There were 2 ways to create a set of sysfs files for a kobj_type,
through the default_attrs field, and the default_groups field.
The default_groups field was introduced in aa30f47cf666
("kobject: Add support for default attribute groups to kobj_type") # v5.2
The default_attrs field was dropped in cdb4f26a63c3
("kobject: kobj_type: remove default_attrs") # v5.18
Hence move the scst_sysfs sysfs code to use default_groups field
since Linux Kernel v5.2
There were 2 ways to create a set of sysfs files for a kobj_type,
through the default_attrs field, and the default_groups field.
The default_groups field was introduced in aa30f47cf666
("kobject: Add support for default attribute groups to kobj_type") # v5.2
The default_attrs field was dropped in cdb4f26a63c3
("kobject: kobj_type: remove default_attrs") # v5.18
Hence move the scst_mem sysfs code to use default_groups field
since Linux Kernel v5.2
For session recovery, driver relies on the dpc thread to initiate certain
operations. The dpc thread runs exclusively without the Mailbox interface
being occupied. A recent code change for heartbeat check via mailbox cmd 0
is preventing the dpc thread from carrying out its operation. This patch
allows the higher priority error recovery to run first before running the
lower priority heartbeat check.
Link: https://lore.kernel.org/r/20220310092604.22950-9-njavali@marvell.com
Fixes: d94d8158e184 ("scsi: qla2xxx: Add heartbeat check")
Cc: stable@vger.kernel.org
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
[ commit 713b415726f1 upstream ]
User experienced device lost. The log shows Get port data base command was
queued up, failed, and requeued again. Every time it is requeued, it set
the FCF_ASYNC_ACTIVE. This prevents any recovery code from occurring
because driver thinks a recovery is in progress for this session. In
essence, this session is hung. The reason it gets into this place is the
session deletion got in front of this call due to link perturbation.
Break the requeue cycle and exit. The session deletion code will trigger a
session relogin.
Link: https://lore.kernel.org/r/20220310092604.22950-8-njavali@marvell.com
Fixes: 726b85487067 ("qla2xxx: Add framework for async fabric discovery")
Cc: stable@vger.kernel.org
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
[ commit c02aada06d19 upstream ]
During purex packet handling the driver was incorrectly freeing a
pre-allocated structure. Fix this by skipping that entry.
System crashed with the following stack during a module unload test.
Call Trace:
sbitmap_init_node+0x7f/0x1e0
sbitmap_queue_init_node+0x24/0x150
blk_mq_init_bitmaps+0x3d/0xa0
blk_mq_init_tags+0x68/0x90
blk_mq_alloc_map_and_rqs+0x44/0x120
blk_mq_alloc_set_map_and_rqs+0x63/0x150
blk_mq_alloc_tag_set+0x11b/0x230
scsi_add_host_with_dma.cold+0x3f/0x245
qla2x00_probe_one+0xd5a/0x1b80 [qla2xxx]
Call Trace with slub_debug and debug kernel:
kasan_report_invalid_free+0x50/0x80
__kasan_slab_free+0x137/0x150
slab_free_freelist_hook+0xc6/0x190
kfree+0xe8/0x2e0
qla2x00_free_device+0x3bb/0x5d0 [qla2xxx]
qla2x00_remove_one+0x668/0xcf0 [qla2xxx]
Link: https://lore.kernel.org/r/20220310092604.22950-6-njavali@marvell.com
Fixes: 62e9dd177732 ("scsi: qla2xxx: Change in PUREX to handle FPIN ELS requests")
Cc: stable@vger.kernel.org
Reported-by: Marco Patalano <mpatalan@redhat.com>
Tested-by: Marco Patalano <mpatalan@redhat.com>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
[ commit 0972252450f9 upstream ]
User experienced some of the LUN failed to get rediscovered after long
cable pull test. The issue is triggered by a race condition between driver
setting session online state vs starting the LUN scan process at the same
time. Current code set the online state after notifying the session is
available. In this case, trigger to start the LUN scan process happened
before driver could set the session in online state. LUN scan ends up with
failure due to the session online check was failing.
Set the online state before reporting of the availability of the session.
Link: https://lore.kernel.org/r/20220310092604.22950-3-njavali@marvell.com
Fixes: aecf043443d3 ("scsi: qla2xxx: Fix Remote port registration")
Cc: stable@vger.kernel.org
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
[ commit 6a45c8e137d4 upstream ]
The timeout handler and the done function are racing. When
qla2x00_async_iocb_timeout() starts to run it can be preempted by the
normal response path (via the firmware?). qla24xx_async_gpsc_sp_done()
releases the SRB unconditionally. When scheduling back to
qla2x00_async_iocb_timeout() qla24xx_async_abort_cmd() will access an freed
sp->qpair pointer:
qla2xxx [0000:83:00.0]-2871:0: Async-gpsc timeout - hdl=63d portid=234500 50:06:0e:80:08:77:b6:21.
qla2xxx [0000:83:00.0]-2853:0: Async done-gpsc res 0, WWPN 50:06:0e:80:08:77:b6:21
qla2xxx [0000:83:00.0]-2854:0: Async-gpsc OUT WWPN 20:45:00:27:f8:75:33:00 speeds=2c00 speed=0400.
qla2xxx [0000:83:00.0]-28d8:0: qla24xx_handle_gpsc_event 50:06:0e:80:08:77:b6:21 DS 7 LS 6 rc 0 login 1|1 rscn 1|0 lid 5
BUG: unable to handle kernel NULL pointer dereference at 0000000000000004
IP: qla24xx_async_abort_cmd+0x1b/0x1c0 [qla2xxx]
Obvious solution to this is to introduce a reference counter. One reference
is taken for the normal code path (the 'good' case) and one for the timeout
path. As we always race between the normal good case and the timeout/abort
handler we need to serialize it. Also we cannot assume any order between
the handlers. Since this is slow path we can use proper synchronization via
locks.
When we are able to cancel a timer (del_timer returns 1) we know there
can't be any error handling in progress because the timeout handler hasn't
expired yet, thus we can safely decrement the refcounter by one.
If we are not able to cancel the timer, we know an abort handler is
running. We have to make sure we call sp->done() in the abort handlers
before calling kref_put().
Link: https://lore.kernel.org/r/20220110050218.3958-3-njavali@marvell.com
Cc: stable@vger.kernel.org
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Co-developed-by: Daniel Wagner <dwagner@suse.de>
Signed-off-by: Daniel Wagner <dwagner@suse.de>
Signed-off-by: Saurav Kashyap <skashyap@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
[ commit 31e6cdbe0eae upstream ]
Instead of using the SCp.ptr field to track whether or not a command is
in flight, use the sp->type field to track this information. sp->type
must be set for proper operation of the qla2xxx driver. See e.g. the
switch (sp->type) statement in qla2x00_ct_entry().
Port the commit 8cc7b18f ("qla2x00t-32gbit: Stop using the SCSI pointer")
to the old qlogic driver.
Support for the following scsi changes in the Linux Kernel v5.18:
- ce70fd9a551a ("scsi: core: Remove the cmd field from struct scsi_request")
- 5b794f98074a ("scsi: core: Remove the sense and sense_len fields from struct scsi_request")
- 6aded12b10e0 ("scsi: core: Remove struct scsi_request")
The use of set_fs() for scst_write_file_transactional() was dropped in
94018a42 ("scst: Use kernel_{read,write}() instead of scst_{read,write}()").
Hence rename the goto label associated with this functionality.
Support for the following block/fs changes in the Linux Kernel v5.18:
- 609be1066731 ("block: pass a block_device and opf to bio_alloc_bioset")
- 07888c665b40 ("block: pass a block_device and opf to bio_alloc")
- cbcc268bb1ce ("fs: Move many prototypes to pagemap.h")
There is a regular need in the kernel to provide a way to declare
having a dynamically sized set of trailing elements in a structure.
Kernel code should always use “flexible array members”[1] for these
cases. The older style of one-element or zero-length arrays should
no longer be used[2].
This code was transformed with the help of Coccinelle:
($ spatch --jobs $(getconf _NPROCESSORS_ONLN) --sp-file script.cocci --include-headers --dir . > output.patch)
@@
identifier S, member, array;
type T1, T2;
@@
struct S {
...
T1 member;
T2 array[
- 0
];
};
[1] https://en.wikipedia.org/wiki/Flexible_array_member
[2] https://www.kernel.org/doc/html/v5.16/process/deprecated.html#zero-length-and-one-element-arrays
Link: https://github.com/KSPP/linux/issues/78
Add the opportunity to globally specify multiple addresses, which
iSCSI/iSER portal will listen on.
There is a way to specify the addresses for a target, on which
portals it will be available, using the allowed_portal parameter.
There is also a way to set this globally via the iscsi-scstd
--address parameter. The problem is that the last option allows
you to specify only one local address. So extend this functionality
to specify more then one address.
Also increase the maximum number of listening addresses to 32.
Signed-off-by: Aleksandr Dyadyushkin <alextalker@ya.ru>
For the old qla2x00 driver, the hw_target attribute allows you to
distinguish hardware and virtual targets.
But for the new qla2x00t-32gbit driver, this attribute turned out
to be broken. Regardless of whether it is a hardware or virtual target,
this attribute is created for that target and always returns 1.
Restore this broken functionality by returning 1 for the hardware target
and 0 otherwise.
