05 Oct, 2006
15 commits
-
open-iSCSI driver for Qlogic Corporation's iSCSI HBAs
Signed-off-by: Ravi Anand
Signed-off-by: David Somayajulu
Signed-off-by: Doug Maxey
Signed-off-by: Mike Christie
Signed-off-by: James Bottomley -
Signed-off-by: Jeff Garzik
Signed-off-by: James Bottomley -
- handle clear_user() error
- handle and properly unwind from sysfs errors thrown during mod init
- adjust order of calls in megasas_exit() to precisely match
registration order in megasas_init()Signed-off-by: Jeff Garzik
Updated for extra attribute and
Signed-off-by: James Bottomley -
Handle and unwind from errors returned by driver model functions.
Signed-off-by: Jeff Garzik
Signed-off-by: James Bottomley -
- Notice and handle sysfs errors in module init, tape init
- Properly unwind errors in module init
- Remove bogus st_sysfs_class==NULL test, it is guaranteed !NULL at that point
Signed-off-by: Jeff Garzik
Signed-off-by: James Bottomley -
- Properly handle and unwind errors in init_sd(). Fixes leaks on error,
if class_register() or scsi_register_driver() failed.- Ensure that exit_sd() execution order is the perfect inverse of
initialization order.FIXME: If some-but-not-all register_blkdev() calls fail, we wind up
calling unregister_blkdev() for block devices we did not register.
This was a pre-existing bug.Signed-off-by: Jeff Garzik
Signed-off-by: James Bottomley -
- check all sysfs-related return codes, and propagate them back to callers
- properly unwind errors in osst_probe(), init_osst(). This fixes a
leak that occured if scsi driver registration failed, and fixes an
oops if sysfs creation returned an error.(unrelated)
- kzalloc() cleanup in new_tape_buf()Signed-off-by: Jeff Garzik
Signed-off-by: James Bottomley -
Remove the obsolete hosts.h file under drivers/scsi.
Signed-off-by: Henrik Kretzschmar
Signed-off-by: James Bottomley -
Changes the obsolete Scsi_Cmnd to struct scsi_cmnd in aic7xxx_old.c.
Also replacing lots of whitespaces with tabs in structures and functions
which have been changed.Signed-off-by: Henrik Kretzschmar
Signed-off-by: James Bottomley -
This patch sets timeout of max 180 seconds for ioctl completion.
It also updates the Changelog and hikes the version to 3.05.Signed-off-by: Sumant Patro
Signed-off-by: James Bottomley -
This patch adds a tasklet for command completion.
Signed-off-by: Sumant Patro
Signed-off-by: James Bottomley -
This patch adds function to print the pending frame details before returning
failure from the reset routine. It also exposes a new variable megasas_dbg_lvl
that allows the user to set the debug level for logging.Signed-off-by: Sumant Patro
Signed-off-by: James Bottomley -
This patch adds function pointer to invoke disable interrupt for
xscale and ppc IOP based controllers. Removes old implementation that checks
for controller type in megasas_disable_intr.Signed-off-by: Sumant Patro
Signed-off-by: James Bottomley -
This patch removes duplicated code in frame calculation & adds
megasas_get_frame_count() that also takes into account the number of frames
that can be contained in the Main frame.
FW uses the frame count to pull sufficient number of frames from host memory.Signed-off-by: Sumant Patro
Signed-off-by: James Bottomley -
This patch has the following enhancements :
a. handles new transition states of FW to support controller hotplug.
b. It reduces by 1 the maximum cmds that the driver may send to FW.
c. Sends "Stop Processing" cmd to FW before returning failure from reset routine
d. Adds print in megasas_transition routine
e. Sends "RESET" flag to FW to do a soft reset of controller
to move from Operational to Ready state.
f. Sending correct pointer (cmd->sense) to pci_pool_freeSigned-off-by: Sumant Patro
Signed-off-by: James Bottomley
04 Oct, 2006
10 commits
-
Signed-off-by: Andrew Vasquez
Signed-off-by: James Bottomley -
Stall error handler if attempting recovery while an rport is
blocked. This avoids device offline scenarios due to errors in
the error handler.Reference implementation from lpfc/mptfc.
Signed-off-by: Andrew Vasquez
Signed-off-by: James Bottomley -
Signed-off-by: Andrew Vasquez
Signed-off-by: James Bottomley -
Signed-off-by: Andrew Vasquez
Signed-off-by: James Bottomley -
Signed-off-by: Andrew Vasquez
Signed-off-by: James Bottomley -
The system hostname will be used during a subsequent FDMI registration
with the fabric.Signed-off-by: Andrew Vasquez
Signed-off-by: James Bottomley -
Refactored original code from qla_gs.c:qla2x00_rsnn_nn().
Signed-off-by: Andrew Vasquez
Signed-off-by: James Bottomley -
iIDMA (Intelligent Interleaved Direct Memory Access) allows for
the HBA hardware to send FC frames at the rate at which they can
be received by a target device. By taking advantage of the
higher link rate, the HBA can maximize bandwidth utilization in a
heterogeneous multi-speed SAN.Within a fabric topology, port speed detection is done via a Name
Server command (GFPN_ID) followed by a Fabric Management command
(GPSC). In an FCAL/N2N topology, port speed is based on the HBA
link-rate.Signed-off-by: Andrew Vasquez
Signed-off-by: James Bottomley -
Changes the obsolete Scsi_Cmnd to struct scsi_cmnd in the arm subdir
of the scsi-subsys.Signed-off-by: Henrik Kretzschmar
Acked-by: Russell King
Signed-off-by: James Bottomley -
Converts the obsolete Scsi_Cmnd to struct scsi_cmnd in the ips-driver.
Signed-off-by: Henrik Kretzschmar
Signed-off-by: James Bottomley
03 Oct, 2006
1 commit
-
following an email from John Adams to me with a patch
to enable tmscsim to use blocks up to 1MB and a discussion on linux-scsi,
below is a patch to enable clustering for tmscsim. I made it switchable
with a module parameter, with default "enable" - in case somebody gets
problems with it. Unfortunately, I was not able to check if this alone
lets you use any bigger blocks with a tape, as my tape seems to only
support 1 block size - only "mt setblk 1" is successful, any other value
fails. OTOH, testing on a P-133 showed that enabling clustering alone
improves throughput by 10% and reduces CPU load by another 10%, so, seems
a worthy thing to do. As for setting max_sectors, that might become a
separate patch...Signed-off-by: James Bottomley
02 Oct, 2006
9 commits
-
When SCSI-2 they can support luns past 7 and sparse luns.
Signed-off-by: Mike Christie
Signed-off-by: James Bottomley -
support the report luns opcode
.
Signed-off-by: Mike Christie
Signed-off-by: James Bottomley -
This is from RHEL4. I do not have any info from our bugzilla. All
I could find was something like this thread
http://lkml.org/lkml/2005/1/7/346Report lun for linux does not work. It may be our lun format code or
it may be the device. It is probably not worth it to add anything
special for this device, so the patch just adds BLIST_NOREPORTLUN.Signed-off-by: Mike Christie
Signed-off-by: James Bottomley -
This is from RHEL4. This box can support
scsi2 and can also support BLIST_SPARSELUN | BLIST_LARGELUN.Signed-off-by: Mike Christie
Signed-off-by: James Bottomley -
In scsi_execute_async()'s error path, a struct scsi_io_context
allocated with kmem_cache_alloc() is kfree()'d. Obviously
kmem_cache_free() should be used instead.Signed-off-by: Arne Redlich
Signed-off-by: James Bottomley -
A new device (id 0x8650, nickname 'yosemite') support is added.
It's basically the same, except for following items:
- mapping of id and lun by firmware
- special handling for some commands in interrupt routine
- change of internal copy function for these special commands
- different reset handling code
- different shutdown notification commandSigned-off-by: Ed Lin
Signed-off-by: James Bottomley -
The payload_sz field in struct req_msg is not big enough to indicate
the size of req_msg, as its type is u8.
It is confirmed that this field is not used by firmware, so cancel
it here.Signed-off-by: Ed Lin
Signed-off-by: James Bottomley -
this overrun was spotted by coverity (cid #1403).
If type == ARRAY_SIZE(scsi_device_types), we are off by one.
Signed-off-by: Eric Sesterhenn
Signed-off-by: James Bottomley -
Comment says "Read high byte first as some registers increment..."
but code doesn't guarantee that, I think:
return ((ahd_inb(ahd, port+1) << 8) | ahd_inb(ahd, port));
Compiler can reorder it.Make the order explicit.
Signed-off-by: Denis Vlasenko
Signed-off-by: Andrew MortonFixed rejections and added aic7xxx code
Signed-off-by: James Bottomley
01 Oct, 2006
1 commit
-
Signed-off-by: Alan Cox
Signed-off-by: James Bottomley
28 Sep, 2006
1 commit
-
device_reprobe() should return an error code. When it does so,
scsi_device_reprobe() should propagate it back.Signed-off-by: Andrew Morton
Signed-off-by: James Bottomley
27 Sep, 2006
3 commits
-
gcc 4.1 with some extra warnings show the following:
drivers/scsi/ipr.c:6361: warning: comparison of unsigned expression < 0 is always false
drivers/scsi/ipr.c:6385: warning: comparison of unsigned expression < 0 is always false
drivers/scsi/ipr.c:6415: warning: comparison of unsigned expression < 0 is always falseThe problem is that rc is of the type u32, which can never be smaller than
zero, therefore all three error handling checks get useless. This patch
changes it to a normal int, because all usages / all functions it get used
with expect an int.Signed-off-by: Eric Sesterhenn
Acked-by: Brian King
Signed-off-by: Andrew Morton
Signed-off-by: James Bottomley -
Converts pci_module_init() to pci_register_driver() in the scsi subsys on
23 drivers which only return the value of pci_module_init().Signed-off-by: Henrik Kretzschmar
Signed-off-by: Andrew Morton
Signed-off-by: James Bottomley -
another signdness warning from gcc 4.1
drivers/scsi/osst.c:5154: warning: comparison of unsigned expression < 0 is always false
The problem is that blk is defined as unsigned, but all usages of it are
normal int cases. osst_get_frame_position() and osst_get_sector() return ints
and can return negative values. If blk stays an unsigned int, the error check
is useless.Signed-off-by: Eric Sesterhenn
Signed-off-by: Andrew Morton
Signed-off-by: James Bottomley