_error_string() and
cam_error_print() routines for userland and
the kernel.
cam.h: Add a new CAM flag, CAM_RETRY_SELTO.
Add enumerated types for the various options
available with cam_error_print() and
cam_error_string().
cam_ccb.h: Add new transfer negotiation structures/types.
Change inq_len in the ccb_getdev structure to
be "reserved". This field has never been
filled in, and will be removed when we next
bump the CAM version.
cam_debug.h: Fix typo.
cam_periph.c: Modularize cam_periph_error(). The SCSI error
handling part of cam_periph_error() is now
in camperiphscsistatuserror() and
camperiphscsisenseerror().
In cam_periph_lock(), increase the reference
count on the periph while we wait for our lock
attempt to succeed so that the periph won't go
away while we're sleeping.
cam_xpt.c: Add new transfer negotiation code. (ifdefed
out)
Add a new function, xpt_path_string(). This
is a string/sbuf analog to xpt_print_path().
scsi_all.c: Revamp string handing and error printing code.
We now use sbufs for much of the string
formatting code. More of that code is shared
between userland the kernel.
scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly
useful in the first place.
Add a new error action, SS_REQSENSE. (Send a
request sense and then retry the command.)
This is useful when the controller hasn't
performed autosense for some reason.
Change the default actions around a bit.
scsi_cd.c,
scsi_da.c,
scsi_pt.c,
scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection
timeouts shouldn't be covered by a sense flag.
scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO.
Get rid of the last vestiges of a read/write
interface.
libkern/bsearch.c,
sys/libkern.h,
conf/files: Add bsearch.c, which is needed for some of the
new table lookup routines.
aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if
CAM_NEW_TRAN_CODE is defined.
sbuf.h,
subr_sbuf.c: Add the appropriate #ifdefs so sbufs can
compile and run in userland.
Change sbuf_printf() to use vsnprintf()
instead of kvprintf(), which is only available
in the kernel.
Change the source string for sbuf_cpy() and
sbuf_cat() to be a const char *.
Add __BEGIN_DECLS and __END_DECLS around
function prototypes since they're now exported
to userland.
kdump/mkioctls: Include stdio.h before cam.h since cam.h now
includes a function with a FILE * argument.
Submitted by: gibbs (mostly)
Reviewed by: jdp, marcel (libsbuf makefile changes)
Reviewed by: des (sbuf changes)
Reviewed by: ken
2001-03-27 05:45:52 +00:00
|
|
|
DEVSTAT_NO_BLOCKSIZE
|
|
|
|
|
| (no_tags ? DEVSTAT_NO_ORDERED_TAGS : 0),
|
2000-01-17 06:27:37 +00:00
|
|
|
softc->pd_type |
|
2011-04-14 21:25:32 +00:00
|
|
|
XPORT_DEVSTAT_TYPE(cpi.transport) |
|
Add a prioritization field to the devstat_add_entry() call so that
peripheral drivers can determine where in the devstat(9) list they are
inserted.
This requires recompilation of libdevstat, systat, vmstat, rpc.rstatd, and
any ports that depend on the devstat code, since the size of the devstat
structure has changed. The devstat version number has been incremented as
well to reflect the change.
This sorts devices in the devstat list in "more interesting" to "less
interesting" order. So, for instance, da devices are now more important
than floppy drives, and so will appear before floppy drives in the default
output from systat, iostat, vmstat, etc.
The order of devices is, for now, kept in a central table in devicestat.h.
If individual drivers were able to make a meaningful decision on what
priority they should be at attach time, we could consider splitting the
priority information out into the various drivers. For now, though, they
have no way of knowing that, so it's easier to put them in an easy to find
table.
Also, move the checkversion() call in vmstat(8) to a more logical place.
Thanks to Bruce and David O'Brien for suggestions, for reviewing this, and
for putting up with the long time it has taken me to commit it. Bruce did
object somewhat to the central priority table (he would rather the
priorities be distributed in each driver), so his objection is duly noted
here.
Reviewed by: bde, obrien
1999-02-10 00:04:13 +00:00
|
|
|
DEVSTAT_TYPE_PASS,
|
|
|
|
|
DEVSTAT_PRIORITY_PASS);
|
1999-11-17 04:59:09 +00:00
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
/*
|
|
|
|
|
* Initialize the taskqueue handler for shutting down kqueue.
|
|
|
|
|
*/
|
|
|
|
|
TASK_INIT(&softc->shutdown_kqueue_task, /*priority*/ 0,
|
|
|
|
|
pass_shutdown_kqueue, periph);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Acquire a reference to the periph that we can release once we've
|
|
|
|
|
* cleaned up the kqueue.
|
|
|
|
|
*/
|
|
|
|
|
if (cam_periph_acquire(periph) != CAM_REQ_CMP) {
|
|
|
|
|
xpt_print(periph->path, "%s: lost periph during "
|
|
|
|
|
"registration!\n", __func__);
|
|
|
|
|
cam_periph_lock(periph);
|
|
|
|
|
return (CAM_REQ_CMP_ERR);
|
|
|
|
|
}
|
|
|
|
|
|
Fix several reference counting and object lifetime issues between
the pass(4) and enc(4) drivers and devfs.
The pass(4) driver uses the destroy_dev_sched() routine to
schedule its device node for destruction in a separate thread
context. It does this because the passcleanup() routine can get
called indirectly from the passclose() routine, and that would
cause a deadlock if the close routine tried to destroy its own
device node.
In any case, once a particular passthrough driver number, e.g.
pass3, is destroyed, CAM considers that unit number (3 in this
case) available for reuse.
The problem is that devfs may not be done cleaning up the previous
instance of pass3, and will panic if isn't done cleaning up the
previous instance.
The solution is to get a callback from devfs when the device node
is removed, and make sure we hold a reference to the peripheral
until that happens.
Testing exposed some other cases where we have reference counting
issues, and those were also fixed in the pass(4) driver.
cam_periph.c: In camperiphfree(), reorder some of the operations.
The peripheral destructor needs to be called before
the peripheral is removed from the peripheral is
removed from the list. This is because once we
remove the peripheral from the list, and drop the
topology lock, the peripheral number may be reused.
But if the destructor hasn't been called yet, there
may still be resources hanging around (like devfs
nodes) that haven't been fully cleaned up.
cam_xpt.c: Add an argument to xpt_remove_periph() to indicate
whether the topology lock is already held.
scsi_enc.c: Acquire an extra reference to the peripheral during
registration, and release it once we get a callback
from devfs indicating that the device node is gone.
Call destroy_dev_sched_cb() in enc_oninvalidate()
instead of calling destroy_dev() in the cleanup
routine.
scsi_pass.c: Add reference counting to handle peripheral and
devfs object lifetime issues.
Add a reference to the peripheral and the devfs
node in the peripheral registration.
Don't attempt to add a physical path alias if the
peripheral has been marked invalid.
Release the devfs reference once the initial
physical path alias taskqueue run has completed.
Schedule devfs node destruction in the
passoninvalidate(), and release our peripheral
reference in a new routine, passdevgonecb() once
the devfs node is gone. This allows the peripheral
to fully go away, and the peripheral destructor,
passcleanup(), will get called.
MFC after: 3 days
Sponsored by: Spectra Logic
2012-06-20 17:08:00 +00:00
|
|
|
/*
|
|
|
|
|
* Acquire a reference to the periph before we create the devfs
|
|
|
|
|
* instance for it. We'll release this reference once the devfs
|
|
|
|
|
* instance has been freed.
|
|
|
|
|
*/
|
|
|
|
|
if (cam_periph_acquire(periph) != CAM_REQ_CMP) {
|
|
|
|
|
xpt_print(periph->path, "%s: lost periph during "
|
|
|
|
|
"registration!\n", __func__);
|
Fix a device departure bug for the the pass(4), enc(4), sg(4) and ch(4)
drivers.
The bug occurrs when a userland process has the driver instance
open and the underlying device goes away. We get the devfs
callback that the device node has been destroyed, but not all of
the closes necessary to fully decrement the reference count on the
CAM peripheral.
The reason is that once devfs calls back and says the device has
been destroyed, it is moved off to deadfs, and devfs guarantees
that there will be no more open or close calls. So the solution
is to keep track of how many outstanding open calls there are on
the device, and just release that many references when we get the
callback from devfs.
scsi_pass.c,
scsi_enc.c,
scsi_enc_internal.h: Add an open count to the softc in these
drivers. Increment it on open and
decrement it on close.
When we get a devfs callback to say that
the device node has gone away, decrement
the peripheral reference count by the
number of still outstanding opens.
Make sure we don't access the peripheral
with cam_periph_unlock() after what might
be the final call to
cam_periph_release_locked(). The
peripheral might have been freed, and we
will be dereferencing freed memory.
scsi_ch.c,
scsi_sg.c: For the ch(4) and sg(4) drivers, add the
same changes described above, and in
addition, fix another bug that was
previously fixed in the pass(4) and enc(4)
drivers.
These drivers were calling destroy_dev()
from their cleanup routine, but that could
cause a deadlock because the cleanup
routine could be indirectly called from
the driver's close routine. This would
cause a deadlock, because the device node
is being held open by the active close
call, and can't be destroyed.
Sponsored by: Spectra Logic Corporation
MFC after: 1 week
2012-12-08 04:03:04 +00:00
|
|
|
cam_periph_lock(periph);
|
Fix several reference counting and object lifetime issues between
the pass(4) and enc(4) drivers and devfs.
The pass(4) driver uses the destroy_dev_sched() routine to
schedule its device node for destruction in a separate thread
context. It does this because the passcleanup() routine can get
called indirectly from the passclose() routine, and that would
cause a deadlock if the close routine tried to destroy its own
device node.
In any case, once a particular passthrough driver number, e.g.
pass3, is destroyed, CAM considers that unit number (3 in this
case) available for reuse.
The problem is that devfs may not be done cleaning up the previous
instance of pass3, and will panic if isn't done cleaning up the
previous instance.
The solution is to get a callback from devfs when the device node
is removed, and make sure we hold a reference to the peripheral
until that happens.
Testing exposed some other cases where we have reference counting
issues, and those were also fixed in the pass(4) driver.
cam_periph.c: In camperiphfree(), reorder some of the operations.
The peripheral destructor needs to be called before
the peripheral is removed from the peripheral is
removed from the list. This is because once we
remove the peripheral from the list, and drop the
topology lock, the peripheral number may be reused.
But if the destructor hasn't been called yet, there
may still be resources hanging around (like devfs
nodes) that haven't been fully cleaned up.
cam_xpt.c: Add an argument to xpt_remove_periph() to indicate
whether the topology lock is already held.
scsi_enc.c: Acquire an extra reference to the peripheral during
registration, and release it once we get a callback
from devfs indicating that the device node is gone.
Call destroy_dev_sched_cb() in enc_oninvalidate()
instead of calling destroy_dev() in the cleanup
routine.
scsi_pass.c: Add reference counting to handle peripheral and
devfs object lifetime issues.
Add a reference to the peripheral and the devfs
node in the peripheral registration.
Don't attempt to add a physical path alias if the
peripheral has been marked invalid.
Release the devfs reference once the initial
physical path alias taskqueue run has completed.
Schedule devfs node destruction in the
passoninvalidate(), and release our peripheral
reference in a new routine, passdevgonecb() once
the devfs node is gone. This allows the peripheral
to fully go away, and the peripheral destructor,
passcleanup(), will get called.
MFC after: 3 days
Sponsored by: Spectra Logic
2012-06-20 17:08:00 +00:00
|
|
|
return (CAM_REQ_CMP_ERR);
|
|
|
|
|
}
|
|
|
|
|
|
1999-11-17 04:59:09 +00:00
|
|
|
/* Register the device */
|
2016-01-07 20:22:55 +00:00
|
|
|
make_dev_args_init(&args);
|
|
|
|
|
args.mda_devsw = &pass_cdevsw;
|
|
|
|
|
args.mda_unit = periph->unit_number;
|
|
|
|
|
args.mda_uid = UID_ROOT;
|
|
|
|
|
args.mda_gid = GID_OPERATOR;
|
|
|
|
|
args.mda_mode = 0600;
|
|
|
|
|
args.mda_si_drv1 = periph;
|
|
|
|
|
error = make_dev_s(&args, &softc->dev, "%s%d", periph->periph_name,
|
|
|
|
|
periph->unit_number);
|
|
|
|
|
if (error != 0) {
|
|
|
|
|
cam_periph_lock(periph);
|
|
|
|
|
cam_periph_release_locked(periph);
|
|
|
|
|
return (CAM_REQ_CMP_ERR);
|
|
|
|
|
}
|
Fix several reference counting and object lifetime issues between
the pass(4) and enc(4) drivers and devfs.
The pass(4) driver uses the destroy_dev_sched() routine to
schedule its device node for destruction in a separate thread
context. It does this because the passcleanup() routine can get
called indirectly from the passclose() routine, and that would
cause a deadlock if the close routine tried to destroy its own
device node.
In any case, once a particular passthrough driver number, e.g.
pass3, is destroyed, CAM considers that unit number (3 in this
case) available for reuse.
The problem is that devfs may not be done cleaning up the previous
instance of pass3, and will panic if isn't done cleaning up the
previous instance.
The solution is to get a callback from devfs when the device node
is removed, and make sure we hold a reference to the peripheral
until that happens.
Testing exposed some other cases where we have reference counting
issues, and those were also fixed in the pass(4) driver.
cam_periph.c: In camperiphfree(), reorder some of the operations.
The peripheral destructor needs to be called before
the peripheral is removed from the peripheral is
removed from the list. This is because once we
remove the peripheral from the list, and drop the
topology lock, the peripheral number may be reused.
But if the destructor hasn't been called yet, there
may still be resources hanging around (like devfs
nodes) that haven't been fully cleaned up.
cam_xpt.c: Add an argument to xpt_remove_periph() to indicate
whether the topology lock is already held.
scsi_enc.c: Acquire an extra reference to the peripheral during
registration, and release it once we get a callback
from devfs indicating that the device node is gone.
Call destroy_dev_sched_cb() in enc_oninvalidate()
instead of calling destroy_dev() in the cleanup
routine.
scsi_pass.c: Add reference counting to handle peripheral and
devfs object lifetime issues.
Add a reference to the peripheral and the devfs
node in the peripheral registration.
Don't attempt to add a physical path alias if the
peripheral has been marked invalid.
Release the devfs reference once the initial
physical path alias taskqueue run has completed.
Schedule devfs node destruction in the
passoninvalidate(), and release our peripheral
reference in a new routine, passdevgonecb() once
the devfs node is gone. This allows the peripheral
to fully go away, and the peripheral destructor,
passcleanup(), will get called.
MFC after: 3 days
Sponsored by: Spectra Logic
2012-06-20 17:08:00 +00:00
|
|
|
|
|
|
|
|
/*
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
* Hold a reference to the periph before we create the physical
|
|
|
|
|
* path alias so it can't go away.
|
Fix several reference counting and object lifetime issues between
the pass(4) and enc(4) drivers and devfs.
The pass(4) driver uses the destroy_dev_sched() routine to
schedule its device node for destruction in a separate thread
context. It does this because the passcleanup() routine can get
called indirectly from the passclose() routine, and that would
cause a deadlock if the close routine tried to destroy its own
device node.
In any case, once a particular passthrough driver number, e.g.
pass3, is destroyed, CAM considers that unit number (3 in this
case) available for reuse.
The problem is that devfs may not be done cleaning up the previous
instance of pass3, and will panic if isn't done cleaning up the
previous instance.
The solution is to get a callback from devfs when the device node
is removed, and make sure we hold a reference to the peripheral
until that happens.
Testing exposed some other cases where we have reference counting
issues, and those were also fixed in the pass(4) driver.
cam_periph.c: In camperiphfree(), reorder some of the operations.
The peripheral destructor needs to be called before
the peripheral is removed from the peripheral is
removed from the list. This is because once we
remove the peripheral from the list, and drop the
topology lock, the peripheral number may be reused.
But if the destructor hasn't been called yet, there
may still be resources hanging around (like devfs
nodes) that haven't been fully cleaned up.
cam_xpt.c: Add an argument to xpt_remove_periph() to indicate
whether the topology lock is already held.
scsi_enc.c: Acquire an extra reference to the peripheral during
registration, and release it once we get a callback
from devfs indicating that the device node is gone.
Call destroy_dev_sched_cb() in enc_oninvalidate()
instead of calling destroy_dev() in the cleanup
routine.
scsi_pass.c: Add reference counting to handle peripheral and
devfs object lifetime issues.
Add a reference to the peripheral and the devfs
node in the peripheral registration.
Don't attempt to add a physical path alias if the
peripheral has been marked invalid.
Release the devfs reference once the initial
physical path alias taskqueue run has completed.
Schedule devfs node destruction in the
passoninvalidate(), and release our peripheral
reference in a new routine, passdevgonecb() once
the devfs node is gone. This allows the peripheral
to fully go away, and the peripheral destructor,
passcleanup(), will get called.
MFC after: 3 days
Sponsored by: Spectra Logic
2012-06-20 17:08:00 +00:00
|
|
|
*/
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
if (cam_periph_acquire(periph) != CAM_REQ_CMP) {
|
|
|
|
|
xpt_print(periph->path, "%s: lost periph during "
|
|
|
|
|
"registration!\n", __func__);
|
|
|
|
|
cam_periph_lock(periph);
|
|
|
|
|
return (CAM_REQ_CMP_ERR);
|
|
|
|
|
}
|
Fix several reference counting and object lifetime issues between
the pass(4) and enc(4) drivers and devfs.
The pass(4) driver uses the destroy_dev_sched() routine to
schedule its device node for destruction in a separate thread
context. It does this because the passcleanup() routine can get
called indirectly from the passclose() routine, and that would
cause a deadlock if the close routine tried to destroy its own
device node.
In any case, once a particular passthrough driver number, e.g.
pass3, is destroyed, CAM considers that unit number (3 in this
case) available for reuse.
The problem is that devfs may not be done cleaning up the previous
instance of pass3, and will panic if isn't done cleaning up the
previous instance.
The solution is to get a callback from devfs when the device node
is removed, and make sure we hold a reference to the peripheral
until that happens.
Testing exposed some other cases where we have reference counting
issues, and those were also fixed in the pass(4) driver.
cam_periph.c: In camperiphfree(), reorder some of the operations.
The peripheral destructor needs to be called before
the peripheral is removed from the peripheral is
removed from the list. This is because once we
remove the peripheral from the list, and drop the
topology lock, the peripheral number may be reused.
But if the destructor hasn't been called yet, there
may still be resources hanging around (like devfs
nodes) that haven't been fully cleaned up.
cam_xpt.c: Add an argument to xpt_remove_periph() to indicate
whether the topology lock is already held.
scsi_enc.c: Acquire an extra reference to the peripheral during
registration, and release it once we get a callback
from devfs indicating that the device node is gone.
Call destroy_dev_sched_cb() in enc_oninvalidate()
instead of calling destroy_dev() in the cleanup
routine.
scsi_pass.c: Add reference counting to handle peripheral and
devfs object lifetime issues.
Add a reference to the peripheral and the devfs
node in the peripheral registration.
Don't attempt to add a physical path alias if the
peripheral has been marked invalid.
Release the devfs reference once the initial
physical path alias taskqueue run has completed.
Schedule devfs node destruction in the
passoninvalidate(), and release our peripheral
reference in a new routine, passdevgonecb() once
the devfs node is gone. This allows the peripheral
to fully go away, and the peripheral destructor,
passcleanup(), will get called.
MFC after: 3 days
Sponsored by: Spectra Logic
2012-06-20 17:08:00 +00:00
|
|
|
|
2013-04-04 19:07:37 +00:00
|
|
|
cam_periph_lock(periph);
|
1999-11-17 04:59:09 +00:00
|
|
|
|
Plumb device physical path reporting from CAM devices, through GEOM and
DEVFS, and make it accessible via the diskinfo utility.
Extend GEOM's generic attribute query mechanism into generic disk consumers.
sys/geom/geom_disk.c:
sys/geom/geom_disk.h:
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Allow disk providers to implement a new method which can override
the default BIO_GETATTR response, d_getattr(struct bio *). This
function returns -1 if not handled, otherwise it returns 0 or an
errno to be passed to g_io_deliver().
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Don't copy the serial number to dp->d_ident anymore, as the CAM XPT
is now responsible for returning this information via
d_getattr()->(a)dagetattr()->xpt_getatr().
sys/geom/geom_dev.c:
- Implement a new ioctl, DIOCGPHYSPATH, which returns the GEOM
attribute "GEOM::physpath", if possible. If the attribute request
returns a zero-length string, ENOENT is returned.
usr.sbin/diskinfo/diskinfo.c:
- If the DIOCGPHYSPATH ioctl is successful, report physical path
data when diskinfo is executed with the '-v' option.
Submitted by: will
Reviewed by: gibbs
Sponsored by: Spectra Logic Corporation
Add generic attribute change notification support to GEOM.
sys/sys/geom/geom.h:
Add a new attrchanged method field to both g_class
and g_geom.
sys/sys/geom/geom.h:
sys/geom/geom_event.c:
- Provide the g_attr_changed() function that providers
can use to advertise attribute changes.
- Perform delivery of attribute change notifications
from a thread context via the standard GEOM event
mechanism.
sys/geom/geom_subr.c:
Inherit the attrchanged method from class to geom (class instance).
sys/geom/geom_disk.c:
Provide disk_attr_changed() to provide g_attr_changed() access
to consumers of the disk API.
sys/cam/scsi/scsi_pass.c:
sys/cam/scsi/scsi_da.c:
sys/geom/geom_dev.c:
sys/geom/geom_disk.c:
Use attribute changed events to track updates to physical path
information.
sys/cam/scsi/scsi_da.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, and
the updated buffer type references our physical path
attribute, emit a GEOM attribute changed event via the
disk_attr_changed() API.
sys/cam/scsi/scsi_pass.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, update
the physical patch devfs alias for this pass instance.
Submitted by: gibbs
Sponsored by: Spectra Logic Corporation
2011-06-14 17:10:32 +00:00
|
|
|
TASK_INIT(&softc->add_physpath_task, /*priority*/0,
|
|
|
|
|
pass_add_physpath, periph);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* See if physical path information is already available.
|
|
|
|
|
*/
|
|
|
|
|
taskqueue_enqueue(taskqueue_thread, &softc->add_physpath_task);
|
|
|
|
|
|
1998-09-15 06:36:34 +00:00
|
|
|
/*
|
Plumb device physical path reporting from CAM devices, through GEOM and
DEVFS, and make it accessible via the diskinfo utility.
Extend GEOM's generic attribute query mechanism into generic disk consumers.
sys/geom/geom_disk.c:
sys/geom/geom_disk.h:
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Allow disk providers to implement a new method which can override
the default BIO_GETATTR response, d_getattr(struct bio *). This
function returns -1 if not handled, otherwise it returns 0 or an
errno to be passed to g_io_deliver().
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Don't copy the serial number to dp->d_ident anymore, as the CAM XPT
is now responsible for returning this information via
d_getattr()->(a)dagetattr()->xpt_getatr().
sys/geom/geom_dev.c:
- Implement a new ioctl, DIOCGPHYSPATH, which returns the GEOM
attribute "GEOM::physpath", if possible. If the attribute request
returns a zero-length string, ENOENT is returned.
usr.sbin/diskinfo/diskinfo.c:
- If the DIOCGPHYSPATH ioctl is successful, report physical path
data when diskinfo is executed with the '-v' option.
Submitted by: will
Reviewed by: gibbs
Sponsored by: Spectra Logic Corporation
Add generic attribute change notification support to GEOM.
sys/sys/geom/geom.h:
Add a new attrchanged method field to both g_class
and g_geom.
sys/sys/geom/geom.h:
sys/geom/geom_event.c:
- Provide the g_attr_changed() function that providers
can use to advertise attribute changes.
- Perform delivery of attribute change notifications
from a thread context via the standard GEOM event
mechanism.
sys/geom/geom_subr.c:
Inherit the attrchanged method from class to geom (class instance).
sys/geom/geom_disk.c:
Provide disk_attr_changed() to provide g_attr_changed() access
to consumers of the disk API.
sys/cam/scsi/scsi_pass.c:
sys/cam/scsi/scsi_da.c:
sys/geom/geom_dev.c:
sys/geom/geom_disk.c:
Use attribute changed events to track updates to physical path
information.
sys/cam/scsi/scsi_da.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, and
the updated buffer type references our physical path
attribute, emit a GEOM attribute changed event via the
disk_attr_changed() API.
sys/cam/scsi/scsi_pass.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, update
the physical patch devfs alias for this pass instance.
Submitted by: gibbs
Sponsored by: Spectra Logic Corporation
2011-06-14 17:10:32 +00:00
|
|
|
* Add an async callback so that we get notified if
|
|
|
|
|
* this device goes away or its physical path
|
|
|
|
|
* (stored in the advanced info data of the EDT) has
|
|
|
|
|
* changed.
|
1998-09-15 06:36:34 +00:00
|
|
|
*/
|
Plumb device physical path reporting from CAM devices, through GEOM and
DEVFS, and make it accessible via the diskinfo utility.
Extend GEOM's generic attribute query mechanism into generic disk consumers.
sys/geom/geom_disk.c:
sys/geom/geom_disk.h:
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Allow disk providers to implement a new method which can override
the default BIO_GETATTR response, d_getattr(struct bio *). This
function returns -1 if not handled, otherwise it returns 0 or an
errno to be passed to g_io_deliver().
sys/cam/scsi/scsi_da.c:
sys/cam/ata/ata_da.c:
- Don't copy the serial number to dp->d_ident anymore, as the CAM XPT
is now responsible for returning this information via
d_getattr()->(a)dagetattr()->xpt_getatr().
sys/geom/geom_dev.c:
- Implement a new ioctl, DIOCGPHYSPATH, which returns the GEOM
attribute "GEOM::physpath", if possible. If the attribute request
returns a zero-length string, ENOENT is returned.
usr.sbin/diskinfo/diskinfo.c:
- If the DIOCGPHYSPATH ioctl is successful, report physical path
data when diskinfo is executed with the '-v' option.
Submitted by: will
Reviewed by: gibbs
Sponsored by: Spectra Logic Corporation
Add generic attribute change notification support to GEOM.
sys/sys/geom/geom.h:
Add a new attrchanged method field to both g_class
and g_geom.
sys/sys/geom/geom.h:
sys/geom/geom_event.c:
- Provide the g_attr_changed() function that providers
can use to advertise attribute changes.
- Perform delivery of attribute change notifications
from a thread context via the standard GEOM event
mechanism.
sys/geom/geom_subr.c:
Inherit the attrchanged method from class to geom (class instance).
sys/geom/geom_disk.c:
Provide disk_attr_changed() to provide g_attr_changed() access
to consumers of the disk API.
sys/cam/scsi/scsi_pass.c:
sys/cam/scsi/scsi_da.c:
sys/geom/geom_dev.c:
sys/geom/geom_disk.c:
Use attribute changed events to track updates to physical path
information.
sys/cam/scsi/scsi_da.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, and
the updated buffer type references our physical path
attribute, emit a GEOM attribute changed event via the
disk_attr_changed() API.
sys/cam/scsi/scsi_pass.c:
Add AC_ADVINFO_CHANGED to the registered asynchronous CAM
events for this driver. When this event occurs, update
the physical patch devfs alias for this pass instance.
Submitted by: gibbs
Sponsored by: Spectra Logic Corporation
2011-06-14 17:10:32 +00:00
|
|
|
xpt_register_async(AC_LOST_DEVICE | AC_ADVINFO_CHANGED,
|
|
|
|
|
passasync, periph, periph->path);
|
1998-09-15 06:36:34 +00:00
|
|
|
|
|
|
|
|
if (bootverbose)
|
|
|
|
|
xpt_announce_periph(periph, NULL);
|
|
|
|
|
|
|
|
|
|
return(CAM_REQ_CMP);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int
|
2004-06-16 09:47:26 +00:00
|
|
|
passopen(struct cdev *dev, int flags, int fmt, struct thread *td)
|
1998-09-15 06:36:34 +00:00
|
|
|
{
|
|
|
|
|
struct cam_periph *periph;
|
|
|
|
|
struct pass_softc *softc;
|
2002-08-15 20:54:03 +00:00
|
|
|
int error;
|
1998-09-15 06:36:34 +00:00
|
|
|
|
2002-08-15 20:54:03 +00:00
|
|
|
periph = (struct cam_periph *)dev->si_drv1;
|
2007-04-15 08:49:19 +00:00
|
|
|
if (cam_periph_acquire(periph) != CAM_REQ_CMP)
|
1998-09-15 06:36:34 +00:00
|
|
|
return (ENXIO);
|
|
|
|
|
|
2007-04-15 08:49:19 +00:00
|
|
|
cam_periph_lock(periph);
|
|
|
|
|
|
1998-09-15 06:36:34 +00:00
|
|
|
softc = (struct pass_softc *)periph->softc;
|
|
|
|
|
|
Fix a problem with the way we handled device invalidation when attaching
to a device failed.
In theory, the same steps that happen when we get an AC_LOST_DEVICE async
notification should have been taken when a driver fails to attach. In
practice, that wasn't the case.
This only affected the da, cd and ch drivers, but the fix affects all
peripheral drivers.
There were several possible problems:
- In the da driver, we didn't remove the peripheral's softc from the da
driver's linked list of softcs. Once the peripheral and softc got
removed, we'd get a kernel panic the next time the timeout routine
called dasendorderedtag().
- In the da, cd and possibly ch drivers, we didn't remove the
peripheral's devstat structure from the devstat queue. Once the
peripheral and softc were removed, this could cause a panic if anyone
tried to access device statistics. (one component of the linked list
wouldn't exist anymore)
- In the cd driver, we didn't take the peripheral off the changer run
queue if it was scheduled to run. In practice, it's highly unlikely,
and maybe impossible that the peripheral would have been on the
changer run queue at that stage of the probe process.
The fix is:
- Add a new peripheral callback function (the "oninvalidate" function)
that is called the first time cam_periph_invalidate() is called for a
peripheral.
- Create new foooninvalidate() routines for each peripheral driver. This
routine is always called at splsoftcam(), and contains all the stuff
that used to be in the AC_LOST_DEVICE case of the async callback
handler.
- Move the devstat cleanup call to the destructor/cleanup routines, since
some of the drivers do I/O in their close routines.
- Make sure that when we're flushing the buffer queue, we traverse it at
splbio().
- Add a check for the invalid flag in the pt driver's open routine.
Reviewed by: gibbs
1998-10-22 22:16:56 +00:00
|
|
|
if (softc->flags & PASS_FLAG_INVALID) {
|
Work around a race condition in devfs by changing the way closes
are handled in most CAM peripheral drivers that are not handled by
GEOM's disk class.
The usual character driver open and close semantics are that the
driver gets N open calls, but only one close, when the last caller
closes the device.
CAM peripheral drivers expect that behavior to be honored to the
letter, and the CAM peripheral driver code (specifically
cam_periph_release_locked_busses()) panics if it is done incorrectly.
Since devfs has to drop its locks while it calls a driver's close
routine, and it does not have a way to delay or prevent open calls
while it is calling the close routine, there is a race.
The sequence of events, simplified a bit, is:
- devfs acquires a lock
- devfs checks the reference count, and if it is 1, continues to close.
- devfs releases the lock
- 2nd process open call on the device happens here
- devfs calls the driver's close routine
- devfs acquires a lock
- devfs decrements the reference count
- devfs releases the lock
- 2nd process close call on the device happens here
At the second close, we get a panic in
cam_periph_release_locked_busses(), complaining that peripheral
has been released when the reference count is already 0. This is
because we have gotten two closes in a row, which should not
happen.
The fix is to add the D_TRACKCLOSE flag to the driver's cdevsw, so
that we get a close() call for each open(). That does happen
reliably, so we can make sure that our reference counts are
correct.
Note that the sa(4) and pt(4) drivers only allow one context
through the open routine. So these drivers aren't exposed to the
same race condition.
scsi_ch.c,
scsi_enc.c,
scsi_enc_internal.h,
scsi_pass.c,
scsi_sg.c:
For these drivers, change the open() routine to
increment the reference count for every open, and
just decrement the reference count in the close.
Call cam_periph_release_locked() in some scenarios
to avoid additional lock and unlock calls.
scsi_pt.c: Call cam_periph_release_locked() in some scenarios
to avoid additional lock and unlock calls.
MFC after: 3 days
2012-05-27 06:11:09 +00:00
|
|
|
cam_periph_release_locked(periph);
|
2007-04-15 08:49:19 +00:00
|
|
|
cam_periph_unlock(periph);
|
1998-09-15 06:36:34 +00:00
|
|
|
return(ENXIO);
|
Fix a problem with the way we handled device invalidation when attaching
to a device failed.
In theory, the same steps that happen when we get an AC_LOST_DEVICE async
notification should have been taken when a driver fails to attach. In
practice, that wasn't the case.
This only affected the da, cd and ch drivers, but the fix affects all
peripheral drivers.
There were several possible problems:
- In the da driver, we didn't remove the peripheral's softc from the da
driver's linked list of softcs. Once the peripheral and softc got
removed, we'd get a kernel panic the next time the timeout routine
called dasendorderedtag().
- In the da, cd and possibly ch drivers, we didn't remove the
peripheral's devstat structure from the devstat queue. Once the
peripheral and softc were removed, this could cause a panic if anyone
tried to access device statistics. (one component of the linked list
wouldn't exist anymore)
- In the cd driver, we didn't take the peripheral off the changer run
queue if it was scheduled to run. In practice, it's highly unlikely,
and maybe impossible that the peripheral would have been on the
changer run queue at that stage of the probe process.
The fix is:
- Add a new peripheral callback function (the "oninvalidate" function)
that is called the first time cam_periph_invalidate() is called for a
peripheral.
- Create new foooninvalidate() routines for each peripheral driver. This
routine is always called at splsoftcam(), and contains all the stuff
that used to be in the AC_LOST_DEVICE case of the async callback
handler.
- Move the devstat cleanup call to the destructor/cleanup routines, since
some of the drivers do I/O in their close routines.
- Make sure that when we're flushing the buffer queue, we traverse it at
splbio().
- Add a check for the invalid flag in the pt driver's open routine.
Reviewed by: gibbs
1998-10-22 22:16:56 +00:00
|
|
|
}
|
1998-11-22 23:44:47 +00:00
|
|
|
|
|
|
|
|
/*
|
2001-09-25 02:15:00 +00:00
|
|
|
* Don't allow access when we're running at a high securelevel.
|
1998-11-22 23:44:47 +00:00
|
|
|
*/
|
2002-02-27 18:32:23 +00:00
|
|
|
error = securelevel_gt(td->td_ucred, 1);
|
2001-09-26 20:13:16 +00:00
|
|
|
if (error) {
|
Work around a race condition in devfs by changing the way closes
are handled in most CAM peripheral drivers that are not handled by
GEOM's disk class.
The usual character driver open and close semantics are that the
driver gets N open calls, but only one close, when the last caller
closes the device.
CAM peripheral drivers expect that behavior to be honored to the
letter, and the CAM peripheral driver code (specifically
cam_periph_release_locked_busses()) panics if it is done incorrectly.
Since devfs has to drop its locks while it calls a driver's close
routine, and it does not have a way to delay or prevent open calls
while it is calling the close routine, there is a race.
The sequence of events, simplified a bit, is:
- devfs acquires a lock
- devfs checks the reference count, and if it is 1, continues to close.
- devfs releases the lock
- 2nd process open call on the device happens here
- devfs calls the driver's close routine
- devfs acquires a lock
- devfs decrements the reference count
- devfs releases the lock
- 2nd process close call on the device happens here
At the second close, we get a panic in
cam_periph_release_locked_busses(), complaining that peripheral
has been released when the reference count is already 0. This is
because we have gotten two closes in a row, which should not
happen.
The fix is to add the D_TRACKCLOSE flag to the driver's cdevsw, so
that we get a close() call for each open(). That does happen
reliably, so we can make sure that our reference counts are
correct.
Note that the sa(4) and pt(4) drivers only allow one context
through the open routine. So these drivers aren't exposed to the
same race condition.
scsi_ch.c,
scsi_enc.c,
scsi_enc_internal.h,
scsi_pass.c,
scsi_sg.c:
For these drivers, change the open() routine to
increment the reference count for every open, and
just decrement the reference count in the close.
Call cam_periph_release_locked() in some scenarios
to avoid additional lock and unlock calls.
scsi_pt.c: Call cam_periph_release_locked() in some scenarios
to avoid additional lock and unlock calls.
MFC after: 3 days
2012-05-27 06:11:09 +00:00
|
|
|
cam_periph_release_locked(periph);
|
2007-04-15 08:49:19 +00:00
|
|
|
cam_periph_unlock(periph);
|
2001-09-26 20:13:16 +00:00
|
|
|
return(error);
|
1998-11-22 23:44:47 +00:00
|
|
|
}
|
1998-09-15 06:36:34 +00:00
|
|
|
|
1998-09-16 00:11:53 +00:00
|
|
|
/*
|
|
|
|
|
* Only allow read-write access.
|
|
|
|
|
*/
|
1998-11-22 23:44:47 +00:00
|
|
|
if (((flags & FWRITE) == 0) || ((flags & FREAD) == 0)) {
|
Work around a race condition in devfs by changing the way closes
are handled in most CAM peripheral drivers that are not handled by
GEOM's disk class.
The usual character driver open and close semantics are that the
driver gets N open calls, but only one close, when the last caller
closes the device.
CAM peripheral drivers expect that behavior to be honored to the
letter, and the CAM peripheral driver code (specifically
cam_periph_release_locked_busses()) panics if it is done incorrectly.
Since devfs has to drop its locks while it calls a driver's close
routine, and it does not have a way to delay or prevent open calls
while it is calling the close routine, there is a race.
The sequence of events, simplified a bit, is:
- devfs acquires a lock
- devfs checks the reference count, and if it is 1, continues to close.
- devfs releases the lock
- 2nd process open call on the device happens here
- devfs calls the driver's close routine
- devfs acquires a lock
- devfs decrements the reference count
- devfs releases the lock
- 2nd process close call on the device happens here
At the second close, we get a panic in
cam_periph_release_locked_busses(), complaining that peripheral
has been released when the reference count is already 0. This is
because we have gotten two closes in a row, which should not
happen.
The fix is to add the D_TRACKCLOSE flag to the driver's cdevsw, so
that we get a close() call for each open(). That does happen
reliably, so we can make sure that our reference counts are
correct.
Note that the sa(4) and pt(4) drivers only allow one context
through the open routine. So these drivers aren't exposed to the
same race condition.
scsi_ch.c,
scsi_enc.c,
scsi_enc_internal.h,
scsi_pass.c,
scsi_sg.c:
For these drivers, change the open() routine to
increment the reference count for every open, and
just decrement the reference count in the close.
Call cam_periph_release_locked() in some scenarios
to avoid additional lock and unlock calls.
scsi_pt.c: Call cam_periph_release_locked() in some scenarios
to avoid additional lock and unlock calls.
MFC after: 3 days
2012-05-27 06:11:09 +00:00
|
|
|
cam_periph_release_locked(periph);
|
2007-04-15 08:49:19 +00:00
|
|
|
cam_periph_unlock(periph);
|
1998-09-16 00:11:53 +00:00
|
|
|
return(EPERM);
|
1998-11-22 23:44:47 +00:00
|
|
|
}
|
1998-09-16 00:11:53 +00:00
|
|
|
|
1998-09-15 06:36:34 +00:00
|
|
|
/*
|
|
|
|
|
* We don't allow nonblocking access.
|
|
|
|
|
*/
|
|
|
|
|
if ((flags & O_NONBLOCK) != 0) {
|
2006-12-05 07:45:28 +00:00
|
|
|
xpt_print(periph->path, "can't do nonblocking access\n");
|
Work around a race condition in devfs by changing the way closes
are handled in most CAM peripheral drivers that are not handled by
GEOM's disk class.
The usual character driver open and close semantics are that the
driver gets N open calls, but only one close, when the last caller
closes the device.
CAM peripheral drivers expect that behavior to be honored to the
letter, and the CAM peripheral driver code (specifically
cam_periph_release_locked_busses()) panics if it is done incorrectly.
Since devfs has to drop its locks while it calls a driver's close
routine, and it does not have a way to delay or prevent open calls
while it is calling the close routine, there is a race.
The sequence of events, simplified a bit, is:
- devfs acquires a lock
- devfs checks the reference count, and if it is 1, continues to close.
- devfs releases the lock
- 2nd process open call on the device happens here
- devfs calls the driver's close routine
- devfs acquires a lock
- devfs decrements the reference count
- devfs releases the lock
- 2nd process close call on the device happens here
At the second close, we get a panic in
cam_periph_release_locked_busses(), complaining that peripheral
has been released when the reference count is already 0. This is
because we have gotten two closes in a row, which should not
happen.
The fix is to add the D_TRACKCLOSE flag to the driver's cdevsw, so
that we get a close() call for each open(). That does happen
reliably, so we can make sure that our reference counts are
correct.
Note that the sa(4) and pt(4) drivers only allow one context
through the open routine. So these drivers aren't exposed to the
same race condition.
scsi_ch.c,
scsi_enc.c,
scsi_enc_internal.h,
scsi_pass.c,
scsi_sg.c:
For these drivers, change the open() routine to
increment the reference count for every open, and
just decrement the reference count in the close.
Call cam_periph_release_locked() in some scenarios
to avoid additional lock and unlock calls.
scsi_pt.c: Call cam_periph_release_locked() in some scenarios
to avoid additional lock and unlock calls.
MFC after: 3 days
2012-05-27 06:11:09 +00:00
|
|
|
cam_periph_release_locked(periph);
|
2007-04-15 08:49:19 +00:00
|
|
|
cam_periph_unlock(periph);
|
1998-11-22 23:44:47 +00:00
|
|
|
return(EINVAL);
|
1998-09-15 06:36:34 +00:00
|
|
|
}
|
|
|
|
|
|
Fix a device departure bug for the the pass(4), enc(4), sg(4) and ch(4)
drivers.
The bug occurrs when a userland process has the driver instance
open and the underlying device goes away. We get the devfs
callback that the device node has been destroyed, but not all of
the closes necessary to fully decrement the reference count on the
CAM peripheral.
The reason is that once devfs calls back and says the device has
been destroyed, it is moved off to deadfs, and devfs guarantees
that there will be no more open or close calls. So the solution
is to keep track of how many outstanding open calls there are on
the device, and just release that many references when we get the
callback from devfs.
scsi_pass.c,
scsi_enc.c,
scsi_enc_internal.h: Add an open count to the softc in these
drivers. Increment it on open and
decrement it on close.
When we get a devfs callback to say that
the device node has gone away, decrement
the peripheral reference count by the
number of still outstanding opens.
Make sure we don't access the peripheral
with cam_periph_unlock() after what might
be the final call to
cam_periph_release_locked(). The
peripheral might have been freed, and we
will be dereferencing freed memory.
scsi_ch.c,
scsi_sg.c: For the ch(4) and sg(4) drivers, add the
same changes described above, and in
addition, fix another bug that was
previously fixed in the pass(4) and enc(4)
drivers.
These drivers were calling destroy_dev()
from their cleanup routine, but that could
cause a deadlock because the cleanup
routine could be indirectly called from
the driver's close routine. This would
cause a deadlock, because the device node
is being held open by the active close
call, and can't be destroyed.
Sponsored by: Spectra Logic Corporation
MFC after: 1 week
2012-12-08 04:03:04 +00:00
|
|
|
softc->open_count++;
|
|
|
|
|
|
Work around a race condition in devfs by changing the way closes
are handled in most CAM peripheral drivers that are not handled by
GEOM's disk class.
The usual character driver open and close semantics are that the
driver gets N open calls, but only one close, when the last caller
closes the device.
CAM peripheral drivers expect that behavior to be honored to the
letter, and the CAM peripheral driver code (specifically
cam_periph_release_locked_busses()) panics if it is done incorrectly.
Since devfs has to drop its locks while it calls a driver's close
routine, and it does not have a way to delay or prevent open calls
while it is calling the close routine, there is a race.
The sequence of events, simplified a bit, is:
- devfs acquires a lock
- devfs checks the reference count, and if it is 1, continues to close.
- devfs releases the lock
- 2nd process open call on the device happens here
- devfs calls the driver's close routine
- devfs acquires a lock
- devfs decrements the reference count
- devfs releases the lock
- 2nd process close call on the device happens here
At the second close, we get a panic in
cam_periph_release_locked_busses(), complaining that peripheral
has been released when the reference count is already 0. This is
because we have gotten two closes in a row, which should not
happen.
The fix is to add the D_TRACKCLOSE flag to the driver's cdevsw, so
that we get a close() call for each open(). That does happen
reliably, so we can make sure that our reference counts are
correct.
Note that the sa(4) and pt(4) drivers only allow one context
through the open routine. So these drivers aren't exposed to the
same race condition.
scsi_ch.c,
scsi_enc.c,
scsi_enc_internal.h,
scsi_pass.c,
scsi_sg.c:
For these drivers, change the open() routine to
increment the reference count for every open, and
just decrement the reference count in the close.
Call cam_periph_release_locked() in some scenarios
to avoid additional lock and unlock calls.
scsi_pt.c: Call cam_periph_release_locked() in some scenarios
to avoid additional lock and unlock calls.
MFC after: 3 days
2012-05-27 06:11:09 +00:00
|
|
|
cam_periph_unlock(periph);
|
1998-09-15 06:36:34 +00:00
|
|
|
|
|
|
|
|
return (error);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int
|
2004-06-16 09:47:26 +00:00
|
|
|
passclose(struct cdev *dev, int flag, int fmt, struct thread *td)
|
1998-09-15 06:36:34 +00:00
|
|
|
{
|
|
|
|
|
struct cam_periph *periph;
|
Fix a device departure bug for the the pass(4), enc(4), sg(4) and ch(4)
drivers.
The bug occurrs when a userland process has the driver instance
open and the underlying device goes away. We get the devfs
callback that the device node has been destroyed, but not all of
the closes necessary to fully decrement the reference count on the
CAM peripheral.
The reason is that once devfs calls back and says the device has
been destroyed, it is moved off to deadfs, and devfs guarantees
that there will be no more open or close calls. So the solution
is to keep track of how many outstanding open calls there are on
the device, and just release that many references when we get the
callback from devfs.
scsi_pass.c,
scsi_enc.c,
scsi_enc_internal.h: Add an open count to the softc in these
drivers. Increment it on open and
decrement it on close.
When we get a devfs callback to say that
the device node has gone away, decrement
the peripheral reference count by the
number of still outstanding opens.
Make sure we don't access the peripheral
with cam_periph_unlock() after what might
be the final call to
cam_periph_release_locked(). The
peripheral might have been freed, and we
will be dereferencing freed memory.
scsi_ch.c,
scsi_sg.c: For the ch(4) and sg(4) drivers, add the
same changes described above, and in
addition, fix another bug that was
previously fixed in the pass(4) and enc(4)
drivers.
These drivers were calling destroy_dev()
from their cleanup routine, but that could
cause a deadlock because the cleanup
routine could be indirectly called from
the driver's close routine. This would
cause a deadlock, because the device node
is being held open by the active close
call, and can't be destroyed.
Sponsored by: Spectra Logic Corporation
MFC after: 1 week
2012-12-08 04:03:04 +00:00
|
|
|
struct pass_softc *softc;
|
Merge CAM locking changes from the projects/camlock branch to radically
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
2013-10-21 12:00:26 +00:00
|
|
|
struct mtx *mtx;
|
1998-09-15 06:36:34 +00:00
|
|
|
|
2002-08-15 20:54:03 +00:00
|
|
|
periph = (struct cam_periph *)dev->si_drv1;
|
Merge CAM locking changes from the projects/camlock branch to radically
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
2013-10-21 12:00:26 +00:00
|
|
|
mtx = cam_periph_mtx(periph);
|
|
|
|
|
mtx_lock(mtx);
|
1998-09-15 06:36:34 +00:00
|
|
|
|
Fix a device departure bug for the the pass(4), enc(4), sg(4) and ch(4)
drivers.
The bug occurrs when a userland process has the driver instance
open and the underlying device goes away. We get the devfs
callback that the device node has been destroyed, but not all of
the closes necessary to fully decrement the reference count on the
CAM peripheral.
The reason is that once devfs calls back and says the device has
been destroyed, it is moved off to deadfs, and devfs guarantees
that there will be no more open or close calls. So the solution
is to keep track of how many outstanding open calls there are on
the device, and just release that many references when we get the
callback from devfs.
scsi_pass.c,
scsi_enc.c,
scsi_enc_internal.h: Add an open count to the softc in these
drivers. Increment it on open and
decrement it on close.
When we get a devfs callback to say that
the device node has gone away, decrement
the peripheral reference count by the
number of still outstanding opens.
Make sure we don't access the peripheral
with cam_periph_unlock() after what might
be the final call to
cam_periph_release_locked(). The
peripheral might have been freed, and we
will be dereferencing freed memory.
scsi_ch.c,
scsi_sg.c: For the ch(4) and sg(4) drivers, add the
same changes described above, and in
addition, fix another bug that was
previously fixed in the pass(4) and enc(4)
drivers.
These drivers were calling destroy_dev()
from their cleanup routine, but that could
cause a deadlock because the cleanup
routine could be indirectly called from
the driver's close routine. This would
cause a deadlock, because the device node
is being held open by the active close
call, and can't be destroyed.
Sponsored by: Spectra Logic Corporation
MFC after: 1 week
2012-12-08 04:03:04 +00:00
|
|
|
softc = periph->softc;
|
|
|
|
|
softc->open_count--;
|
|
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
if (softc->open_count == 0) {
|
|
|
|
|
struct pass_io_req *io_req, *io_req2;
|
|
|
|
|
int need_unlock;
|
|
|
|
|
|
|
|
|
|
need_unlock = 0;
|
|
|
|
|
|
|
|
|
|
TAILQ_FOREACH_SAFE(io_req, &softc->done_queue, links, io_req2) {
|
|
|
|
|
TAILQ_REMOVE(&softc->done_queue, io_req, links);
|
|
|
|
|
passiocleanup(softc, io_req);
|
|
|
|
|
uma_zfree(softc->pass_zone, io_req);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
TAILQ_FOREACH_SAFE(io_req, &softc->incoming_queue, links,
|
|
|
|
|
io_req2) {
|
|
|
|
|
TAILQ_REMOVE(&softc->incoming_queue, io_req, links);
|
|
|
|
|
passiocleanup(softc, io_req);
|
|
|
|
|
uma_zfree(softc->pass_zone, io_req);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* If there are any active I/Os, we need to forcibly acquire a
|
|
|
|
|
* reference to the peripheral so that we don't go away
|
|
|
|
|
* before they complete. We'll release the reference when
|
|
|
|
|
* the abandoned queue is empty.
|
|
|
|
|
*/
|
|
|
|
|
io_req = TAILQ_FIRST(&softc->active_queue);
|
|
|
|
|
if ((io_req != NULL)
|
|
|
|
|
&& (softc->flags & PASS_FLAG_ABANDONED_REF_SET) == 0) {
|
|
|
|
|
cam_periph_doacquire(periph);
|
|
|
|
|
softc->flags |= PASS_FLAG_ABANDONED_REF_SET;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Since the I/O in the active queue is not under our
|
|
|
|
|
* control, just set a flag so that we can clean it up when
|
|
|
|
|
* it completes and put it on the abandoned queue. This
|
|
|
|
|
* will prevent our sending spurious completions in the
|
|
|
|
|
* event that the device is opened again before these I/Os
|
|
|
|
|
* complete.
|
|
|
|
|
*/
|
|
|
|
|
TAILQ_FOREACH_SAFE(io_req, &softc->active_queue, links,
|
|
|
|
|
io_req2) {
|
|
|
|
|
TAILQ_REMOVE(&softc->active_queue, io_req, links);
|
|
|
|
|
io_req->flags |= PASS_IO_ABANDONED;
|
|
|
|
|
TAILQ_INSERT_TAIL(&softc->abandoned_queue, io_req,
|
|
|
|
|
links);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
Fix a device departure bug for the the pass(4), enc(4), sg(4) and ch(4)
drivers.
The bug occurrs when a userland process has the driver instance
open and the underlying device goes away. We get the devfs
callback that the device node has been destroyed, but not all of
the closes necessary to fully decrement the reference count on the
CAM peripheral.
The reason is that once devfs calls back and says the device has
been destroyed, it is moved off to deadfs, and devfs guarantees
that there will be no more open or close calls. So the solution
is to keep track of how many outstanding open calls there are on
the device, and just release that many references when we get the
callback from devfs.
scsi_pass.c,
scsi_enc.c,
scsi_enc_internal.h: Add an open count to the softc in these
drivers. Increment it on open and
decrement it on close.
When we get a devfs callback to say that
the device node has gone away, decrement
the peripheral reference count by the
number of still outstanding opens.
Make sure we don't access the peripheral
with cam_periph_unlock() after what might
be the final call to
cam_periph_release_locked(). The
peripheral might have been freed, and we
will be dereferencing freed memory.
scsi_ch.c,
scsi_sg.c: For the ch(4) and sg(4) drivers, add the
same changes described above, and in
addition, fix another bug that was
previously fixed in the pass(4) and enc(4)
drivers.
These drivers were calling destroy_dev()
from their cleanup routine, but that could
cause a deadlock because the cleanup
routine could be indirectly called from
the driver's close routine. This would
cause a deadlock, because the device node
is being held open by the active close
call, and can't be destroyed.
Sponsored by: Spectra Logic Corporation
MFC after: 1 week
2012-12-08 04:03:04 +00:00
|
|
|
cam_periph_release_locked(periph);
|
|
|
|
|
|
|
|
|
|
/*
|
Merge CAM locking changes from the projects/camlock branch to radically
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
2013-10-21 12:00:26 +00:00
|
|
|
* We reference the lock directly here, instead of using
|
Fix a device departure bug for the the pass(4), enc(4), sg(4) and ch(4)
drivers.
The bug occurrs when a userland process has the driver instance
open and the underlying device goes away. We get the devfs
callback that the device node has been destroyed, but not all of
the closes necessary to fully decrement the reference count on the
CAM peripheral.
The reason is that once devfs calls back and says the device has
been destroyed, it is moved off to deadfs, and devfs guarantees
that there will be no more open or close calls. So the solution
is to keep track of how many outstanding open calls there are on
the device, and just release that many references when we get the
callback from devfs.
scsi_pass.c,
scsi_enc.c,
scsi_enc_internal.h: Add an open count to the softc in these
drivers. Increment it on open and
decrement it on close.
When we get a devfs callback to say that
the device node has gone away, decrement
the peripheral reference count by the
number of still outstanding opens.
Make sure we don't access the peripheral
with cam_periph_unlock() after what might
be the final call to
cam_periph_release_locked(). The
peripheral might have been freed, and we
will be dereferencing freed memory.
scsi_ch.c,
scsi_sg.c: For the ch(4) and sg(4) drivers, add the
same changes described above, and in
addition, fix another bug that was
previously fixed in the pass(4) and enc(4)
drivers.
These drivers were calling destroy_dev()
from their cleanup routine, but that could
cause a deadlock because the cleanup
routine could be indirectly called from
the driver's close routine. This would
cause a deadlock, because the device node
is being held open by the active close
call, and can't be destroyed.
Sponsored by: Spectra Logic Corporation
MFC after: 1 week
2012-12-08 04:03:04 +00:00
|
|
|
* cam_periph_unlock(). The reason is that the call to
|
|
|
|
|
* cam_periph_release_locked() above could result in the periph
|
|
|
|
|
* getting freed. If that is the case, dereferencing the periph
|
|
|
|
|
* with a cam_periph_unlock() call would cause a page fault.
|
|
|
|
|
*
|
|
|
|
|
* cam_periph_release() avoids this problem using the same method,
|
|
|
|
|
* but we're manually acquiring and dropping the lock here to
|
|
|
|
|
* protect the open count and avoid another lock acquisition and
|
|
|
|
|
* release.
|
|
|
|
|
*/
|
Merge CAM locking changes from the projects/camlock branch to radically
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
2013-10-21 12:00:26 +00:00
|
|
|
mtx_unlock(mtx);
|
1998-09-15 06:36:34 +00:00
|
|
|
|
|
|
|
|
return (0);
|
|
|
|
|
}
|
|
|
|
|
|
2013-06-17 08:57:09 +00:00
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
static void
|
|
|
|
|
passstart(struct cam_periph *periph, union ccb *start_ccb)
|
1998-09-15 06:36:34 +00:00
|
|
|
{
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
struct pass_softc *softc;
|
1998-09-15 06:36:34 +00:00
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
softc = (struct pass_softc *)periph->softc;
|
1998-09-15 06:36:34 +00:00
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
switch (softc->state) {
|
|
|
|
|
case PASS_STATE_NORMAL: {
|
|
|
|
|
struct pass_io_req *io_req;
|
Add a number of interrelated CAM feature enhancements and bug fixes.
NOTE: These changes will require recompilation of any userland
applications, like cdrecord, xmcd, etc., that use the CAM passthrough
interface. A make world is recommended.
camcontrol.[c8]:
- We now support two new commands, "tags" and "negotiate".
- The tags commands allows users to view the number of tagged
openings for a device as well as a number of other related
parameters, and it allows users to set tagged openings for
a device.
- The negotiate command allows users to enable and disable
disconnection and tagged queueing, set sync rates, offsets
and bus width. Note that not all of those features are
available for all controllers. Only the adv, ahc, and ncr
drivers fully support all of the features at this point.
Some cards do not allow the setting of sync rates, offsets and
the like, and some of the drivers don't have any facilities to
do so. Some drivers, like the adw driver, only support enabling
or disabling sync negotiation, but do not support setting sync
rates.
- new description in the camcontrol man page of how to format a disk
- cleanup of the camcontrol inquiry command
- add support in the 'devlist' command for skipping unconfigured devices if
-v was not specified on the command line.
- make use of the new base_transfer_speed in the path inquiry CCB.
- fix CCB bzero cases
cam_xpt.c, cam_sim.[ch], cam_ccb.h:
- new flags on many CCB function codes to designate whether they're
non-immediate, use a user-supplied CCB, and can only be passed from
userland programs via the xpt device. Use these flags in the transport
layer and pass driver to categorize CCBs.
- new flag in the transport layer device matching code for device nodes
that indicates whether a device is unconfigured
- bump the CAM version from 0x10 to 0x11
- Change the CAM ioctls to use the version as their group code, so we can
force users to recompile code even when the CCB size doesn't change.
- add + fill in a new value in the path inquiry CCB, base_transfer_speed.
Remove a corresponding field from the cam_sim structure, and add code to
every SIM to set this field to the proper value.
- Fix the set transfer settings code in the transport layer.
scsi_cd.c:
- make some variables volatile instead of just casting them in various
places
- fix a race condition in the changer code
- attach unless we get a "logical unit not supported" error. This should
fix all of the cases where people have devices that return weird errors
when they don't have media in the drive.
scsi_da.c:
- attach unless we get a "logical unit not supported" error
scsi_pass.c:
- for immediate CCBs, just malloc a CCB to send the user request in. This
gets rid of the 'held' count problem in camcontrol tags.
scsi_pass.h:
- change the CAM ioctls to use the CAM version as their group code.
adv driver:
- Allow changing the sync rate and offset separately.
adw driver
- Allow changing the sync rate and offset separately.
aha driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
ahc driver:
- Allow setting offset and sync rate separately
bt driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
NCR driver:
- Fix the ultra/ultra 2 negotiation bug
- allow setting both the sync rate and offset separately
Other HBA drivers:
- Put code in to set the base_transfer_speed field for
XPT_GET_TRAN_SETTINGS CCBs.
Reviewed by: gibbs, mjacob (isp), imp (aha)
1999-05-06 20:16:39 +00:00
|
|
|
|
|
|
|
|
/*
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
* Check for any queued I/O requests that require an
|
|
|
|
|
* allocated slot.
|
Add a number of interrelated CAM feature enhancements and bug fixes.
NOTE: These changes will require recompilation of any userland
applications, like cdrecord, xmcd, etc., that use the CAM passthrough
interface. A make world is recommended.
camcontrol.[c8]:
- We now support two new commands, "tags" and "negotiate".
- The tags commands allows users to view the number of tagged
openings for a device as well as a number of other related
parameters, and it allows users to set tagged openings for
a device.
- The negotiate command allows users to enable and disable
disconnection and tagged queueing, set sync rates, offsets
and bus width. Note that not all of those features are
available for all controllers. Only the adv, ahc, and ncr
drivers fully support all of the features at this point.
Some cards do not allow the setting of sync rates, offsets and
the like, and some of the drivers don't have any facilities to
do so. Some drivers, like the adw driver, only support enabling
or disabling sync negotiation, but do not support setting sync
rates.
- new description in the camcontrol man page of how to format a disk
- cleanup of the camcontrol inquiry command
- add support in the 'devlist' command for skipping unconfigured devices if
-v was not specified on the command line.
- make use of the new base_transfer_speed in the path inquiry CCB.
- fix CCB bzero cases
cam_xpt.c, cam_sim.[ch], cam_ccb.h:
- new flags on many CCB function codes to designate whether they're
non-immediate, use a user-supplied CCB, and can only be passed from
userland programs via the xpt device. Use these flags in the transport
layer and pass driver to categorize CCBs.
- new flag in the transport layer device matching code for device nodes
that indicates whether a device is unconfigured
- bump the CAM version from 0x10 to 0x11
- Change the CAM ioctls to use the version as their group code, so we can
force users to recompile code even when the CCB size doesn't change.
- add + fill in a new value in the path inquiry CCB, base_transfer_speed.
Remove a corresponding field from the cam_sim structure, and add code to
every SIM to set this field to the proper value.
- Fix the set transfer settings code in the transport layer.
scsi_cd.c:
- make some variables volatile instead of just casting them in various
places
- fix a race condition in the changer code
- attach unless we get a "logical unit not supported" error. This should
fix all of the cases where people have devices that return weird errors
when they don't have media in the drive.
scsi_da.c:
- attach unless we get a "logical unit not supported" error
scsi_pass.c:
- for immediate CCBs, just malloc a CCB to send the user request in. This
gets rid of the 'held' count problem in camcontrol tags.
scsi_pass.h:
- change the CAM ioctls to use the CAM version as their group code.
adv driver:
- Allow changing the sync rate and offset separately.
adw driver
- Allow changing the sync rate and offset separately.
aha driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
ahc driver:
- Allow setting offset and sync rate separately
bt driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
NCR driver:
- Fix the ultra/ultra 2 negotiation bug
- allow setting both the sync rate and offset separately
Other HBA drivers:
- Put code in to set the base_transfer_speed field for
XPT_GET_TRAN_SETTINGS CCBs.
Reviewed by: gibbs, mjacob (isp), imp (aha)
1999-05-06 20:16:39 +00:00
|
|
|
*/
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
io_req = TAILQ_FIRST(&softc->incoming_queue);
|
|
|
|
|
if (io_req == NULL) {
|
|
|
|
|
xpt_release_ccb(start_ccb);
|
Add a number of interrelated CAM feature enhancements and bug fixes.
NOTE: These changes will require recompilation of any userland
applications, like cdrecord, xmcd, etc., that use the CAM passthrough
interface. A make world is recommended.
camcontrol.[c8]:
- We now support two new commands, "tags" and "negotiate".
- The tags commands allows users to view the number of tagged
openings for a device as well as a number of other related
parameters, and it allows users to set tagged openings for
a device.
- The negotiate command allows users to enable and disable
disconnection and tagged queueing, set sync rates, offsets
and bus width. Note that not all of those features are
available for all controllers. Only the adv, ahc, and ncr
drivers fully support all of the features at this point.
Some cards do not allow the setting of sync rates, offsets and
the like, and some of the drivers don't have any facilities to
do so. Some drivers, like the adw driver, only support enabling
or disabling sync negotiation, but do not support setting sync
rates.
- new description in the camcontrol man page of how to format a disk
- cleanup of the camcontrol inquiry command
- add support in the 'devlist' command for skipping unconfigured devices if
-v was not specified on the command line.
- make use of the new base_transfer_speed in the path inquiry CCB.
- fix CCB bzero cases
cam_xpt.c, cam_sim.[ch], cam_ccb.h:
- new flags on many CCB function codes to designate whether they're
non-immediate, use a user-supplied CCB, and can only be passed from
userland programs via the xpt device. Use these flags in the transport
layer and pass driver to categorize CCBs.
- new flag in the transport layer device matching code for device nodes
that indicates whether a device is unconfigured
- bump the CAM version from 0x10 to 0x11
- Change the CAM ioctls to use the version as their group code, so we can
force users to recompile code even when the CCB size doesn't change.
- add + fill in a new value in the path inquiry CCB, base_transfer_speed.
Remove a corresponding field from the cam_sim structure, and add code to
every SIM to set this field to the proper value.
- Fix the set transfer settings code in the transport layer.
scsi_cd.c:
- make some variables volatile instead of just casting them in various
places
- fix a race condition in the changer code
- attach unless we get a "logical unit not supported" error. This should
fix all of the cases where people have devices that return weird errors
when they don't have media in the drive.
scsi_da.c:
- attach unless we get a "logical unit not supported" error
scsi_pass.c:
- for immediate CCBs, just malloc a CCB to send the user request in. This
gets rid of the 'held' count problem in camcontrol tags.
scsi_pass.h:
- change the CAM ioctls to use the CAM version as their group code.
adv driver:
- Allow changing the sync rate and offset separately.
adw driver
- Allow changing the sync rate and offset separately.
aha driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
ahc driver:
- Allow setting offset and sync rate separately
bt driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
NCR driver:
- Fix the ultra/ultra 2 negotiation bug
- allow setting both the sync rate and offset separately
Other HBA drivers:
- Put code in to set the base_transfer_speed field for
XPT_GET_TRAN_SETTINGS CCBs.
Reviewed by: gibbs, mjacob (isp), imp (aha)
1999-05-06 20:16:39 +00:00
|
|
|
break;
|
|
|
|
|
}
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
TAILQ_REMOVE(&softc->incoming_queue, io_req, links);
|
|
|
|
|
TAILQ_INSERT_TAIL(&softc->active_queue, io_req, links);
|
|
|
|
|
/*
|
|
|
|
|
* Merge the user's CCB into the allocated CCB.
|
|
|
|
|
*/
|
|
|
|
|
xpt_merge_ccb(start_ccb, &io_req->ccb);
|
|
|
|
|
start_ccb->ccb_h.ccb_type = PASS_CCB_QUEUED_IO;
|
|
|
|
|
start_ccb->ccb_h.ccb_ioreq = io_req;
|
|
|
|
|
start_ccb->ccb_h.cbfcnp = passdone;
|
|
|
|
|
io_req->alloced_ccb = start_ccb;
|
|
|
|
|
binuptime(&io_req->start_time);
|
|
|
|
|
devstat_start_transaction(softc->device_stats,
|
|
|
|
|
&io_req->start_time);
|
Add a number of interrelated CAM feature enhancements and bug fixes.
NOTE: These changes will require recompilation of any userland
applications, like cdrecord, xmcd, etc., that use the CAM passthrough
interface. A make world is recommended.
camcontrol.[c8]:
- We now support two new commands, "tags" and "negotiate".
- The tags commands allows users to view the number of tagged
openings for a device as well as a number of other related
parameters, and it allows users to set tagged openings for
a device.
- The negotiate command allows users to enable and disable
disconnection and tagged queueing, set sync rates, offsets
and bus width. Note that not all of those features are
available for all controllers. Only the adv, ahc, and ncr
drivers fully support all of the features at this point.
Some cards do not allow the setting of sync rates, offsets and
the like, and some of the drivers don't have any facilities to
do so. Some drivers, like the adw driver, only support enabling
or disabling sync negotiation, but do not support setting sync
rates.
- new description in the camcontrol man page of how to format a disk
- cleanup of the camcontrol inquiry command
- add support in the 'devlist' command for skipping unconfigured devices if
-v was not specified on the command line.
- make use of the new base_transfer_speed in the path inquiry CCB.
- fix CCB bzero cases
cam_xpt.c, cam_sim.[ch], cam_ccb.h:
- new flags on many CCB function codes to designate whether they're
non-immediate, use a user-supplied CCB, and can only be passed from
userland programs via the xpt device. Use these flags in the transport
layer and pass driver to categorize CCBs.
- new flag in the transport layer device matching code for device nodes
that indicates whether a device is unconfigured
- bump the CAM version from 0x10 to 0x11
- Change the CAM ioctls to use the version as their group code, so we can
force users to recompile code even when the CCB size doesn't change.
- add + fill in a new value in the path inquiry CCB, base_transfer_speed.
Remove a corresponding field from the cam_sim structure, and add code to
every SIM to set this field to the proper value.
- Fix the set transfer settings code in the transport layer.
scsi_cd.c:
- make some variables volatile instead of just casting them in various
places
- fix a race condition in the changer code
- attach unless we get a "logical unit not supported" error. This should
fix all of the cases where people have devices that return weird errors
when they don't have media in the drive.
scsi_da.c:
- attach unless we get a "logical unit not supported" error
scsi_pass.c:
- for immediate CCBs, just malloc a CCB to send the user request in. This
gets rid of the 'held' count problem in camcontrol tags.
scsi_pass.h:
- change the CAM ioctls to use the CAM version as their group code.
adv driver:
- Allow changing the sync rate and offset separately.
adw driver
- Allow changing the sync rate and offset separately.
aha driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
ahc driver:
- Allow setting offset and sync rate separately
bt driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
NCR driver:
- Fix the ultra/ultra 2 negotiation bug
- allow setting both the sync rate and offset separately
Other HBA drivers:
- Put code in to set the base_transfer_speed field for
XPT_GET_TRAN_SETTINGS CCBs.
Reviewed by: gibbs, mjacob (isp), imp (aha)
1999-05-06 20:16:39 +00:00
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
xpt_action(start_ccb);
|
2012-10-27 10:14:12 +00:00
|
|
|
|
Add a number of interrelated CAM feature enhancements and bug fixes.
NOTE: These changes will require recompilation of any userland
applications, like cdrecord, xmcd, etc., that use the CAM passthrough
interface. A make world is recommended.
camcontrol.[c8]:
- We now support two new commands, "tags" and "negotiate".
- The tags commands allows users to view the number of tagged
openings for a device as well as a number of other related
parameters, and it allows users to set tagged openings for
a device.
- The negotiate command allows users to enable and disable
disconnection and tagged queueing, set sync rates, offsets
and bus width. Note that not all of those features are
available for all controllers. Only the adv, ahc, and ncr
drivers fully support all of the features at this point.
Some cards do not allow the setting of sync rates, offsets and
the like, and some of the drivers don't have any facilities to
do so. Some drivers, like the adw driver, only support enabling
or disabling sync negotiation, but do not support setting sync
rates.
- new description in the camcontrol man page of how to format a disk
- cleanup of the camcontrol inquiry command
- add support in the 'devlist' command for skipping unconfigured devices if
-v was not specified on the command line.
- make use of the new base_transfer_speed in the path inquiry CCB.
- fix CCB bzero cases
cam_xpt.c, cam_sim.[ch], cam_ccb.h:
- new flags on many CCB function codes to designate whether they're
non-immediate, use a user-supplied CCB, and can only be passed from
userland programs via the xpt device. Use these flags in the transport
layer and pass driver to categorize CCBs.
- new flag in the transport layer device matching code for device nodes
that indicates whether a device is unconfigured
- bump the CAM version from 0x10 to 0x11
- Change the CAM ioctls to use the version as their group code, so we can
force users to recompile code even when the CCB size doesn't change.
- add + fill in a new value in the path inquiry CCB, base_transfer_speed.
Remove a corresponding field from the cam_sim structure, and add code to
every SIM to set this field to the proper value.
- Fix the set transfer settings code in the transport layer.
scsi_cd.c:
- make some variables volatile instead of just casting them in various
places
- fix a race condition in the changer code
- attach unless we get a "logical unit not supported" error. This should
fix all of the cases where people have devices that return weird errors
when they don't have media in the drive.
scsi_da.c:
- attach unless we get a "logical unit not supported" error
scsi_pass.c:
- for immediate CCBs, just malloc a CCB to send the user request in. This
gets rid of the 'held' count problem in camcontrol tags.
scsi_pass.h:
- change the CAM ioctls to use the CAM version as their group code.
adv driver:
- Allow changing the sync rate and offset separately.
adw driver
- Allow changing the sync rate and offset separately.
aha driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
ahc driver:
- Allow setting offset and sync rate separately
bt driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
NCR driver:
- Fix the ultra/ultra 2 negotiation bug
- allow setting both the sync rate and offset separately
Other HBA drivers:
- Put code in to set the base_transfer_speed field for
XPT_GET_TRAN_SETTINGS CCBs.
Reviewed by: gibbs, mjacob (isp), imp (aha)
1999-05-06 20:16:39 +00:00
|
|
|
/*
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
* If we have any more I/O waiting, schedule ourselves again.
|
Add a number of interrelated CAM feature enhancements and bug fixes.
NOTE: These changes will require recompilation of any userland
applications, like cdrecord, xmcd, etc., that use the CAM passthrough
interface. A make world is recommended.
camcontrol.[c8]:
- We now support two new commands, "tags" and "negotiate".
- The tags commands allows users to view the number of tagged
openings for a device as well as a number of other related
parameters, and it allows users to set tagged openings for
a device.
- The negotiate command allows users to enable and disable
disconnection and tagged queueing, set sync rates, offsets
and bus width. Note that not all of those features are
available for all controllers. Only the adv, ahc, and ncr
drivers fully support all of the features at this point.
Some cards do not allow the setting of sync rates, offsets and
the like, and some of the drivers don't have any facilities to
do so. Some drivers, like the adw driver, only support enabling
or disabling sync negotiation, but do not support setting sync
rates.
- new description in the camcontrol man page of how to format a disk
- cleanup of the camcontrol inquiry command
- add support in the 'devlist' command for skipping unconfigured devices if
-v was not specified on the command line.
- make use of the new base_transfer_speed in the path inquiry CCB.
- fix CCB bzero cases
cam_xpt.c, cam_sim.[ch], cam_ccb.h:
- new flags on many CCB function codes to designate whether they're
non-immediate, use a user-supplied CCB, and can only be passed from
userland programs via the xpt device. Use these flags in the transport
layer and pass driver to categorize CCBs.
- new flag in the transport layer device matching code for device nodes
that indicates whether a device is unconfigured
- bump the CAM version from 0x10 to 0x11
- Change the CAM ioctls to use the version as their group code, so we can
force users to recompile code even when the CCB size doesn't change.
- add + fill in a new value in the path inquiry CCB, base_transfer_speed.
Remove a corresponding field from the cam_sim structure, and add code to
every SIM to set this field to the proper value.
- Fix the set transfer settings code in the transport layer.
scsi_cd.c:
- make some variables volatile instead of just casting them in various
places
- fix a race condition in the changer code
- attach unless we get a "logical unit not supported" error. This should
fix all of the cases where people have devices that return weird errors
when they don't have media in the drive.
scsi_da.c:
- attach unless we get a "logical unit not supported" error
scsi_pass.c:
- for immediate CCBs, just malloc a CCB to send the user request in. This
gets rid of the 'held' count problem in camcontrol tags.
scsi_pass.h:
- change the CAM ioctls to use the CAM version as their group code.
adv driver:
- Allow changing the sync rate and offset separately.
adw driver
- Allow changing the sync rate and offset separately.
aha driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
ahc driver:
- Allow setting offset and sync rate separately
bt driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
NCR driver:
- Fix the ultra/ultra 2 negotiation bug
- allow setting both the sync rate and offset separately
Other HBA drivers:
- Put code in to set the base_transfer_speed field for
XPT_GET_TRAN_SETTINGS CCBs.
Reviewed by: gibbs, mjacob (isp), imp (aha)
1999-05-06 20:16:39 +00:00
|
|
|
*/
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
if (!TAILQ_EMPTY(&softc->incoming_queue))
|
|
|
|
|
xpt_schedule(periph, CAM_PRIORITY_NORMAL);
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
default:
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
Add a number of interrelated CAM feature enhancements and bug fixes.
NOTE: These changes will require recompilation of any userland
applications, like cdrecord, xmcd, etc., that use the CAM passthrough
interface. A make world is recommended.
camcontrol.[c8]:
- We now support two new commands, "tags" and "negotiate".
- The tags commands allows users to view the number of tagged
openings for a device as well as a number of other related
parameters, and it allows users to set tagged openings for
a device.
- The negotiate command allows users to enable and disable
disconnection and tagged queueing, set sync rates, offsets
and bus width. Note that not all of those features are
available for all controllers. Only the adv, ahc, and ncr
drivers fully support all of the features at this point.
Some cards do not allow the setting of sync rates, offsets and
the like, and some of the drivers don't have any facilities to
do so. Some drivers, like the adw driver, only support enabling
or disabling sync negotiation, but do not support setting sync
rates.
- new description in the camcontrol man page of how to format a disk
- cleanup of the camcontrol inquiry command
- add support in the 'devlist' command for skipping unconfigured devices if
-v was not specified on the command line.
- make use of the new base_transfer_speed in the path inquiry CCB.
- fix CCB bzero cases
cam_xpt.c, cam_sim.[ch], cam_ccb.h:
- new flags on many CCB function codes to designate whether they're
non-immediate, use a user-supplied CCB, and can only be passed from
userland programs via the xpt device. Use these flags in the transport
layer and pass driver to categorize CCBs.
- new flag in the transport layer device matching code for device nodes
that indicates whether a device is unconfigured
- bump the CAM version from 0x10 to 0x11
- Change the CAM ioctls to use the version as their group code, so we can
force users to recompile code even when the CCB size doesn't change.
- add + fill in a new value in the path inquiry CCB, base_transfer_speed.
Remove a corresponding field from the cam_sim structure, and add code to
every SIM to set this field to the proper value.
- Fix the set transfer settings code in the transport layer.
scsi_cd.c:
- make some variables volatile instead of just casting them in various
places
- fix a race condition in the changer code
- attach unless we get a "logical unit not supported" error. This should
fix all of the cases where people have devices that return weird errors
when they don't have media in the drive.
scsi_da.c:
- attach unless we get a "logical unit not supported" error
scsi_pass.c:
- for immediate CCBs, just malloc a CCB to send the user request in. This
gets rid of the 'held' count problem in camcontrol tags.
scsi_pass.h:
- change the CAM ioctls to use the CAM version as their group code.
adv driver:
- Allow changing the sync rate and offset separately.
adw driver
- Allow changing the sync rate and offset separately.
aha driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
ahc driver:
- Allow setting offset and sync rate separately
bt driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
NCR driver:
- Fix the ultra/ultra 2 negotiation bug
- allow setting both the sync rate and offset separately
Other HBA drivers:
- Put code in to set the base_transfer_speed field for
XPT_GET_TRAN_SETTINGS CCBs.
Reviewed by: gibbs, mjacob (isp), imp (aha)
1999-05-06 20:16:39 +00:00
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
static void
|
|
|
|
|
passdone(struct cam_periph *periph, union ccb *done_ccb)
|
|
|
|
|
{
|
|
|
|
|
struct pass_softc *softc;
|
|
|
|
|
struct ccb_scsiio *csio;
|
Add a number of interrelated CAM feature enhancements and bug fixes.
NOTE: These changes will require recompilation of any userland
applications, like cdrecord, xmcd, etc., that use the CAM passthrough
interface. A make world is recommended.
camcontrol.[c8]:
- We now support two new commands, "tags" and "negotiate".
- The tags commands allows users to view the number of tagged
openings for a device as well as a number of other related
parameters, and it allows users to set tagged openings for
a device.
- The negotiate command allows users to enable and disable
disconnection and tagged queueing, set sync rates, offsets
and bus width. Note that not all of those features are
available for all controllers. Only the adv, ahc, and ncr
drivers fully support all of the features at this point.
Some cards do not allow the setting of sync rates, offsets and
the like, and some of the drivers don't have any facilities to
do so. Some drivers, like the adw driver, only support enabling
or disabling sync negotiation, but do not support setting sync
rates.
- new description in the camcontrol man page of how to format a disk
- cleanup of the camcontrol inquiry command
- add support in the 'devlist' command for skipping unconfigured devices if
-v was not specified on the command line.
- make use of the new base_transfer_speed in the path inquiry CCB.
- fix CCB bzero cases
cam_xpt.c, cam_sim.[ch], cam_ccb.h:
- new flags on many CCB function codes to designate whether they're
non-immediate, use a user-supplied CCB, and can only be passed from
userland programs via the xpt device. Use these flags in the transport
layer and pass driver to categorize CCBs.
- new flag in the transport layer device matching code for device nodes
that indicates whether a device is unconfigured
- bump the CAM version from 0x10 to 0x11
- Change the CAM ioctls to use the version as their group code, so we can
force users to recompile code even when the CCB size doesn't change.
- add + fill in a new value in the path inquiry CCB, base_transfer_speed.
Remove a corresponding field from the cam_sim structure, and add code to
every SIM to set this field to the proper value.
- Fix the set transfer settings code in the transport layer.
scsi_cd.c:
- make some variables volatile instead of just casting them in various
places
- fix a race condition in the changer code
- attach unless we get a "logical unit not supported" error. This should
fix all of the cases where people have devices that return weird errors
when they don't have media in the drive.
scsi_da.c:
- attach unless we get a "logical unit not supported" error
scsi_pass.c:
- for immediate CCBs, just malloc a CCB to send the user request in. This
gets rid of the 'held' count problem in camcontrol tags.
scsi_pass.h:
- change the CAM ioctls to use the CAM version as their group code.
adv driver:
- Allow changing the sync rate and offset separately.
adw driver
- Allow changing the sync rate and offset separately.
aha driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
ahc driver:
- Allow setting offset and sync rate separately
bt driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
NCR driver:
- Fix the ultra/ultra 2 negotiation bug
- allow setting both the sync rate and offset separately
Other HBA drivers:
- Put code in to set the base_transfer_speed field for
XPT_GET_TRAN_SETTINGS CCBs.
Reviewed by: gibbs, mjacob (isp), imp (aha)
1999-05-06 20:16:39 +00:00
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
softc = (struct pass_softc *)periph->softc;
|
|
|
|
|
|
|
|
|
|
cam_periph_assert(periph, MA_OWNED);
|
|
|
|
|
|
|
|
|
|
csio = &done_ccb->csio;
|
|
|
|
|
switch (csio->ccb_h.ccb_type) {
|
|
|
|
|
case PASS_CCB_QUEUED_IO: {
|
|
|
|
|
struct pass_io_req *io_req;
|
|
|
|
|
|
|
|
|
|
io_req = done_ccb->ccb_h.ccb_ioreq;
|
|
|
|
|
#if 0
|
|
|
|
|
xpt_print(periph->path, "%s: called for user CCB %p\n",
|
|
|
|
|
__func__, io_req->user_ccb_ptr);
|
|
|
|
|
#endif
|
|
|
|
|
if (((done_ccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP)
|
|
|
|
|
&& (done_ccb->ccb_h.flags & CAM_PASS_ERR_RECOVER)
|
|
|
|
|
&& ((io_req->flags & PASS_IO_ABANDONED) == 0)) {
|
|
|
|
|
int error;
|
|
|
|
|
|
|
|
|
|
error = passerror(done_ccb, CAM_RETRY_SELTO,
|
|
|
|
|
SF_RETRY_UA | SF_NO_PRINT);
|
|
|
|
|
|
|
|
|
|
if (error == ERESTART) {
|
|
|
|
|
/*
|
|
|
|
|
* A retry was scheduled, so
|
|
|
|
|
* just return.
|
|
|
|
|
*/
|
|
|
|
|
return;
|
|
|
|
|
}
|
Add a number of interrelated CAM feature enhancements and bug fixes.
NOTE: These changes will require recompilation of any userland
applications, like cdrecord, xmcd, etc., that use the CAM passthrough
interface. A make world is recommended.
camcontrol.[c8]:
- We now support two new commands, "tags" and "negotiate".
- The tags commands allows users to view the number of tagged
openings for a device as well as a number of other related
parameters, and it allows users to set tagged openings for
a device.
- The negotiate command allows users to enable and disable
disconnection and tagged queueing, set sync rates, offsets
and bus width. Note that not all of those features are
available for all controllers. Only the adv, ahc, and ncr
drivers fully support all of the features at this point.
Some cards do not allow the setting of sync rates, offsets and
the like, and some of the drivers don't have any facilities to
do so. Some drivers, like the adw driver, only support enabling
or disabling sync negotiation, but do not support setting sync
rates.
- new description in the camcontrol man page of how to format a disk
- cleanup of the camcontrol inquiry command
- add support in the 'devlist' command for skipping unconfigured devices if
-v was not specified on the command line.
- make use of the new base_transfer_speed in the path inquiry CCB.
- fix CCB bzero cases
cam_xpt.c, cam_sim.[ch], cam_ccb.h:
- new flags on many CCB function codes to designate whether they're
non-immediate, use a user-supplied CCB, and can only be passed from
userland programs via the xpt device. Use these flags in the transport
layer and pass driver to categorize CCBs.
- new flag in the transport layer device matching code for device nodes
that indicates whether a device is unconfigured
- bump the CAM version from 0x10 to 0x11
- Change the CAM ioctls to use the version as their group code, so we can
force users to recompile code even when the CCB size doesn't change.
- add + fill in a new value in the path inquiry CCB, base_transfer_speed.
Remove a corresponding field from the cam_sim structure, and add code to
every SIM to set this field to the proper value.
- Fix the set transfer settings code in the transport layer.
scsi_cd.c:
- make some variables volatile instead of just casting them in various
places
- fix a race condition in the changer code
- attach unless we get a "logical unit not supported" error. This should
fix all of the cases where people have devices that return weird errors
when they don't have media in the drive.
scsi_da.c:
- attach unless we get a "logical unit not supported" error
scsi_pass.c:
- for immediate CCBs, just malloc a CCB to send the user request in. This
gets rid of the 'held' count problem in camcontrol tags.
scsi_pass.h:
- change the CAM ioctls to use the CAM version as their group code.
adv driver:
- Allow changing the sync rate and offset separately.
adw driver
- Allow changing the sync rate and offset separately.
aha driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
ahc driver:
- Allow setting offset and sync rate separately
bt driver:
- Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs.
NCR driver:
- Fix the ultra/ultra 2 negotiation bug
- allow setting both the sync rate and offset separately
Other HBA drivers:
- Put code in to set the base_transfer_speed field for
XPT_GET_TRAN_SETTINGS CCBs.
Reviewed by: gibbs, mjacob (isp), imp (aha)
1999-05-06 20:16:39 +00:00
|
|
|
}
|
1998-09-15 06:36:34 +00:00
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
/*
|
|
|
|
|
* Copy the allocated CCB contents back to the malloced CCB
|
|
|
|
|
* so we can give status back to the user when he requests it.
|
|
|
|
|
*/
|
|
|
|
|
bcopy(done_ccb, &io_req->ccb, sizeof(*done_ccb));
|
1998-09-15 06:36:34 +00:00
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
/*
|
|
|
|
|
* Log data/transaction completion with devstat(9).
|
|
|
|
|
*/
|
|
|
|
|
switch (done_ccb->ccb_h.func_code) {
|
|
|
|
|
case XPT_SCSI_IO:
|
|
|
|
|
devstat_end_transaction(softc->device_stats,
|
|
|
|
|
done_ccb->csio.dxfer_len - done_ccb->csio.resid,
|
|
|
|
|
done_ccb->csio.tag_action & 0x3,
|
|
|
|
|
((done_ccb->ccb_h.flags & CAM_DIR_MASK) ==
|
|
|
|
|
CAM_DIR_NONE) ? DEVSTAT_NO_DATA :
|
|
|
|
|
(done_ccb->ccb_h.flags & CAM_DIR_OUT) ?
|
|
|
|
|
DEVSTAT_WRITE : DEVSTAT_READ, NULL,
|
|
|
|
|
&io_req->start_time);
|
|
|
|
|
break;
|
|
|
|
|
case XPT_ATA_IO:
|
|
|
|
|
devstat_end_transaction(softc->device_stats,
|
|
|
|
|
done_ccb->ataio.dxfer_len - done_ccb->ataio.resid,
|
|
|
|
|
done_ccb->ataio.tag_action & 0x3,
|
|
|
|
|
((done_ccb->ccb_h.flags & CAM_DIR_MASK) ==
|
|
|
|
|
CAM_DIR_NONE) ? DEVSTAT_NO_DATA :
|
|
|
|
|
(done_ccb->ccb_h.flags & CAM_DIR_OUT) ?
|
|
|
|
|
DEVSTAT_WRITE : DEVSTAT_READ, NULL,
|
|
|
|
|
&io_req->start_time);
|
|
|
|
|
break;
|
|
|
|
|
case XPT_SMP_IO:
|
|
|
|
|
/*
|
|
|
|
|
* XXX KDM this isn't quite right, but there isn't
|
|
|
|
|
* currently an easy way to represent a bidirectional
|
|
|
|
|
* transfer in devstat. The only way to do it
|
|
|
|
|
* and have the byte counts come out right would
|
|
|
|
|
* mean that we would have to record two
|
|
|
|
|
* transactions, one for the request and one for the
|
|
|
|
|
* response. For now, so that we report something,
|
|
|
|
|
* just treat the entire thing as a read.
|
|
|
|
|
*/
|
|
|
|
|
devstat_end_transaction(softc->device_stats,
|
|
|
|
|
done_ccb->smpio.smp_request_len +
|
|
|
|
|
done_ccb->smpio.smp_response_len,
|
|
|
|
|
DEVSTAT_TAG_SIMPLE, DEVSTAT_READ, NULL,
|
|
|
|
|
&io_req->start_time);
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
devstat_end_transaction(softc->device_stats, 0,
|
|
|
|
|
DEVSTAT_TAG_NONE, DEVSTAT_NO_DATA, NULL,
|
|
|
|
|
&io_req->start_time);
|
|
|
|
|
break;
|
|
|
|
|
}
|
1998-09-15 06:36:34 +00:00
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
/*
|
|
|
|
|
* In the normal case, take the completed I/O off of the
|
|
|
|
|
* active queue and put it on the done queue. Notitfy the
|
|
|
|
|
* user that we have a completed I/O.
|
|
|
|
|
*/
|
|
|
|
|
if ((io_req->flags & PASS_IO_ABANDONED) == 0) {
|
|
|
|
|
TAILQ_REMOVE(&softc->active_queue, io_req, links);
|
|
|
|
|
TAILQ_INSERT_TAIL(&softc->done_queue, io_req, links);
|
|
|
|
|
selwakeuppri(&softc->read_select, PRIBIO);
|
|
|
|
|
KNOTE_LOCKED(&softc->read_select.si_note, 0);
|
|
|
|
|
} else {
|
|
|
|
|
/*
|
|
|
|
|
* In the case of an abandoned I/O (final close
|
|
|
|
|
* without fetching the I/O), take it off of the
|
|
|
|
|
* abandoned queue and free it.
|
|
|
|
|
*/
|
|
|
|
|
TAILQ_REMOVE(&softc->abandoned_queue, io_req, links);
|
|
|
|
|
passiocleanup(softc, io_req);
|
|
|
|
|
uma_zfree(softc->pass_zone, io_req);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Release the done_ccb here, since we may wind up
|
|
|
|
|
* freeing the peripheral when we decrement the
|
|
|
|
|
* reference count below.
|
|
|
|
|
*/
|
|
|
|
|
xpt_release_ccb(done_ccb);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* If the abandoned queue is empty, we can release
|
|
|
|
|
* our reference to the periph since we won't have
|
|
|
|
|
* any more completions coming.
|
|
|
|
|
*/
|
|
|
|
|
if ((TAILQ_EMPTY(&softc->abandoned_queue))
|
|
|
|
|
&& (softc->flags & PASS_FLAG_ABANDONED_REF_SET)) {
|
|
|
|
|
softc->flags &= ~PASS_FLAG_ABANDONED_REF_SET;
|
|
|
|
|
cam_periph_release_locked(periph);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* We have already released the CCB, so we can
|
|
|
|
|
* return.
|
|
|
|
|
*/
|
|
|
|
|
return;
|
|
|
|
|
}
|
1998-09-15 06:36:34 +00:00
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
xpt_release_ccb(done_ccb);
|
1998-09-15 06:36:34 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
passcreatezone(struct cam_periph *periph)
|
1998-09-15 06:36:34 +00:00
|
|
|
{
|
|
|
|
|
struct pass_softc *softc;
|
2013-06-07 00:22:38 +00:00
|
|
|
int error;
|
1998-09-15 06:36:34 +00:00
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
error = 0;
|
1998-09-15 06:36:34 +00:00
|
|
|
softc = (struct pass_softc *)periph->softc;
|
|
|
|
|
|
Add asynchronous command support to the pass(4) driver, and the new
camdd(8) utility.
CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and
completed CCBs may be retrieved via the CAMIOGET ioctl. User
processes can use poll(2) or kevent(2) to get notification when
I/O has completed.
While the existing CAMIOCOMMAND blocking ioctl interface only
supports user virtual data pointers in a CCB (generally only
one per CCB), the new CAMIOQUEUE ioctl supports user virtual and
physical address pointers, as well as user virtual and physical
scatter/gather lists. This allows user applications to have more
flexibility in their data handling operations.
Kernel memory for data transferred via the queued interface is
allocated from the zone allocator in MAXPHYS sized chunks, and user
data is copied in and out. This is likely faster than the
vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in
configurations with many processors (there are more TLB shootdowns
caused by the mapping/unmapping operation) but may not be as fast
as running with unmapped I/O.
The new memory handling model for user requests also allows
applications to send CCBs with request sizes that are larger than
MAXPHYS. The pass(4) driver now limits queued requests to the I/O
size listed by the SIM driver in the maxio field in the Path
Inquiry (XPT_PATH_INQ) CCB.
There are some things things would be good to add:
1. Come up with a way to do unmapped I/O on multiple buffers.
Currently the unmapped I/O interface operates on a struct bio,
which includes only one address and length. It would be nice
to be able to send an unmapped scatter/gather list down to
busdma. This would allow eliminating the copy we currently do
for data.
2. Add an ioctl to list currently outstanding CCBs in the various
queues.
3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do
that.
4. Test physical address support. Virtual pointers and scatter
gather lists have been tested, but I have not yet tested
physical addresses or scatter/gather lists.
5. Investigate multiple queue support. At the moment there is one
queue of commands per pass(4) device. If multiple processes
open the device, they will submit I/O into the same queue and
get events for the same completions. This is probably the right
model for most applications, but it is something that could be
changed later on.
Also, add a new utility, camdd(8) that uses the asynchronous pass(4)
driver interface.
This utility is intended to be a basic data transfer/copy utility,
a simple benchmark utility, and an example of how to use the
asynchronous pass(4) interface.
It can copy data to and from pass(4) devices using any target queue
depth, starting offset and blocksize for the input and ouptut devices.
It currently only supports SCSI devices, but could be easily extended
to support ATA devices.
It can also copy data to and from regular files, block devices, tape
devices, pipes, stdin, and stdout. It does not support queueing
multiple commands to any of those targets, since it uses the standard
read(2)/write(2)/writev(2)/readv(2) system calls.
The I/O is done by two threads, one for the reader and one for the
writer. The reader thread sends completed read requests to the
writer thread in strictly sequential order, even if they complete
out of order. That could be modified later on for random I/O patterns
or slightly out of order I/O.
camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from
the pass(4) driver and also to send request notifications internally.
For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR)
per CAM CCB on the reading side, and a scatter/gather list
(CAM_DATA_SG) on the writing side. In addition to testing both
interfaces, this makes any potential reblocking of I/O easier. No
data is copied between the reader and the writer, but rather the
reader's buffers are split into multiple I/O requests or combined
into a single I/O request depending on the input and output blocksize.
For the file I/O path, camdd(8) also uses a single buffer (read(2),
write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list
(readv(2), writev(2), preadv(2), pwritev(2)) on writes.
Things that would be nice to do for camdd(8) eventually:
1. Add support for I/O pattern generation. Patterns like all
zeros, all ones, LBA-based patterns, random patterns, etc. Right
Now you can always use /dev/zero, /dev/random, etc.
2. Add support for a "sink" mode, so we do only reads with no
writes. Right now, you can use /dev/null.
3. Add support for automatic queue depth probing, so that we can
figure out the right queue depth on the input and output side
for maximum throughput. At the moment it defaults to 6.
4. Add support for SATA device passthrough I/O.
5. Add support for random LBAs and/or lengths on the input and
output sides.
6. Track average per-I/O latency and busy time. The busy time
and latency could also feed in to the automatic queue depth
determination.
sys/cam/scsi/scsi_pass.h:
Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue
and fetch asynchronous CAM CCBs respectively.
Although these ioctls do not have a declared argument, they
both take a union ccb pointer. If we declare a size here,
the ioctl code in sys/kern/sys_generic.c will malloc and free
a buffer for either the CCB or the CCB pointer (depending on
how it is declared). Since we have to keep a copy of the
CCB (which is fairly large) anyway, having the ioctl malloc
and free a CCB for each call is wasteful.
sys/cam/scsi/scsi_pass.c:
Add asynchronous CCB support.
Add two new ioctls, CAMIOQUEUE and CAMIOGET.
CAMIOQUEUE adds a CCB to the incoming queue. The CCB is
executed immediately (and moved to the active queue) if it
is an immediate CCB, but otherwise it will be executed
in passstart() when a CCB is available from the transport layer.
When CCBs are completed (because they are immediate or
passdone() if they are queued), they are put on the done
queue.
If we get the final close on the device before all pending
I/O is complete, all active I/O is moved to the abandoned
queue and we increment the peripheral reference count so
that the peripheral driver instance doesn't go away before
all pending I/O is done.
The new passcreatezone() function is called on the first
call to the CAMIOQUEUE ioctl on a given device to allocate
the UMA zones for I/O requests and S/G list buffers. This
may be good to move off to a taskqueue at some point.
The new passmemsetup() function allocates memory and
scatter/gather lists to hold the user's data, and copies
in any data that needs to be written. For virtual pointers
(CAM_DATA_VADDR), the kernel buffer is malloced from the
new pass(4) driver malloc bucket. For virtual
scatter/gather lists (CAM_DATA_SG), buffers are allocated
from a new per-pass(9) UMA zone in MAXPHYS-sized chunks.
Physical pointers are passed in unchanged. We have support
for up to 16 scatter/gather segments (for the user and
kernel S/G lists) in the default struct pass_io_req, so
requests with longer S/G lists require an extra kernel malloc.
The new passcopysglist() function copies a user scatter/gather
list to a kernel scatter/gather list. The number of elements
in each list may be different, but (obviously) the amount of data
stored has to be identical.
The new passmemdone() function copies data out for the
CAM_DATA_VADDR and CAM_DATA_SG cases.
The new passiocleanup() function restores data pointers in
user CCBs and frees memory.
Add new functions to support kqueue(2)/kevent(2):
passreadfilt() tells kevent whether or not the done
queue is empty.
passkqfilter() adds a knote to our list.
passreadfiltdetach() removes a knote from our list.
Add a new function, passpoll(), for poll(2)/select(2)
to use.
Add devstat(9) support for the queued CCB path.
sys/cam/ata/ata_da.c:
Add support for the BIO_VLIST bio type.
sys/cam/cam_ccb.h:
Add a new enumeration for the xflags field in the CCB header.
(This doesn't change the CCB header, just adds an enumeration to
use.)
sys/cam/cam_xpt.c:
Add a new function, xpt_setup_ccb_flags(), that allows specifying
CCB flags.
sys/cam/cam_xpt.h:
Add a prototype for xpt_setup_ccb_flags().
sys/cam/scsi/scsi_da.c:
Add support for BIO_VLIST.
sys/dev/md/md.c:
Add BIO_VLIST support to md(4).
sys/geom/geom_disk.c:
Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size
limiting code in g_disk_start() a bit.
sys/kern/subr_bus_dma.c:
Change _bus_dmamap_load_vlist() to take a starting offset and
length.
Add a new function, _bus_dmamap_load_pages(), that will load a list
of physical pages starting at an offset.
Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios.
Allow unmapped I/O to start at an offset.
sys/kern/subr_uio.c:
Add two new functions, physcopyin_vlist() and physcopyout_vlist().
sys/pc98/include/bus.h:
Guard kernel-only parts of the pc98 machine/bus.h header with
#ifdef _KERNEL.
This allows userland programs to include <machine/bus.h> to get the
definition of bus_addr_t and bus_size_t.
sys/sys/bio.h:
Add a new bio flag, BIO_VLIST.
sys/sys/uio.h:
Add prototypes for physcopyin_vlist() and physcopyout_vlist().
share/man/man4/pass.4:
Document the CAMIOQUEUE and CAMIOGET ioctls.
usr.sbin/Makefile:
Add camdd.
usr.sbin/camdd/Makefile:
Add a makefile for camdd(8).
usr.sbin/camdd/camdd.8:
Man page for camdd(8).
usr.sbin/camdd/camdd.c:
The new camdd(8) utility.
Sponsored by: Spectra Logic
MFC after: 1 week
2015-12-03 20:54:55 +00:00
|
|
|
cam_periph_assert(periph, MA_OWNED);
|
|
|
|
|
KASSERT(((softc->flags & PASS_FLAG_ZONE_VALID) == 0),
|
|
|
|
|
("%s called when the pass(4) zone is valid!\n", __func__));
|
|
|
|
|
KASSERT((softc->pass_zone == NULL),
|
|
|
|
|
("%s called when the pass(4) zone is allocated!\n", __func__));
|
|
|
|
|
|
|
|
|
|
if ((softc->flags & PASS_FLAG_ZONE_INPROG) == 0) {
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* We're the first context through, so we need to create
|
|
|
|
|
* the pass(4) UMA zone for I/O requests.
|
|
|
|
|
*/
|
|
|
|
|
softc->flags |= PASS_FLAG_ZONE_INPROG;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* uma_zcreate() does a blocking (M_WAITOK) allocation,
|
|
|
|
|
* so we cannot hold a mutex while we call it.
|
|
|
|
|
*/
|
|
|
|
|
cam_periph_unlock(periph);
|
|
|
|
|
|
|
|
|
|
softc->pass_zone = uma_zcreate(softc->zone_name,
|
|
|
|
|
sizeof(struct pass_io_req), NULL, NULL, NULL, NULL,
|
|
|
|
|
/*align*/ 0, /*flags*/ 0);
|
|
|
|
|
|
|
|
|
|
softc->pass_io_zone = uma_zcreate(softc->io_zone_name,
|
|
|
|
|
softc->io_zone_size, NULL, NULL, NULL, NULL,
|
|
|
|
|
/*align*/ 0, /*flags*/ 0);
|
|
|
|
|
|
|
|
|
|
cam_periph_lock(periph);
|
|
|
|
|
|
|
|
|
|
if ((softc->pass_zone == NULL)
|
|
|
|
|
|| (softc->pass_io_zone == NULL)) {
|
|
|
|
|
if (softc->pass_zone == NULL)
|
|
|
|
|
xpt_print(periph->path, "unable to allocate "
|
|
|
|
|
"IO Req UMA zone\n");
|
|
|
|
|
else
|
|
|
|
|
xpt_print(periph->path, "unable to allocate "
|
|
|
|
|
"IO UMA zone\n");
|
|
|
|
|
softc->flags &= ~PASS_FLAG_ZONE_INPROG;
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Set the flags appropriately and notify any other waiters.
|
|
|
|
|
*/
|
|
|
|
|
softc->flags &= PASS_FLAG_ZONE_INPROG;
|
|
|
|
|
softc->flags |= PASS_FLAG_ZONE_VALID;
|
|
|
|
|
wakeup(&softc->pass_zone);
|
|
|
|
|
} else {
|
|
|
|
|
/*
|
|
|
|
|
* In this case, the UMA zone has not yet been created, but
|
|
|
|
|
* another context is in the process of creating it. We
|
|
|
|
|
* need to sleep until the creation is either done or has
|
|
|
|
|
* failed.
|
|
|
|
|
*/
|
|
|
|
|
while ((softc->flags & PASS_FLAG_ZONE_INPROG)
|
|
|
|
|
&& ((softc->flags & PASS_FLAG_ZONE_VALID) == 0)) {
|
|
|
|
|
error = msleep(&softc->pass_zone,
|
|
|
|
|
cam_periph_mtx(periph), PRIBIO,
|
|
|
|
|
"paszon", 0);
|
|
|
|
|
if (error != 0)
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
/*
|
|
|
|
|
* If the zone creation failed, no luck for the user.
|
|
|
|
|
*/
|
|
|
|
|
if ((softc->flags & PASS_FLAG_ZONE_VALID) == 0){
|
|
|
|
|
error = ENOMEM;
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
bailout:
|
|
|
|
|
return (error);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
|
passiocleanup(struct pass_softc *softc, struct pass_io_req *io_req)
|
|
|
|
|
{
|
|
|
|
|
union ccb *ccb;
|
|
|
|
|
u_int8_t **data_ptrs[CAM_PERIPH_MAXMAPS];
|
|
|
|
|
int i, numbufs;
|
|
|
|
|
|
|
|
|
|
ccb = &io_req->ccb;
|
|
|
|
|
|
|
|
|
|
switch (ccb->ccb_h.func_code) {
|
|
|
|
|
case XPT_DEV_MATCH:
|
|
|
|
|
numbufs = min(io_req->num_bufs, 2);
|
|
|
|
|
|
|
|
|
|
if (numbufs == 1) {
|
|
|
|
|
data_ptrs[0] = (u_int8_t **)&ccb->cdm.matches;
|
|
|
|
|
} else {
|
|
|
|
|
data_ptrs[0] = (u_int8_t **)&ccb->cdm.patterns;
|
|
|
|
|
data_ptrs[1] = (u_int8_t **)&ccb->cdm.matches;
|
|
|
|
|
}
|
|
|
|
|
break;
|
|
|
|
|
case XPT_SCSI_IO:
|
|
|
|
|
case XPT_CONT_TARGET_IO:
|
|
|
|
|
data_ptrs[0] = &ccb->csio.data_ptr;
|
|
|
|
|
numbufs = min(io_req->num_bufs, 1);
|
|
|
|
|
break;
|
|
|
|
|
case XPT_ATA_IO:
|
|
|
|
|
data_ptrs[0] = &ccb->ataio.data_ptr;
|
|
|
|
|
numbufs = min(io_req->num_bufs, 1);
|
|
|
|
|
break;
|
|
|
|
|
case XPT_SMP_IO:
|
|
|
|
|
numbufs = min(io_req->num_bufs, 2);
|
|
|
|
|
data_ptrs[0] = &ccb->smpio.smp_request;
|
|
|
|
|
data_ptrs[1] = &ccb->smpio.smp_response;
|
|
|
|
|
break;
|
|
|
|
|
case XPT_DEV_ADVINFO:
|
|
|
|
|
numbufs = min(io_req->num_bufs, 1);
|
|
|
|
|
data_ptrs[0] = (uint8_t **)&ccb->cdai.buf;
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
/* allow ourselves to be swapped once again */
|
|
|
|
|
return;
|
|
|
|
|
break; /* NOTREACHED */
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (io_req->flags & PASS_IO_USER_SEG_MALLOC) {
|
|
|
|
|
free(io_req->user_segptr, M_SCSIPASS);
|
|
|
|
|
io_req->user_segptr = NULL;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* We only want to free memory we malloced.
|
|
|
|
|
*/
|
|
|
|
|
if (io_req->data_flags == CAM_DATA_VADDR) {
|
|
|
|
|
for (i = 0; i < io_req->num_bufs; i++) {
|
|
|
|
|
if (io_req->kern_bufs[i] == NULL)
|
|
|
|
|
continue;
|
|
|
|
|
|
|
|
|
|
free(io_req->kern_bufs[i], M_SCSIPASS);
|
|
|
|
|
io_req->kern_bufs[i] = NULL;
|
|
|
|
|
}
|
|
|
|
|
} else if (io_req->data_flags == CAM_DATA_SG) {
|
|
|
|
|
for (i = 0; i < io_req->num_kern_segs; i++) {
|
|
|
|
|
if ((uint8_t *)(uintptr_t)
|
|
|
|
|
io_req->kern_segptr[i].ds_addr == NULL)
|
|
|
|
|
continue;
|
|
|
|
|
|
|
|
|
|
uma_zfree(softc->pass_io_zone, (uint8_t *)(uintptr_t)
|
|
|
|
|
io_req->kern_segptr[i].ds_addr);
|
|
|
|
|
io_req->kern_segptr[i].ds_addr = 0;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (io_req->flags & PASS_IO_KERN_SEG_MALLOC) {
|
|
|
|
|
free(io_req->kern_segptr, M_SCSIPASS);
|
|
|
|
|
io_req->kern_segptr = NULL;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (io_req->data_flags != CAM_DATA_PADDR) {
|
|
|
|
|
for (i = 0; i < numbufs; i++) {
|
|
|
|
|
/*
|
|
|
|
|
* Restore the user's buffer pointers to their
|
|
|
|
|
* previous values.
|
|
|
|
|
*/
|
|
|
|
|
if (io_req->user_bufs[i] != NULL)
|
|
|
|
|
*data_ptrs[i] = io_req->user_bufs[i];
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int
|
|
|
|
|
passcopysglist(struct cam_periph *periph, struct pass_io_req *io_req,
|
|
|
|
|
ccb_flags direction)
|
|
|
|
|
{
|
|
|
|
|
bus_size_t kern_watermark, user_watermark, len_copied, len_to_copy;
|
|
|
|
|
bus_dma_segment_t *user_sglist, *kern_sglist;
|
|
|
|
|
int i, j, error;
|
|
|
|
|
|
|
|
|
|
error = 0;
|
|
|
|
|
kern_watermark = 0;
|
|
|
|
|
user_watermark = 0;
|
|
|
|
|
len_to_copy = 0;
|
|
|
|
|
len_copied = 0;
|
|
|
|
|
user_sglist = io_req->user_segptr;
|
|
|
|
|
kern_sglist = io_req->kern_segptr;
|
|
|
|
|
|
|
|
|
|
for (i = 0, j = 0; i < io_req->num_user_segs &&
|
|
|
|
|
j < io_req->num_kern_segs;) {
|
|
|
|
|
uint8_t *user_ptr, *kern_ptr;
|
|
|
|
|
|
|
|
|
|
len_to_copy = min(user_sglist[i].ds_len -user_watermark,
|
|
|
|
|
kern_sglist[j].ds_len - kern_watermark);
|
|
|
|
|
|
|
|
|
|
user_ptr = (uint8_t *)(uintptr_t)user_sglist[i].ds_addr;
|
|
|
|
|
user_ptr = user_ptr + user_watermark;
|
|
|
|
|
kern_ptr = (uint8_t *)(uintptr_t)kern_sglist[j].ds_addr;
|
|
|
|
|
kern_ptr = kern_ptr + kern_watermark;
|
|
|
|
|
|
|
|
|
|
user_watermark += len_to_copy;
|
|
|
|
|
kern_watermark += len_to_copy;
|
|
|
|
|
|
|
|
|
|
if (!useracc(user_ptr, len_to_copy,
|
|
|
|
|
(direction == CAM_DIR_IN) ? VM_PROT_WRITE : VM_PROT_READ)) {
|
|
|
|
|
xpt_print(periph->path, "%s: unable to access user "
|
|
|
|
|
"S/G list element %p len %zu\n", __func__,
|
|
|
|
|
user_ptr, len_to_copy);
|
|
|
|
|
error = EFAULT;
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (direction == CAM_DIR_IN) {
|
|
|
|
|
error = copyout(kern_ptr, user_ptr, len_to_copy);
|
|
|
|
|
if (error != 0) {
|
|
|
|
|
xpt_print(periph->path, "%s: copyout of %u "
|
|
|
|
|
"bytes from %p to %p failed with "
|
|
|
|
|
"error %d\n", __func__, len_to_copy,
|
|
|
|
|
kern_ptr, user_ptr, error);
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
} else {
|
|
|
|
|
error = copyin(user_ptr, kern_ptr, len_to_copy);
|
|
|
|
|
if (error != 0) {
|
|
|
|
|
xpt_print(periph->path, "%s: copyin of %u "
|
|
|
|
|
"bytes from %p to %p failed with "
|
|
|
|
|
"error %d\n", __func__, len_to_copy,
|
|
|
|
|
user_ptr, kern_ptr, error);
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
len_copied += len_to_copy;
|
|
|
|
|
|
|
|
|
|
if (user_sglist[i].ds_len == user_watermark) {
|
|
|
|
|
i++;
|
|
|
|
|
user_watermark = 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (kern_sglist[j].ds_len == kern_watermark) {
|
|
|
|
|
j++;
|
|
|
|
|
kern_watermark = 0;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
bailout:
|
|
|
|
|
|
|
|
|
|
return (error);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int
|
|
|
|
|
passmemsetup(struct cam_periph *periph, struct pass_io_req *io_req)
|
|
|
|
|
{
|
|
|
|
|
union ccb *ccb;
|
|
|
|
|
struct pass_softc *softc;
|
|
|
|
|
int numbufs, i;
|
|
|
|
|
uint8_t **data_ptrs[CAM_PERIPH_MAXMAPS];
|
|
|
|
|
uint32_t lengths[CAM_PERIPH_MAXMAPS];
|
|
|
|
|
uint32_t dirs[CAM_PERIPH_MAXMAPS];
|
|
|
|
|
uint32_t num_segs;
|
|
|
|
|
uint16_t *seg_cnt_ptr;
|
|
|
|
|
size_t maxmap;
|
|
|
|
|
int error;
|
|
|
|
|
|
|
|
|
|
cam_periph_assert(periph, MA_NOTOWNED);
|
|
|
|
|
|
|
|
|
|
softc = periph->softc;
|
|
|
|
|
|
|
|
|
|
error = 0;
|
|
|
|
|
ccb = &io_req->ccb;
|
|
|
|
|
maxmap = 0;
|
|
|
|
|
num_segs = 0;
|
|
|
|
|
seg_cnt_ptr = NULL;
|
|
|
|
|
|
|
|
|
|
switch(ccb->ccb_h.func_code) {
|
|
|
|
|
case XPT_DEV_MATCH:
|
|
|
|
|
if (ccb->cdm.match_buf_len == 0) {
|
|
|
|
|
printf("%s: invalid match buffer length 0\n", __func__);
|
|
|
|
|
return(EINVAL);
|
|
|
|
|
}
|
|
|
|
|
if (ccb->cdm.pattern_buf_len > 0) {
|
|
|
|
|
data_ptrs[0] = (u_int8_t **)&ccb->cdm.patterns;
|
|
|
|
|
lengths[0] = ccb->cdm.pattern_buf_len;
|
|
|
|
|
dirs[0] = CAM_DIR_OUT;
|
|
|
|
|
data_ptrs[1] = (u_int8_t **)&ccb->cdm.matches;
|
|
|
|
|
lengths[1] = ccb->cdm.match_buf_len;
|
|
|
|
|
dirs[1] = CAM_DIR_IN;
|
|
|
|
|
numbufs = 2;
|
|
|
|
|
} else {
|
|
|
|
|
data_ptrs[0] = (u_int8_t **)&ccb->cdm.matches;
|
|
|
|
|
lengths[0] = ccb->cdm.match_buf_len;
|
|
|
|
|
dirs[0] = CAM_DIR_IN;
|
|
|
|
|
numbufs = 1;
|
|
|
|
|
}
|
|
|
|
|
io_req->data_flags = CAM_DATA_VADDR;
|
|
|
|
|
break;
|
|
|
|
|
case XPT_SCSI_IO:
|
|
|
|
|
case XPT_CONT_TARGET_IO:
|
|
|
|
|
if ((ccb->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_NONE)
|
|
|
|
|
return(0);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* The user shouldn't be able to supply a bio.
|
|
|
|
|
*/
|
|
|
|
|
if ((ccb->ccb_h.flags & CAM_DATA_MASK) == CAM_DATA_BIO)
|
|
|
|
|
return (EINVAL);
|
|
|
|
|
|
|
|
|
|
io_req->data_flags = ccb->ccb_h.flags & CAM_DATA_MASK;
|
|
|
|
|
|
|
|
|
|
data_ptrs[0] = &ccb->csio.data_ptr;
|
|
|
|
|
lengths[0] = ccb->csio.dxfer_len;
|
|
|
|
|
dirs[0] = ccb->ccb_h.flags & CAM_DIR_MASK;
|
|
|
|
|
num_segs = ccb->csio.sglist_cnt;
|
|
|
|
|
seg_cnt_ptr = &ccb->csio.sglist_cnt;
|
|
|
|
|
numbufs = 1;
|
|
|
|
|
maxmap = softc->maxio;
|
|
|
|
|
break;
|
|
|
|
|
case XPT_ATA_IO:
|
|
|
|
|
if ((ccb->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_NONE)
|
|
|
|
|
return(0);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* We only support a single virtual address for ATA I/O.
|
|
|
|
|
*/
|
|
|
|
|
if ((ccb->ccb_h.flags & CAM_DATA_MASK) != CAM_DATA_VADDR)
|
|
|
|
|
return (EINVAL);
|
|
|
|
|
|
|
|
|
|
io_req->data_flags = CAM_DATA_VADDR;
|
|
|
|
|
|
|
|
|
|
data_ptrs[0] = &ccb->ataio.data_ptr;
|
|
|
|
|
lengths[0] = ccb->ataio.dxfer_len;
|
|
|
|
|
dirs[0] = ccb->ccb_h.flags & CAM_DIR_MASK;
|
|
|
|
|
numbufs = 1;
|
|
|
|
|
maxmap = softc->maxio;
|
|
|
|
|
break;
|
|
|
|
|
case XPT_SMP_IO:
|
|
|
|
|
io_req->data_flags = CAM_DATA_VADDR;
|
|
|
|
|
|
|
|
|
|
data_ptrs[0] = &ccb->smpio.smp_request;
|
|
|
|
|
lengths[0] = ccb->smpio.smp_request_len;
|
|
|
|
|
dirs[0] = CAM_DIR_OUT;
|
|
|
|
|
data_ptrs[1] = &ccb->smpio.smp_response;
|
|
|
|
|
lengths[1] = ccb->smpio.smp_response_len;
|
|
|
|
|
dirs[1] = CAM_DIR_IN;
|
|
|
|
|
numbufs = 2;
|
|
|
|
|
maxmap = softc->maxio;
|
|
|
|
|
break;
|
|
|
|
|
case XPT_DEV_ADVINFO:
|
|
|
|
|
if (ccb->cdai.bufsiz == 0)
|
|
|
|
|
return (0);
|
|
|
|
|
|
|
|
|
|
io_req->data_flags = CAM_DATA_VADDR;
|
|
|
|
|
|
|
|
|
|
data_ptrs[0] = (uint8_t **)&ccb->cdai.buf;
|
|
|
|
|
lengths[0] = ccb->cdai.bufsiz;
|
|
|
|
|
dirs[0] = CAM_DIR_IN;
|
|
|
|
|
numbufs = 1;
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
return(EINVAL);
|
|
|
|
|
break; /* NOTREACHED */
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
io_req->num_bufs = numbufs;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* If there is a maximum, check to make sure that the user's
|
|
|
|
|
* request fits within the limit. In general, we should only have
|
|
|
|
|
* a maximum length for requests that go to hardware. Otherwise it
|
|
|
|
|
* is whatever we're able to malloc.
|
|
|
|
|
*/
|
|
|
|
|
for (i = 0; i < numbufs; i++) {
|
|
|
|
|
io_req->user_bufs[i] = *data_ptrs[i];
|
|
|
|
|
io_req->dirs[i] = dirs[i];
|
|
|
|
|
io_req->lengths[i] = lengths[i];
|
|
|
|
|
|
|
|
|
|
if (maxmap == 0)
|
|
|
|
|
continue;
|
|
|
|
|
|
|
|
|
|
if (lengths[i] <= maxmap)
|
|
|
|
|
continue;
|
|
|
|
|
|
|
|
|
|
xpt_print(periph->path, "%s: data length %u > max allowed %u "
|
|
|
|
|
"bytes\n", __func__, lengths[i], maxmap);
|
|
|
|
|
error = EINVAL;
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
switch (io_req->data_flags) {
|
|
|
|
|
case CAM_DATA_VADDR:
|
|
|
|
|
/* Map or copy the buffer into kernel address space */
|
|
|
|
|
for (i = 0; i < numbufs; i++) {
|
|
|
|
|
uint8_t *tmp_buf;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* If for some reason no length is specified, we
|
|
|
|
|
* don't need to allocate anything.
|
|
|
|
|
*/
|
|
|
|
|
if (io_req->lengths[i] == 0)
|
|
|
|
|
continue;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Make sure that the user's buffer is accessible
|
|
|
|
|
* to that process.
|
|
|
|
|
*/
|
|
|
|
|
if (!useracc(io_req->user_bufs[i], io_req->lengths[i],
|
|
|
|
|
(io_req->dirs[i] == CAM_DIR_IN) ? VM_PROT_WRITE :
|
|
|
|
|
VM_PROT_READ)) {
|
|
|
|
|
xpt_print(periph->path, "%s: user address %p "
|
|
|
|
|
"length %u is not accessible\n", __func__,
|
|
|
|
|
io_req->user_bufs[i], io_req->lengths[i]);
|
|
|
|
|
error = EFAULT;
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
tmp_buf = malloc(lengths[i], M_SCSIPASS,
|
|
|
|
|
M_WAITOK | M_ZERO);
|
|
|
|
|
io_req->kern_bufs[i] = tmp_buf;
|
|
|
|
|
*data_ptrs[i] = tmp_buf;
|
|
|
|
|
|
|
|
|
|
#if 0
|
|
|
|
|
xpt_print(periph->path, "%s: malloced %p len %u, user "
|
|
|
|
|
"buffer %p, operation: %s\n", __func__,
|
|
|
|
|
tmp_buf, lengths[i], io_req->user_bufs[i],
|
|
|
|
|
(dirs[i] == CAM_DIR_IN) ? "read" : "write");
|
|
|
|
|
#endif
|
|
|
|
|
/*
|
|
|
|
|
* We only need to copy in if the user is writing.
|
|
|
|
|
*/
|
|
|
|
|
if (dirs[i] != CAM_DIR_OUT)
|
|
|
|
|
continue;
|
|
|
|
|
|
|
|
|
|
error = copyin(io_req->user_bufs[i],
|
|
|
|
|
io_req->kern_bufs[i], lengths[i]);
|
|
|
|
|
if (error != 0) {
|
|
|
|
|
xpt_print(periph->path, "%s: copy of user "
|
|
|
|
|
"buffer from %p to %p failed with "
|
|
|
|
|
"error %d\n", __func__,
|
|
|
|
|
io_req->user_bufs[i],
|
|
|
|
|
io_req->kern_bufs[i], error);
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
break;
|
|
|
|
|
case CAM_DATA_PADDR:
|
|
|
|
|
/* Pass down the pointer as-is */
|
|
|
|
|
break;
|
|
|
|
|
case CAM_DATA_SG: {
|
|
|
|
|
size_t sg_length, size_to_go, alloc_size;
|
|
|
|
|
uint32_t num_segs_needed;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Copy the user S/G list in, and then copy in the
|
|
|
|
|
* individual segments.
|
|
|
|
|
*/
|
|
|
|
|
/*
|
|
|
|
|
* We shouldn't see this, but check just in case.
|
|
|
|
|
*/
|
|
|
|
|
if (numbufs != 1) {
|
|
|
|
|
xpt_print(periph->path, "%s: cannot currently handle "
|
|
|
|
|
"more than one S/G list per CCB\n", __func__);
|
|
|
|
|
error = EINVAL;
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* We have to have at least one segment.
|
|
|
|
|
*/
|
|
|
|
|
if (num_segs == 0) {
|
|
|
|
|
xpt_print(periph->path, "%s: CAM_DATA_SG flag set, "
|
|
|
|
|
"but sglist_cnt=0!\n", __func__);
|
|
|
|
|
error = EINVAL;
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Make sure the user specified the total length and didn't
|
|
|
|
|
* just leave it to us to decode the S/G list.
|
|
|
|
|
*/
|
|
|
|
|
if (lengths[0] == 0) {
|
|
|
|
|
xpt_print(periph->path, "%s: no dxfer_len specified, "
|
|
|
|
|
"but CAM_DATA_SG flag is set!\n", __func__);
|
|
|
|
|
error = EINVAL;
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* We allocate buffers in io_zone_size increments for an
|
|
|
|
|
* S/G list. This will generally be MAXPHYS.
|
|
|
|
|
*/
|
|
|
|
|
if (lengths[0] <= softc->io_zone_size)
|
|
|
|
|
num_segs_needed = 1;
|
|
|
|
|
else {
|
|
|
|
|
num_segs_needed = lengths[0] / softc->io_zone_size;
|
|
|
|
|
if ((lengths[0] % softc->io_zone_size) != 0)
|
|
|
|
|
num_segs_needed++;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Figure out the size of the S/G list */
|
|
|
|
|
sg_length = num_segs * sizeof(bus_dma_segment_t);
|
|
|
|
|
io_req->num_user_segs = num_segs;
|
|
|
|
|
io_req->num_kern_segs = num_segs_needed;
|
|
|
|
|
|
|
|
|
|
/* Save the user's S/G list pointer for later restoration */
|
|
|
|
|
io_req->user_bufs[0] = *data_ptrs[0];
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* If we have enough segments allocated by default to handle
|
|
|
|
|
* the length of the user's S/G list,
|
|
|
|
|
*/
|
|
|
|
|
if (num_segs > PASS_MAX_SEGS) {
|
|
|
|
|
io_req->user_segptr = malloc(sizeof(bus_dma_segment_t) *
|
|
|
|
|
num_segs, M_SCSIPASS, M_WAITOK | M_ZERO);
|
|
|
|
|
io_req->flags |= PASS_IO_USER_SEG_MALLOC;
|
|
|
|
|
} else
|
|
|
|
|
io_req->user_segptr = io_req->user_segs;
|
|
|
|
|
|
|
|
|
|
if (!useracc(*data_ptrs[0], sg_length, VM_PROT_READ)) {
|
|
|
|
|
xpt_print(periph->path, "%s: unable to access user "
|
|
|
|
|
"S/G list at %p\n", __func__, *data_ptrs[0]);
|
|
|
|
|
error = EFAULT;
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
error = copyin(*data_ptrs[0], io_req->user_segptr, sg_length);
|
|
|
|
|
if (error != 0) {
|
|
|
|
|
xpt_print(periph->path, "%s: copy of user S/G list "
|
|
|
|
|
"from %p to %p failed with error %d\n",
|
|
|
|
|
__func__, *data_ptrs[0], io_req->user_segptr,
|
|
|
|
|
error);
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (num_segs_needed > PASS_MAX_SEGS) {
|
|
|
|
|
io_req->kern_segptr = malloc(sizeof(bus_dma_segment_t) *
|
|
|
|
|
num_segs_needed, M_SCSIPASS, M_WAITOK | M_ZERO);
|
|
|
|
|
io_req->flags |= PASS_IO_KERN_SEG_MALLOC;
|
|
|
|
|
} else {
|
|
|
|
|
io_req->kern_segptr = io_req->kern_segs;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Allocate the kernel S/G list.
|
|
|
|
|
*/
|
|
|
|
|
for (size_to_go = lengths[0], i = 0;
|
|
|
|
|
size_to_go > 0 && i < num_segs_needed;
|
|
|
|
|
i++, size_to_go -= alloc_size) {
|
|
|
|
|
uint8_t *kern_ptr;
|
|
|
|
|
|
|
|
|
|
alloc_size = min(size_to_go, softc->io_zone_size);
|
|
|
|
|
kern_ptr = uma_zalloc(softc->pass_io_zone, M_WAITOK);
|
|
|
|
|
io_req->kern_segptr[i].ds_addr =
|
|
|
|
|
(bus_addr_t)(uintptr_t)kern_ptr;
|
|
|
|
|
io_req->kern_segptr[i].ds_len = alloc_size;
|
|
|
|
|
}
|
|
|
|
|
if (size_to_go > 0) {
|
|
|
|
|
printf("%s: size_to_go = %zu, software error!\n",
|
|
|
|
|
__func__, size_to_go);
|
|
|
|
|
error = EINVAL;
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
*data_ptrs[0] = (uint8_t *)io_req->kern_segptr;
|
|
|
|
|
*seg_cnt_ptr = io_req->num_kern_segs;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* We only need to copy data here if the user is writing.
|
|
|
|
|
*/
|
|
|
|
|
if (dirs[0] == CAM_DIR_OUT)
|
|
|
|
|
error = passcopysglist(periph, io_req, dirs[0]);
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
case CAM_DATA_SG_PADDR: {
|
|
|
|
|
size_t sg_length;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* We shouldn't see this, but check just in case.
|
|
|
|
|
*/
|
|
|
|
|
if (numbufs != 1) {
|
|
|
|
|
printf("%s: cannot currently handle more than one "
|
|
|
|
|
"S/G list per CCB\n", __func__);
|
|
|
|
|
error = EINVAL;
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* We have to have at least one segment.
|
|
|
|
|
*/
|
|
|
|
|
if (num_segs == 0) {
|
|
|
|
|
xpt_print(periph->path, "%s: CAM_DATA_SG_PADDR flag "
|
|
|
|
|
"set, but sglist_cnt=0!\n", __func__);
|
|
|
|
|
error = EINVAL;
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Make sure the user specified the total length and didn't
|
|
|
|
|
* just leave it to us to decode the S/G list.
|
|
|
|
|
*/
|
|
|
|
|
if (lengths[0] == 0) {
|
|
|
|
|
xpt_print(periph->path, "%s: no dxfer_len specified, "
|
|
|
|
|
"but CAM_DATA_SG flag is set!\n", __func__);
|
|
|
|
|
error = EINVAL;
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Figure out the size of the S/G list */
|
|
|
|
|
sg_length = num_segs * sizeof(bus_dma_segment_t);
|
|
|
|
|
io_req->num_user_segs = num_segs;
|
|
|
|
|
io_req->num_kern_segs = io_req->num_user_segs;
|
|
|
|
|
|
|
|
|
|
/* Save the user's S/G list pointer for later restoration */
|
|
|
|
|
io_req->user_bufs[0] = *data_ptrs[0];
|
|
|
|
|
|
|
|
|
|
if (num_segs > PASS_MAX_SEGS) {
|
|
|
|
|
io_req->user_segptr = malloc(sizeof(bus_dma_segment_t) *
|
|
|
|
|
num_segs, M_SCSIPASS, M_WAITOK | M_ZERO);
|
|
|
|
|
io_req->flags |= PASS_IO_USER_SEG_MALLOC;
|
|
|
|
|
} else
|
|
|
|
|
io_req->user_segptr = io_req->user_segs;
|
|
|
|
|
|
|
|
|
|
io_req->kern_segptr = io_req->user_segptr;
|
|
|
|
|
|
|
|
|
|
error = copyin(*data_ptrs[0], io_req->user_segptr, sg_length);
|
|
|
|
|
if (error != 0) {
|
|
|
|
|
xpt_print(periph->path, "%s: copy of user S/G list "
|
|
|
|
|
"from %p to %p failed with error %d\n",
|
|
|
|
|
__func__, *data_ptrs[0], io_req->user_segptr,
|
|
|
|
|
error);
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
default:
|
|
|
|
|
case CAM_DATA_BIO:
|
|
|
|
|
/*
|
|
|
|
|
* A user shouldn't be attaching a bio to the CCB. It
|
|
|
|
|
* isn't a user-accessible structure.
|
|
|
|
|
*/
|
|
|
|
|
error = EINVAL;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
bailout:
|
|
|
|
|
if (error != 0)
|
|
|
|
|
passiocleanup(softc, io_req);
|
|
|
|
|
|
|
|
|
|
return (error);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int
|
|
|
|
|
passmemdone(struct cam_periph *periph, struct pass_io_req *io_req)
|
|
|
|
|
{
|
|
|
|
|
struct pass_softc *softc;
|
|
|
|
|
union ccb *ccb;
|
|
|
|
|
int error;
|
|
|
|
|
int i;
|
|
|
|
|
|
|
|
|
|
error = 0;
|
|
|
|
|
softc = (struct pass_softc *)periph->softc;
|
|
|
|
|
ccb = &io_req->ccb;
|
|
|
|
|
|
|
|
|
|
switch (io_req->data_flags) {
|
|
|
|
|
case CAM_DATA_VADDR:
|
|
|
|
|
/*
|
|
|
|
|
* Copy back to the user buffer if this was a read.
|
|
|
|
|
*/
|
|
|
|
|
for (i = 0; i < io_req->num_bufs; i++) {
|
|
|
|
|
if (io_req->dirs[i] != CAM_DIR_IN)
|
|
|
|
|
continue;
|
|
|
|
|
|
|
|
|
|
error = copyout(io_req->kern_bufs[i],
|
|
|
|
|
io_req->user_bufs[i], io_req->lengths[i]);
|
|
|
|
|
if (error != 0) {
|
|
|
|
|
xpt_print(periph->path, "Unable to copy %u "
|
|
|
|
|
"bytes from %p to user address %p\n",
|
|
|
|
|
io_req->lengths[i],
|
|
|
|
|
io_req->kern_bufs[i],
|
|
|
|
|
io_req->user_bufs[i]);
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
}
|
|
|
|
|
break;
|
|
|
|
|
case CAM_DATA_PADDR:
|
|
|
|
|
/* Do nothing. The pointer is a physical address already */
|
|
|
|
|
break;
|
|
|
|
|
case CAM_DATA_SG:
|
|
|
|
|
/*
|
|
|
|
|
* Copy back to the user buffer if this was a read.
|
|
|
|
|
* Restore the user's S/G list buffer pointer.
|
|
|
|
|
*/
|
|
|
|
|
if (io_req->dirs[0] == CAM_DIR_IN)
|
|
|
|
|
error = passcopysglist(periph, io_req, io_req->dirs[0]);
|
|
|
|
|
break;
|
|
|
|
|
case CAM_DATA_SG_PADDR:
|
|
|
|
|
/*
|
|
|
|
|
* Restore the user's S/G list buffer pointer. No need to
|
|
|
|
|
* copy.
|
|
|
|
|
*/
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
case CAM_DATA_BIO:
|
|
|
|
|
error = EINVAL;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
bailout:
|
|
|
|
|
/*
|
|
|
|
|
* Reset the user's pointers to their original values and free
|
|
|
|
|
* allocated memory.
|
|
|
|
|
*/
|
|
|
|
|
passiocleanup(softc, io_req);
|
|
|
|
|
|
|
|
|
|
return (error);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int
|
|
|
|
|
passioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flag, struct thread *td)
|
|
|
|
|
{
|
|
|
|
|
int error;
|
|
|
|
|
|
|
|
|
|
if ((error = passdoioctl(dev, cmd, addr, flag, td)) == ENOTTY) {
|
|
|
|
|
error = cam_compat_ioctl(dev, cmd, addr, flag, td, passdoioctl);
|
|
|
|
|
}
|
|
|
|
|
return (error);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int
|
|
|
|
|
passdoioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flag, struct thread *td)
|
|
|
|
|
{
|
|
|
|
|
struct cam_periph *periph;
|
|
|
|
|
struct pass_softc *softc;
|
|
|
|
|
int error;
|
|
|
|
|
uint32_t priority;
|
|
|
|
|
|
|
|
|
|
periph = (struct cam_periph *)dev->si_drv1;
|
|
|
|
|
cam_periph_lock(periph);
|
|
|
|
|
softc = (struct pass_softc *)periph->softc;
|
|
|
|
|
|
|
|
|
|
error = 0;
|
|
|
|
|
|
|
|
|
|
switch (cmd) {
|
|
|
|
|
|
|
|
|
|
case CAMIOCOMMAND:
|
|
|
|
|
{
|
|
|
|
|
union ccb *inccb;
|
|
|
|
|
union ccb *ccb;
|
|
|
|
|
int ccb_malloced;
|
|
|
|
|
|
|
|
|
|
inccb = (union ccb *)addr;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Some CCB types, like scan bus and scan lun can only go
|
|
|
|
|
* through the transport layer device.
|
|
|
|
|
*/
|
|
|
|
|
if (inccb->ccb_h.func_code & XPT_FC_XPT_ONLY) {
|
|
|
|
|
xpt_print(periph->path, "CCB function code %#x is "
|
|
|
|
|
"restricted to the XPT device\n",
|
|
|
|
|
inccb->ccb_h.func_code);
|
|
|
|
|
error = ENODEV;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Compatibility for RL/priority-unaware code. */
|
|
|
|
|
priority = inccb->ccb_h.pinfo.priority;
|
|
|
|
|
if (priority <= CAM_PRIORITY_OOB)
|
|
|
|
|
priority += CAM_PRIORITY_OOB + 1;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Non-immediate CCBs need a CCB from the per-device pool
|
|
|
|
|
* of CCBs, which is scheduled by the transport layer.
|
|
|
|
|
* Immediate CCBs and user-supplied CCBs should just be
|
|
|
|
|
* malloced.
|
|
|
|
|
*/
|
|
|
|
|
if ((inccb->ccb_h.func_code & XPT_FC_QUEUED)
|
|
|
|
|
&& ((inccb->ccb_h.func_code & XPT_FC_USER_CCB) == 0)) {
|
|
|
|
|
ccb = cam_periph_getccb(periph, priority);
|
|
|
|
|
ccb_malloced = 0;
|
|
|
|
|
} else {
|
|
|
|
|
ccb = xpt_alloc_ccb_nowait();
|
|
|
|
|
|
|
|
|
|
if (ccb != NULL)
|
|
|
|
|
xpt_setup_ccb(&ccb->ccb_h, periph->path,
|
|
|
|
|
priority);
|
|
|
|
|
ccb_malloced = 1;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (ccb == NULL) {
|
|
|
|
|
xpt_print(periph->path, "unable to allocate CCB\n");
|
|
|
|
|
error = ENOMEM;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
error = passsendccb(periph, ccb, inccb);
|
|
|
|
|
|
|
|
|
|
if (ccb_malloced)
|
|
|
|
|
xpt_free_ccb(ccb);
|
|
|
|
|
else
|
|
|
|
|
xpt_release_ccb(ccb);
|
|
|
|
|
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
case CAMIOQUEUE:
|
|
|
|
|
{
|
|
|
|
|
struct pass_io_req *io_req;
|
|
|
|
|
union ccb **user_ccb, *ccb;
|
|
|
|
|
xpt_opcode fc;
|
|
|
|
|
|
|
|
|
|
if ((softc->flags & PASS_FLAG_ZONE_VALID) == 0) {
|
|
|
|
|
error = passcreatezone(periph);
|
|
|
|
|
if (error != 0)
|
|
|
|
|
goto bailout;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* We're going to do a blocking allocation for this I/O
|
|
|
|
|
* request, so we have to drop the lock.
|
|
|
|
|
*/
|
|
|
|
|
cam_periph_unlock(periph);
|
|
|
|
|
|
|
|
|
|
io_req = uma_zalloc(softc->pass_zone, M_WAITOK | M_ZERO);
|
|
|
|
|
ccb = &io_req->ccb;
|
|
|
|
|
user_ccb = (union ccb **)addr;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Unlike the CAMIOCOMMAND ioctl above, we only have a
|
|
|
|
|
* pointer to the user's CCB, so we have to copy the whole
|
|
|
|
|
* thing in to a buffer we have allocated (above) instead
|
|
|
|
|
* of allowing the ioctl code to malloc a buffer and copy
|
|
|
|
|
* it in.
|
|
|
|
|
*
|
|
|
|
|
* This is an advantage for this asynchronous interface,
|
|
|
|
|
* since we don't want the memory to get freed while the
|
|
|
|
|
* CCB is outstanding.
|
|
|
|
|
*/
|
|
|
|
|
#if 0
|
|
|
|
|
xpt_print(periph->path, "Copying user CCB %p to "
|
|
|
|
|
"kernel address %p\n", *user_ccb, ccb);
|
|
|
|
|
#endif
|
|
|
|
|
error = copyin(*user_ccb, ccb, sizeof(*ccb));
|
|
|
|
|
if (error != 0) {
|
|
|
|
|
xpt_print(periph->path, "Copy of user CCB %p to "
|
|
|
|
|
"kernel address %p failed with error %d\n",
|
|
|
|
|
*user_ccb, ccb, error);
|
|
|
|
|
uma_zfree(softc->pass_zone, io_req);
|
|
|
|
|
cam_periph_lock(periph);
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Some CCB types, like scan bus and scan lun can only go
|
|
|
|
|
* through the transport layer device.
|
|
|
|
|
*/
|
|
|
|
|
if (ccb->ccb_h.func_code & XPT_FC_XPT_ONLY) {
|
|
|
|
|
xpt_print(periph->path, "CCB function code %#x is "
|
|
|
|
|
"restricted to the XPT device\n",
|
|
|
|
|
ccb->ccb_h.func_code);
|
|
|
|
|
uma_zfree(softc->pass_zone, io_req);
|
|
|
|
|
cam_periph_lock(periph);
|
|
|
|
|
error = ENODEV;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Save the user's CCB pointer as well as his linked list
|
|
|
|
|
* pointers and peripheral private area so that we can
|
|
|
|
|
* restore these later.
|
|
|
|
|
*/
|
|
|
|
|
io_req->user_ccb_ptr = *user_ccb;
|
|
|
|
|
io_req->user_periph_links = ccb->ccb_h.periph_links;
|
|
|
|
|
io_req->user_periph_priv = ccb->ccb_h.periph_priv;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Now that we've saved the user's values, we can set our
|
|
|
|
|
* own peripheral private entry.
|
|
|
|
|
*/
|
|
|
|
|
ccb->ccb_h.ccb_ioreq = io_req;
|
|
|
|
|
|
|
|
|
|
/* Compatibility for RL/priority-unaware code. */
|
|
|
|
|
priority = ccb->ccb_h.pinfo.priority;
|
|
|
|
|
if (priority <= CAM_PRIORITY_OOB)
|
|
|
|
|
priority += CAM_PRIORITY_OOB + 1;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Setup fields in the CCB like the path and the priority.
|
|
|
|
|
* The path in particular cannot be done in userland, since
|
|
|
|
|
* it is a pointer to a kernel data structure.
|
|
|
|
|
*/
|
|
|
|
|
xpt_setup_ccb_flags(&ccb->ccb_h, periph->path, priority,
|
|
|
|
|
ccb->ccb_h.flags);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Setup our done routine. There is no way for the user to
|
|
|
|
|
* have a valid pointer here.
|
|
|
|
|
*/
|
|
|
|
|
ccb->ccb_h.cbfcnp = passdone;
|
|
|
|
|
|
|
|
|
|
fc = ccb->ccb_h.func_code;
|
|
|
|
|
/*
|
|
|
|
|
* If this function code has memory that can be mapped in
|
|
|
|
|
* or out, we need to call passmemsetup().
|
|
|
|
|
*/
|
|
|
|
|
if ((fc == XPT_SCSI_IO) || (fc == XPT_ATA_IO)
|
|
|
|
|
|| (fc == XPT_SMP_IO) || (fc == XPT_DEV_MATCH)
|
|
|
|
|
|| (fc == XPT_DEV_ADVINFO)) {
|
|
|
|
|
error = passmemsetup(periph, io_req);
|
|
|
|
|
if (error != 0) {
|
|
|
|
|
uma_zfree(softc->pass_zone, io_req);
|
|
|
|
|
cam_periph_lock(periph);
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
} else
|
|
|
|
|
io_req->mapinfo.num_bufs_used = 0;
|
|
|
|
|
|
|
|
|
|
cam_periph_lock(periph);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Everything goes on the incoming queue initially.
|
|
|
|
|
*/
|
|
|
|
|
TAILQ_INSERT_TAIL(&softc->incoming_queue, io_req, links);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* If the CCB is queued, and is not a user CCB, then
|
|
|
|
|
* we need to allocate a slot for it. Call xpt_schedule()
|
|
|
|
|
* so that our start routine will get called when a CCB is
|
|
|
|
|
* available.
|
|
|
|
|
*/
|
|
|
|
|
if ((fc & XPT_FC_QUEUED)
|
|
|
|
|
&& ((fc & XPT_FC_USER_CCB) == 0)) {
|
|
|
|
|
xpt_schedule(periph, priority);
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* At this point, the CCB in question is either an
|
|
|
|
|
* immediate CCB (like XPT_DEV_ADVINFO) or it is a user CCB
|
|
|
|
|
* and therefore should be malloced, not allocated via a slot.
|
|
|
|
|
* Remove the CCB from the incoming queue and add it to the
|
|
|
|
|
* active queue.
|
|
|
|
|
*/
|
|
|
|
|
TAILQ_REMOVE(&softc->incoming_queue, io_req, links);
|
|
|
|
|
TAILQ_INSERT_TAIL(&softc->active_queue, io_req, links);
|
|
|
|
|
|
|
|
|
|
xpt_action(ccb);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* If this is not a queued CCB (i.e. it is an immediate CCB),
|
|
|
|
|
* then it is already done. We need to put it on the done
|
|
|
|
|
* queue for the user to fetch.
|
|
|
|
|
*/
|
|
|
|
|
if ((fc & XPT_FC_QUEUED) == 0) {
|
|
|
|
|
TAILQ_REMOVE(&softc->active_queue, io_req, links);
|
|
|
|
|
TAILQ_INSERT_TAIL(&softc->done_queue, io_req, links);
|
|
|
|
|
}
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
case CAMIOGET:
|
|
|
|
|
{
|
|
|
|
|
union ccb **user_ccb;
|
|
|
|
|
struct pass_io_req *io_req;
|
|
|
|
|
int old_error;
|
|
|
|
|
|
|
|
|
|
user_ccb = (union ccb **)addr;
|
|
|
|
|
old_error = 0;
|
|
|
|
|
|
|
|
|
|
io_req = TAILQ_FIRST(&softc->done_queue);
|
|
|
|
|
if (io_req == NULL) {
|
|
|
|
|
error = ENOENT;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Remove the I/O from the done queue.
|
|
|
|
|
*/
|
|
|
|
|
TAILQ_REMOVE(&softc->done_queue, io_req, links);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* We have to drop the lock during the copyout because the
|
|
|
|
|
* copyout can result in VM faults that require sleeping.
|
|
|
|
|
*/
|
|
|
|
|
cam_periph_unlock(periph);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Do any needed copies (e.g. for reads) and revert the
|
|
|
|
|
* pointers in the CCB back to the user's pointers.
|
|
|
|
|
*/
|
|
|
|
|
error = passmemdone(periph, io_req);
|
|
|
|
|
|
|
|
|
|
old_error = error;
|
|
|
|
|
|
|
|
|
|
io_req->ccb.ccb_h.periph_links = io_req->user_periph_links;
|
|
|
|
|
io_req->ccb.ccb_h.periph_priv = io_req->user_periph_priv;
|
|
|
|
|
|
|
|
|
|
#if 0
|
|
|
|
|
xpt_print(periph->path, "Copying to user CCB %p from "
|
|
|
|
|
"kernel address %p\n", *user_ccb, &io_req->ccb);
|
|
|
|
|
#endif
|
|
|
|
|
|
|
|
|
|
error = copyout(&io_req->ccb, *user_ccb, sizeof(union ccb));
|
|
|
|
|
if (error != 0) {
|
|
|
|
|
xpt_print(periph->path, "Copy to user CCB %p from "
|
|
|
|
|
"kernel address %p failed with error %d\n",
|
|
|
|
|
*user_ccb, &io_req->ccb, error);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Prefer the first error we got back, and make sure we
|
|
|
|
|
* don't overwrite bad status with good.
|
|
|
|
|
*/
|
|
|
|
|
if (old_error != 0)
|
|
|
|
|
error = old_error;
|
|
|
|
|
|
|
|
|
|
cam_periph_lock(periph);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* At this point, if there was an error, we could potentially
|
|
|
|
|
* re-queue the I/O and try again. But why? The error
|
|
|
|
|
* would almost certainly happen again. We might as well
|
|
|
|
|
* not leak memory.
|
|
|
|
|
*/
|
|
|
|
|
uma_zfree(softc->pass_zone, io_req);
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
default:
|
|
|
|
|
error = cam_periph_ioctl(periph, cmd, addr, passerror);
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
bailout:
|
|
|
|
|
cam_periph_unlock(periph);
|
|
|
|
|
|
|
|
|
|
return(error);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int
|
|
|
|
|
passpoll(struct cdev *dev, int poll_events, struct thread *td)
|
|
|
|
|
{
|
|
|
|
|
struct cam_periph *periph;
|
|
|
|
|
struct pass_softc *softc;
|
|
|
|
|
int revents;
|
|
|
|
|
|
|
|
|
|
periph = (struct cam_periph *)dev->si_drv1;
|
|
|
|
|
softc = (struct pass_softc *)periph->softc;
|
|
|
|
|
|
|
|
|
|
revents = poll_events & (POLLOUT | POLLWRNORM);
|
|
|
|
|
if ((poll_events & (POLLIN | POLLRDNORM)) != 0) {
|
|
|
|
|
cam_periph_lock(periph);
|
|
|
|
|
|
|
|
|
|
if (!TAILQ_EMPTY(&softc->done_queue)) {
|
|
|
|
|
revents |= poll_events & (POLLIN | POLLRDNORM);
|
|
|
|
|
}
|
|
|
|
|
cam_periph_unlock(periph);
|
|
|
|
|
if (revents == 0)
|
|
|
|
|
selrecord(td, &softc->read_select);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return (revents);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int
|
|
|
|
|
passkqfilter(struct cdev *dev, struct knote *kn)
|
|
|
|
|
{
|
|
|
|
|
struct cam_periph *periph;
|
|
|
|
|
struct pass_softc *softc;
|
|
|
|
|
|
|
|
|
|
periph = (struct cam_periph *)dev->si_drv1;
|
|
|
|
|
softc = (struct pass_softc *)periph->softc;
|
|
|
|
|
|
|
|
|
|
kn->kn_hook = (caddr_t)periph;
|
|
|
|
|
kn->kn_fop = &passread_filtops;
|
|
|
|
|
knlist_add(&softc->read_select.si_note, kn, 0);
|
|
|
|
|
|
|
|
|
|
return (0);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
|
passreadfiltdetach(struct knote *kn)
|
|
|
|
|
{
|
|
|
|
|
struct cam_periph *periph;
|
|
|
|
|
struct pass_softc *softc;
|
|
|
|
|
|
|
|
|
|
periph = (struct cam_periph *)kn->kn_hook;
|
|
|
|
|
softc = (struct pass_softc *)periph->softc;
|
|
|
|
|
|
|
|
|
|
knlist_remove(&softc->read_select.si_note, kn, 0);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int
|
|
|
|
|
passreadfilt(struct knote *kn, long hint)
|
|
|
|
|
{
|
|
|
|
|
struct cam_periph *periph;
|
|
|
|
|
struct pass_softc *softc;
|
|
|
|
|
int retval;
|
|
|
|
|
|
|
|
|
|
periph = (struct cam_periph *)kn->kn_hook;
|
|
|
|
|
softc = (struct pass_softc *)periph->softc;
|
|
|
|
|
|
|
|
|
|
cam_periph_assert(periph, MA_OWNED);
|
|
|
|
|
|
|
|
|
|
if (TAILQ_EMPTY(&softc->done_queue))
|
|
|
|
|
retval = 0;
|
|
|
|
|
else
|
|
|
|
|
retval = 1;
|
|
|
|
|
|
|
|
|
|
return (retval);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Generally, "ccb" should be the CCB supplied by the kernel. "inccb"
|
|
|
|
|
* should be the CCB that is copied in from the user.
|
|
|
|
|
*/
|
|
|
|
|
static int
|
|
|
|
|
passsendccb(struct cam_periph *periph, union ccb *ccb, union ccb *inccb)
|
|
|
|
|
{
|
|
|
|
|
struct pass_softc *softc;
|
|
|
|
|
struct cam_periph_map_info mapinfo;
|
|
|
|
|
xpt_opcode fc;
|
|
|
|
|
int error;
|
|
|
|
|
|
|
|
|
|
softc = (struct pass_softc *)periph->softc;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* There are some fields in the CCB header that need to be
|
|
|
|
|
* preserved, the rest we get from the user.
|
|
|
|
|
*/
|
|
|
|
|
xpt_merge_ccb(ccb, inccb);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
*/
|
|
|
|
|
ccb->ccb_h.cbfcnp = passdone;
|
1998-09-15 06:36:34 +00:00
|
|
|
|
|
|
|
|
/*
|
2013-06-07 00:22:38 +00:00
|
|
|
* Let cam_periph_mapmem do a sanity check on the data pointer format.
|
|
|
|
|
* Even if no data transfer is needed, it's a cheap check and it
|
|
|
|
|
* simplifies the code.
|
1998-09-15 06:36:34 +00:00
|
|
|
*/
|
2013-06-07 00:22:38 +00:00
|
|
|
fc = ccb->ccb_h.func_code;
|
|
|
|
|
if ((fc == XPT_SCSI_IO) || (fc == XPT_ATA_IO) || (fc == XPT_SMP_IO)
|
|
|
|
|
|| (fc == XPT_DEV_MATCH) || (fc == XPT_DEV_ADVINFO)) {
|
1998-09-15 06:36:34 +00:00
|
|
|
bzero(&mapinfo, sizeof(mapinfo));
|
|
|
|
|
|
2007-04-15 08:49:19 +00:00
|
|
|
/*
|
|
|
|
|
* cam_periph_mapmem calls into proc and vm functions that can
|
|
|
|
|
* sleep as well as trigger I/O, so we can't hold the lock.
|
|
|
|
|
* Dropping it here is reasonably safe.
|
|
|
|
|
*/
|
|
|
|
|
cam_periph_unlock(periph);
|
2015-09-30 13:31:37 +00:00
|
|
|
error = cam_periph_mapmem(ccb, &mapinfo, softc->maxio);
|
2007-04-15 08:49:19 +00:00
|
|
|
cam_periph_lock(periph);
|
1998-09-15 06:36:34 +00:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* cam_periph_mapmem returned an error, we can't continue.
|
|
|
|
|
* Return the error to the user.
|
|
|
|
|
*/
|
|
|
|
|
if (error)
|
|
|
|
|
return(error);
|
2013-06-07 00:22:38 +00:00
|
|
|
} else
|
|
|
|
|
/* Ensure that the unmap call later on is a no-op. */
|
|
|
|
|
mapinfo.num_bufs_used = 0;
|
1998-09-15 06:36:34 +00:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* If the user wants us to perform any error recovery, then honor
|
|
|
|
|
* that request. Otherwise, it's up to the user to perform any
|
|
|
|
|
* error recovery.
|
|
|
|
|
*/
|
One more major cam_periph_error() rewrite to improve error handling and
reporting. It includes:
- removing of error messages controlled by bootverbose, replacing them
with more universal and informative debugging on CAM_DEBUG_INFO level,
that is now built into the kernel by default;
- more close following to the arguments submitted by caller, such as
SF_PRINT_ALWAYS, SF_QUIET_IR and SF_NO_PRINT; consumer knows better which
errors are usual/expected at this point and which are really informative;
- adding two new flags SF_NO_RECOVERY and SF_NO_RETRY to allow caller
specify how much assistance it needs at this point; previously consumers
controlled that by not calling cam_periph_error() at all, but that made
behavior inconsistent and debugging complicated;
- tuning debug messages and taken actions order to make debugging output
more readable and cause-effect relationships visible;
- making camperiphdone() (common device recovery completion handler) to
also use cam_periph_error() in most cases, instead of own dumb code;
- removing manual sense fetching code from cam_periph_error(); I was told
by number of people that it is SIM obligation to fetch sense data, so this
code is useless and only significantly complicates recovery logic;
- making ada, da and pass driver to use cam_periph_error() with new limited
recovery options to handle error recovery and debugging in common way;
as one of results, CAM_REQUEUE_REQ and other retrying statuses are now
working fine with pass driver, that caused many problems before.
- reverting r186891 by raj@ to avoid burning few seconds in tight DELAY()
loops on device probe, while device simply loads media; I think that problem
may already be fixed in other way, and even if it is not, solution must be
different.
Sponsored by: iXsystems, Inc.
MFC after: 2 weeks
2012-06-09 13:07:44 +00:00
|
|
|
cam_periph_runccb(ccb, passerror, /* cam_flags */ CAM_RETRY_SELTO,
|
|
|
|
|
/* sense_flags */ ((ccb->ccb_h.flags & CAM_PASS_ERR_RECOVER) ?
|
|
|
|
|
SF_RETRY_UA : SF_NO_RECOVERY) | SF_NO_PRINT,
|
MFp4: Large set of CAM inprovements.
- Unify bus reset/probe sequence. Whenever bus attached at boot or later,
CAM will automatically reset and scan it. It allows to remove duplicate
code from many drivers.
- Any bus, attached before CAM completed it's boot-time initialization,
will equally join to the process, delaying boot if needed.
- New kern.cam.boot_delay loader tunable should help controllers that
are still unable to register their buses in time (such as slow USB/
PCCard/ CardBus devices), by adding one more event to wait on boot.
- To allow synchronization between different CAM levels, concept of
requests priorities was extended. Priorities now split between several
"run levels". Device can be freezed at specified level, allowing higher
priority requests to pass. For example, no payload requests allowed,
until PMP driver enable port. ATA XPT negotiate transfer parameters,
periph driver configure caching and so on.
- Frozen requests are no more counted by request allocation scheduler.
It fixes deadlocks, when frozen low priority payload requests occupying
slots, required by higher levels to manage theit execution.
- Two last changes were holding proper ATA reinitialization and error
recovery implementation. Now it is done: SATA controllers and Port
Multipliers now implement automatic hot-plug and should correctly
recover from timeouts and bus resets.
- Improve SCSI error recovery for devices on buses without automatic sense
reporting, such as ATAPI or USB. For example, it allows CAM to wait, while
CD drive loads disk, instead of immediately return error status.
- Decapitalize diagnostic messages and make them more readable and sensible.
- Teach PMP driver to limit maximum speed on fan-out ports.
- Make boot wait for PMP scan completes, and make rescan more reliable.
- Fix pass driver, to return CCB to user level in case of error.
- Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
|
|
|
softc->device_stats);
|
1998-09-15 06:36:34 +00:00
|
|
|
|
2013-06-07 00:22:38 +00:00
|
|
|
cam_periph_unmapmem(ccb, &mapinfo);
|
1998-09-15 06:36:34 +00:00
|
|
|
|
|
|
|
|
ccb->ccb_h.cbfcnp = NULL;
|
|
|
|
|
ccb->ccb_h.periph_priv = inccb->ccb_h.periph_priv;
|
|
|
|
|
bcopy(ccb, inccb, sizeof(union ccb));
|
|
|
|
|
|
MFp4: Large set of CAM inprovements.
- Unify bus reset/probe sequence. Whenever bus attached at boot or later,
CAM will automatically reset and scan it. It allows to remove duplicate
code from many drivers.
- Any bus, attached before CAM completed it's boot-time initialization,
will equally join to the process, delaying boot if needed.
- New kern.cam.boot_delay loader tunable should help controllers that
are still unable to register their buses in time (such as slow USB/
PCCard/ CardBus devices), by adding one more event to wait on boot.
- To allow synchronization between different CAM levels, concept of
requests priorities was extended. Priorities now split between several
"run levels". Device can be freezed at specified level, allowing higher
priority requests to pass. For example, no payload requests allowed,
until PMP driver enable port. ATA XPT negotiate transfer parameters,
periph driver configure caching and so on.
- Frozen requests are no more counted by request allocation scheduler.
It fixes deadlocks, when frozen low priority payload requests occupying
slots, required by higher levels to manage theit execution.
- Two last changes were holding proper ATA reinitialization and error
recovery implementation. Now it is done: SATA controllers and Port
Multipliers now implement automatic hot-plug and should correctly
recover from timeouts and bus resets.
- Improve SCSI error recovery for devices on buses without automatic sense
reporting, such as ATAPI or USB. For example, it allows CAM to wait, while
CD drive loads disk, instead of immediately return error status.
- Decapitalize diagnostic messages and make them more readable and sensible.
- Teach PMP driver to limit maximum speed on fan-out ports.
- Make boot wait for PMP scan completes, and make rescan more reliable.
- Fix pass driver, to return CCB to user level in case of error.
- Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
|
|
|
return(0);
|
1998-09-15 06:36:34 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int
|
|
|
|
|
passerror(union ccb *ccb, u_int32_t cam_flags, u_int32_t sense_flags)
|
|
|
|
|
{
|
|
|
|
|
struct cam_periph *periph;
|
|
|
|
|
struct pass_softc *softc;
|
|
|
|
|
|
|
|
|
|
periph = xpt_path_periph(ccb->ccb_h.path);
|
|
|
|
|
softc = (struct pass_softc *)periph->softc;
|
|
|
|
|
|
|
|
|
|
return(cam_periph_error(ccb, cam_flags, sense_flags,
|
|
|
|
|
&softc->saved_ccb));
|
|
|
|
|
}
|
