26 Oct, 2010
40 commits
-
This patch adds the Scatter-Gather (sg) API to libosd.
Scatter-gather enables a write/read of multiple none-contiguous
areas of an object, in a single call. The extents may overlap
and/or be in any order.The Scatter-Gather list is sent to the target in what is called
a "cdb continuation segment". This is yet another possible segment
in the osd-out-buffer. It is unlike all other segments in that it
sits before the actual "data" segment (which until now was always
first), and that it is signed by itself and not part of the data
buffer. This is because the cdb-continuation-segment is considered
a spill-over of the CDB data, and is therefor signed under
OSD_SEC_CAPKEY and higher.TODO: A new osd_finalize_request_ex version should be supplied so
the @caps received on the network also contains a size parameter
and can be spilled over into the "cdb continuation segment".Thanks to John Chandy for the original
code, and investigations. And the implementation of SG support
in the osd-target.Original-coded-by: John Chandy
Signed-off-by: Boaz Harrosh
Signed-off-by: James Bottomley -
At osd_end_request first free the request that might
point to pages, then free these pages. In reverse order
of allocation. For now it's just anal neatness. When we'll
use mempools It'll also pay in performance.Signed-off-by: Boaz Harrosh
Signed-off-by: James Bottomley -
The _osd_req_finalize_attr_page was off by a mile, when trying to
append the enc_get_attr segment instead of the proper set_attr segment.Also properly support when we don't have any attribute to set while
getting a full page. And when clearing an attribute by setting it's
size to zero.Reported-by: John Chandy
Signed-off-by: Boaz Harrosh
Signed-off-by: James Bottomley -
Signed-off-by: Alex Iannicelli
Signed-off-by: James Smart
Signed-off-by: James Bottomley -
- Add new WQE fields as defined by new SLI interface to support new hardware.
Signed-off-by: Alex Iannicelli
Signed-off-by: James Smart
Signed-off-by: James Bottomley -
Fix critical errors
- Update send_scsi_event to validate pnode pointer active before copying
the wwpn information.
- Add a message, mailbox_idle, and unlock before failing SECURITY_MGMT
or AUTH_PORT mailbox commands
- Prevent spin_lock_irqsave from being called twice in a row.Signed-off-by: Alex Iannicelli
Signed-off-by: James Smart
Signed-off-by: James Bottomley -
Adapter Shutdown and Unregistration cleanup
- Correct the logic around hba shutdown. Prior to final reset, the
driver must wait for all XRIs to return from the adapter. Added logic
to poll, progressively slowing the poll rate as delay gets longer.
- Correct behavior around the rsvd1 field in UNREG_RPI_ALL mailbox
completion and final rpi cleanup.
- Updated logic to move pending VPI registrations to their completion
in cases where a CVL may be received while registration in progress.
- Added unreg all rpi mailbox command before unreg vpi.Signed-off-by: Alex Iannicelli
Signed-off-by: James Smart
Signed-off-by: James Bottomley -
Added driver logic to detect the last devloss timeout of remote nodes which
was still in use of FCF. At that point, the driver should set the last
in-use remote node devloss timeout flag if it was not already set and should
perform proper action on the in-use FCF and recover of FCF from firmware,
depending on the state the driver's FIP engine is in.Find eligible FCF through FCF table rescan or the next new FCF event when
FCF table rescan turned out empty eligible FCF, and the successful flogi
into an FCF shall clear the HBA_DEVLOSS_TMO flag, indicating the successful
recovery from devloss timeout.[jejb: add delay.h include to lpfc_hbadisc.c to fix ppc compile]
Signed-off-by: Alex Iannicelli
Signed-off-by: James Smart
Signed-off-by: James Bottomley -
Add support of received ELS commands
- Add support for received RLS ELS command
- Add support for received ECHO ELS command
- Add support for received RTV ELS commandSigned-off-by: Alex Iannicelli
Signed-off-by: James Smart
Signed-off-by: James Bottomley -
FC/FCoE Discovery fixes:
- Call the lpfc_drain_txq only for SLI4 hba
- In lpfc_cmpl_els_fdisc, fix code path that does not free IOCB.
- Treated firmware matching FCF property with different index as error
- Propagate error returns from lpfc_issue_els_flogi()
- Refactored lpfc_unregister_unused_fcf() to create a post
lpfc_dev_loss_tmo handler call for SLI-4 devices. Allows checking of
fcf after last ndlp released so that fcf can be released if no longer
in use.
- Replaced individual FCF_XXXX_DISC flag clearing in lieu of aggregate
FCF_DISCOVERY flag upon succesful completion of flogi.
- Correct setting of altBbCredit value in sparams to correct issue with
logins with remote loop-based devices.Signed-off-by: Alex Iannicelli
Signed-off-by: James Smart
Signed-off-by: James Bottomley -
There was an addition to the hardware roadmap that includes a new adapter.
This patch adds the new definitions for the adapter.Signed-off-by: Wayne Boyer
Acked-by: Brian King
Signed-off-by: James Bottomley -
This patch addresses the comments from Randy Dunlap (Randy.Dunlap@oracle.com)
regarding comment blocks that begining with "/**". bfa driver comments
currently do not follow kernel-doc convention, we hence replace all
/** with /* and **/ with */.Signed-off-by: Jing Huang
Signed-off-by: James Bottomley -
This patch addresses the comments from Randy Dunlap (Randy.Dunlap@oracle.com)
regarding comment blocks that begining with "/**". bfa driver comments
currently do not follow kernel-doc convention, we hence replace all
/** with /* and **/ with */.Signed-off-by: Jing Huang
Signed-off-by: James Bottomley -
Fix compile warning for frame size over 1024 in gcc 4.4.
Signed-off-by: Jing Huang
Signed-off-by: James Bottomley -
This patch replaces register access functions and macros with the the ones
provided by linux.Signed-off-by: Jing Huang
Signed-off-by: James Bottomley -
Signed-off-by: Jing Huang
Signed-off-by: James Bottomley -
This patch removes os wrapper and unused functions.
bfa_os_assign(), bfa_os_memset(), bfa_os_memcpy(), bfa_os_udelay()
bfa_os_vsprintf(), bfa_os_snprintf(), and bfa_os_get_clock() are replaced with
direct assignment or native linux functions. Some unused functions related to VF
(Vitual fabric) are also removed.Signed-off-by: Jing Huang
Signed-off-by: James Bottomley -
Signed-off-by: Vijay Chauhan
Acked-by: Chandra Seetharaman
Signed-off-by: James Bottomley -
Ignore active open reply with status negative advice. This is an
informational message.Signed-off-by: Karen Xie
Reviewed-by: Mike Christie
Signed-off-by: James Bottomley -
Signed-off-by: Giridhar Malavali
Signed-off-by: Madhuranath Iyengar
Signed-off-by: James Bottomley -
This patch fixes an issue which causes the firmware to fail with a
'PRLI failed' status code (iop1 = 405). This status triggers the
driver to fall into an incorrect code-path which does not attempt
a login retry.Signed-off-by: Andrew Vasquez
Signed-off-by: Madhuranath Iyengar
Signed-off-by: James Bottomley -
This patch fixes a regression introduced by commit
083a469db4ecf3b286a96b5b722c37fc1affe0beqla2xxx_eh_wait_on_command() is waiting for an srb to
complete, which will never happen as the routine took
a reference to the srb previously and will only drop it
after this function. So every command abort will fail.Signed-off-by: Mike Christie
Signed-off-by: Giridhar Malavali
Signed-off-by: Madhuranath Iyengar
Signed-off-by: James Bottomley -
This patch adds a shutdown handler to qla2xxx driver to make sure that all
DMA and firmware activities are stopped, and any associated driver resources
are released. The need for this handler arose when executing kexec in specific
environments caused the data of the 2nd kernel to be corrupted, due to DMA
activities.Signed-off-by: Madhuranath Iyengar
Signed-off-by: James Bottomley -
Commit feafb7b1714cf599a6d0fed45801ab3f66046cbd neglected to initialize
the spinlock.Signed-off-by: Andrew Vasquez
Signed-off-by: Madhuranath Iyengar
Signed-off-by: James Bottomley -
This patch cleans up any printk or debug tracing of the the
serial_number field in the qla2xxx driver.Signed-off-by: Madhuranath Iyengar
Signed-off-by: James Bottomley -
Signed-off-by: Harish Zunjarrao
Signed-off-by: Madhuranath Iyengar
Signed-off-by: James Bottomley -
Currently when we receive a CS_RESET as a response for a SCSI command the
driver will return DID_TRANSPORT_DISRUPTED back to the SCSI mid-layer. There
are certain circumstances where this could cause the mid-layer to exhaust all of
its retries if the FC port goes away for a short time. This will result in
commands being prematurly failed. Moving the CS_RESET return code to be
grouped with other link level events will cause the FC transport layer to block
that target's queue thus preventing the premature exhaustion of retries.Signed-off-by: Chad Dupuis
Signed-off-by: Madhuranath Iyengar
Signed-off-by: James Bottomley -
Using del_timer_sync() in the qla2x00_ctx_sp_free() function may cause a kernel
panic as it is not interrupt context safe and qla2x00_ctx_sp_free() may be
called from a softirq context. Changing the call from del_timer_sync() to
del_timer() will make the function interrupt context safe.Signed-off-by: Chad Dupuis
Signed-off-by: Madhuranath Iyengar
Signed-off-by: James Bottomley -
Add the module parameter ql2xgffidenable to disable/enable the use of the
GFF_ID name server command to prevent non FCP SCSI devices from being added to
the driver's internal fc_port database.Signed-off-by: Chad Dupuis
Signed-off-by: Madhuranath Iyengar
Signed-off-by: James Bottomley -
This patch removes the use of the port down retry counter as a mechanism to
update a fcport state. The internal driver counter is a residual carry-over
from pre-FC-transport aware driver inteaction. The ql2xport_down_retry module
parameter and NVRAM set ha->port_down_retry_count remain in order to seed the
fc-host's default dev-loss-tmo.Signed-off-by: Chad Dupuis
Signed-off-by: Madhuranath Iyengar
Signed-off-by: James Bottomley -
IRQs are already disabled here so we don't need to disable them again.
But more importantly, the spin_lock_irqsave() overwrites "flags" and
that breaks things when we want to re-enable the IRQs when we call
spin_unlock_irqrestore(&ha->hardware_lock, flags);Signed-off-by: Dan Carpenter
Signed-off-by: Madhuranath Iyengar
Signed-off-by: James Bottomley -
An sr device that reports sense data with SK/ASC/ASCQ of 2/4/2 (Not ready,
Logical unit not ready, Initializing command required) will be handled
in sr_drive_status as (2/4/!1) and assumed to be a 'format in progress'
which returns CDS_DISC_OK. The drive will not be made ready in this case.Prior to 210ba1d1724f5c4ed87a2ab1a21ca861a915f734 sr_drive_status would
have returned CDS_TRAY_OPEN and this results in an START_STOP_UNIT to
close the tray, which resolves the initialization requirement.This patch adds handling for SK/ASC/ASCQ of 2/4/2 where it will return
CDS_TRAY_OPEN as a means of triggering a START_STOP_UNIT.This issue is seen on the IBM POWER platform when using a file-backed,
virtual optical device. The device does not support media queries
through the Get Event Status Notification command which could otherwise
trigger a START_STOP_UNIT call to close an open tray.Signed-off-by: Robert Jennings
Signed-off-by: James Bottomley -
A previous patch attempted to validate the destination
MAC address of a FCoE frame by checking that MAC
address against the received port's MAC address. The
implementation seems fine on the surface, but any
VN_Ports added using the NPIV feature will have their
own MAC addresses and these MACs were not being checked,
which prevented any NPIV VN_Ports from receiving frames.In other words, the following patch has broken NPIV.
519e5135e2537c9dbc1cbcc0891b0a936ff5dcd2
[SCSI] fcoe: adds src and dest mac address
checking for fcoe framesPart of the offending patch is correct, but the part
that broke NPIV was attempting to satisfy FC-BB-5
section D.5, 2.1-(discard frames that) "contain a destination MAC
address/destination N_Port_ID pair that was not
assigned by an FCF to one of the VN_Ports on the ENode"The language does _not_ say to compare the destination
FC-MAP/destination N_Port_ID, but instead to compare
the destination MAC address/destination N_Port_ID.>From the FC-BB-5 specification,
"A properly formed FPMA is one in which the 24 most
significant bits equal the Fabric’s FC-MAP value and
the least significant 24 bits equal the N_Port_ID
assigned to the VN_Port by the FCF."This means that we need to compare the FC Frame's
destination FCID against the embedded FCID in the
destination MAC address. This patch checks the lower
24 bits of the destination MAC address against
destination FCID in the Fibre Channel frame.For MAC validation the first line of defense is the
hardware MAC filtering. Each VN_Port will have a
unicast MAC addresses added to the hardware's
filtering table. The Ethernet driver should drop any
MACs not destined for a programmed MAC. This patch
adds a second line of defense that very specfically
compares an element in the FC frame against an element
in the Ethernet header, which is appropriate for the
FCoE layer.Many alternative approaches were considered, including
a LLD callback from libfc. The second most reasonable
approach seemed to be walking the list of NPIV ports
and check each of their MAC addresses against the
destination MAC address of the received frame. The
problem with this approach was that it is likely that
performance would suffer with the more NPIV ports added
to the system since every received frame would need to
walk this list, comparing each entry's MAC.Signed-off-by: Robert Love
Signed-off-by: James Bottomley -
Fix: When FIP frame is received, function fcoe_ctlr_vn_recv calls function
fcoe_ctlr_vn_parse which does memset for addr (&buf.rdata) which leads to
memory corruption. Code was trying to treat "buf" as struct but it was defined
as union. Fix is to change from union to struct for "buf" in function fcoe_ctlr_vn_recv.Technical Details: N/A
Signed-off-by: Kiran Patil
Acked-by: Joe Eykholt
Signed-off-by: Robert Love
Signed-off-by: James Bottomley -
When number of NPIV ports created are greater than the xids
allocated per pool -- for eg., creating 255 NPIV ports on a
system with nr_cpu_ids of 32, with each pool containing 128
xids -- and then generating a link event - for eg.,
shutdown/no shutdown -- on the switch port causes the hang
with the following stack trace.Call Trace:
schedule_timeout+0x19d/0x230
wait_for_common+0xc0/0x170
__cancel_work_timer+0xcf/0x1b0
fc_disc_stop+0x16/0x30 [libfc]
fc_lport_reset_locked+0x47/0x90 [libfc]
fc_lport_enter_reset+0x67/0xe0 [libfc]
fc_lport_disc_callback+0xbc/0xe0 [libfc]
fc_disc_done+0xa8/0xf0 [libfc]
fc_disc_timeout+0x29/0x40 [libfc]
run_workqueue+0xb8/0x140
worker_thread+0x96/0x110
kthread+0x96/0xa0
child_rip+0xa/0x20Fix is to not cancel the disc_work if discovery is already
stopped, thus allowing lport state machine to restart and try
discovery again.Signed-off-by: Bhanu Prakash Gollapudi
Acked-by: Joe Eykholt
Signed-off-by: Robert Love
Signed-off-by: James Bottomley -
It is unlikely but in case if it hits then it would cause panic
due to null cmd ptr, so far only one instance seen recently with
ESX though this was introduced long ago with this commit:-commit c1ecb90a66c5afc7cc5c9349f9c3714eef4a5cfb
Author: Chris Leech
Date: Thu Dec 10 09:59:26 2009 -0800
[SCSI] libfc: reduce hold time on SCSI host lockCurrently fsp->cmd is set to NULL w/o scsi_queue_lock before
dequeuing from scsi_pkt_queue and that could cause NULL
fsp->cmd in fc_fcp_cleanup_each_cmd for cmd completing
with fsp->cmd = NULL after fc_fcp_cleanup_each_cmd taken
reference. No need to set fsp->cmd to NULL as this is also
protected by fc_fcp_lock_pkt(), for above race the
fc_fcp_lock_pkt() in fc_fcp_cleanup_each_cmd() will fail
as that cmd is already done.Mike mentioned same issue at
http://www.open-fcoe.org/pipermail/devel/2010-September/010533.htmlSimilarly moved sc_cmd->SCp.ptr = NULL under scsi_queue_lock so
that scsi abort error handler won't abort on completed cmds.Signed-off-by: Vasu Dev
Signed-off-by: Robert Love
Signed-off-by: James Bottomley -
Since sometimes current FIP_MODE_AUTO mode falls back to non-FIP
mode while DCB link still getting ready in fabric mode with
its peer switch, it falls back after few libfc flogi retries
and that is not we want while working with FIP enabled
switches in FABRIC mode, therefore sets default as FIP_MODE_FABRIC
as discussed and agreed before in this mail thread
http://www.open-fcoe.org/pipermail/devel/2010-August/010511.htmlSigned-off-by: Vasu Dev
Signed-off-by: Robert Love
Signed-off-by: James Bottomley -
Sometimes switch in NPV mode rejects flogi request with DID
zero and in that case flogi is not tried again and port
remains offline, so this patch validates DID for non zero
along with only ACC response to allow flogi retry
for RJT with DID=0 also succeed FLOGI in next try.Signed-off-by: Vasu Dev
Signed-off-by: Robert Love
Signed-off-by: James Bottomley -
This is per Mile Christie feedback since in this case IO
could get retried for tape devices and therefore DID_REQUEUE
cannot be used, more details in this thread.http://marc.info/?l=linux-scsi&m=127970522630136&w=2
Signed-off-by: Vasu Dev
Signed-off-by: Robert Love
Signed-off-by: James Bottomley -
There does not seem to be a reason why libfc adds a 5
second delay to the user requested value for the dev loss
tmo. There also does not seem to be a reason to allow
setting it to 0 (or really close).This patch removes the extra 5 sec delay, and for 0 it
sets it to 1 like other fc drivers. We should actually
be able to set it to 0 since the queue_delayed_work API
will just call queue_work, but other drivers set it to 1 in
that case.Signed-off-by: Mike Christie
Signed-off-by: Robert Love
Signed-off-by: James Bottomley
