According to kernel documentation, this flag should be set if the
workqueue will be involved in the kernel's memory reclamation flow.
Since it is not, there is no need for the driver's WQ to have this
flag set so remove it.
create_workqueue() was replaced with alloc_workqueue() with max_active set
to 0. However, the original create_workqueue() implicitly set max_active
to 1.
This change has led to unexpected bugs because previously, work items
could only be executed one by one. With the change, they can now be
executed simultaneously.
This patch fixes the issue by restoring max_active to 1.
Fixes: f4686e9102 ("scst: Remove deprecated create_workqueue()")
Fixes: https://github.com/SCST-project/scst/issues/179
strlcpy() reads the entire source buffer first. This read may exceed the
destination size limit. This is both inefficient and can lead to linear
read overflows if a source string is not NULL-terminated [1].
[1] https://www.kernel.org/doc/html/latest/process/deprecated.html#strlcpy
Use sendmsg() conditionally with MSG_SPLICE_PAGES in write_data()
rather than calling sendpage().
Support for the following net layer changes in the Linux kernel v6.5:
- dc97391e6610 ("sock: Remove ->sendpage*() in favour of
sendmsg(MSG_SPLICE_PAGES)")
This patch carries out a refactoring of the sendpage functionality in
the write_data() function:
1. Reorganize the logic used to select the sock_sendpage function.
2. Streamline the data sending loop by reducing conditional branches and
eliminating labels.
3. Adjust the error handling for -EINTR and -EAGAIN to make the code
cleaner and easier to follow.
This patch doesn't change any functionality.
This patch introduces several improvements to the 'write iop loop' in
the write_data() function:
1. Move iop-related variables under the scope of the 'write iop loop'.
2. Eliminate the 'retry' label, use 'continue' instead for simplicity.
3. Remove the redundant 'rest' variable, use just 'res' instead.
This patch doesn't change any functionality.
This patch introduces several improvements to the write_data() function:
1. Remove the redundant 'sendpage' function pointer variable.
2. Update variables related to size to use the size_t type for better
type correctness and safety.
3. Introduce a new variable, 'parent_req', to store the
'write_cmnd->parent_req' pointer and reduce redundant accesses.
4. Fix several checkpatch warnings.
This patch doesn't change any functionality.
This patch changes the processing threads to use INTERRUPTIBLE sleep
states in the scst_wait_event_...() functions. This aims to avoid
warnings from the hung task detection checker and to prevent
unnecessary load counting.
Fixes: d8894cbd11 ("scst.h: Refactor wait_event_locked() to enhance usability and clarity")
1. Set the default process state to TASK_UNINTERRUPTIBLE during sleep.
This change is made because our current code does not check whether a
process was interrupted by a signal.
2. Prefix all SCST wait_event-related macros with 'scst_'. This helps to
distinguish SCST-specific macros from those provided by the Linux
kernel itself.
3. Add the capability to return an error code when a process in a
non-TASK_UNINTERRUPTIBLE state is interrupted by a signal.
4. Divide the wait_event_locked function based on each lock type,
resulting in the following new functions: scst_wait_event_lock(),
scst_wait_event_lock_bh(), and scst_wait_event_lock_irq().
Support for the following driver core changes in the Linux kernel v6.4:
- 1aaba11da9aa ("driver core: class: remove module * from class_create()")
- 2243acd50ac4 ("driver core: class: remove struct class_interface * from callbacks")
Remove the else because the if statement has a break or return statement.
This patch fixes the following checkpatch warnings:
WARNING:UNNECESSARY_ELSE: else is not generally useful after a break
or return.
kmem_cache_destroy() can handle NULL pointer correctly, so there is no need
to check NULL pointer before calling kmem_cache_destroy().
For kernel versions before v4.3 there is a backport of
kmem_cache_destroy() that checks for a null pointer.
This patch fixes the following checkpatch warnings:
WARNING:NEEDLESS_IF: kmem_cache_destroy(NULL) is safe and this check
is probably not required
This patch should fix the following bug:
iscsi-scst: ***ERROR***: Sending data failed: initiator ..., write_size 0, write_state 1, res 0
iscsi-scst: ***CRITICAL ERROR***: 0 6 31
------------[ cut here ]------------
kernel BUG at /usr/src/packages/BUILD/scst-3.7.0.8695/iscsi-scst/kernel/nthread.c:1517!
invalid opcode: 0000 [#1] SMP NOPTI
CPU: 12 PID: 997595 Comm: iscsiwr0_14 ...
...
RIP: 0010:iscsi_send+0x877/0x8b0 [iscsi_scst]
Call Trace:
istwr+0x123/0x3b0 [iscsi_scst]
kthread+0x120/0x136
ret_from_fork+0x24/0x36
-------------------------------------
What happens:
- istwr() calls scst_do_job_wr().
- scst_do_job_wr() calls iscsi_send().
- iscsi_send() sets the 'res' variable to 0 during error
in one of three possible places:
iscsi_do_send(), tx_padding(), tx_ddigest().
- All of these functions call exit_tx() which sets conn->write_state to TX_END.
- After iscsi_send() has completed for the current iteration, the next time
it processes iscsi_conn with conn->write_state == TX_END,
which will call BUG() in the switch default case.
Therefore, remove the res == 0 check in iscsi_send() to handle TX_END state.
Fixes: https://github.com/SCST-project/scst/issues/12
Suppress the following (false positive) Coverity complaint:
CID 271578 (#1 of 1): Dereference after null check (FORWARD_NULL)
var_deref_model: Passing null pointer (*ref_cmd).scst_cmd to
scst_set_delivery_status, which dereferences it
(*ref_cmd).scst_aen is set when (*ref_cmd).scst_state == ISCSI_CMD_STATE_AEN
and vice versa, so the Coverity complaint is a false positive. Hence rewrite
the code to suppress this complaint and make the code cleaner.
This patch fixes the following Coverity complaint:
CID 271606 (#1 of 1): Uninitialized scalar variable (UNINIT)
uninit_use_in_call: Using uninitialized value event.
Field event.target_name is uninitialized when calling __event_send.
Move PRINT_ERROR to the body of RDMA_CM_EVENT_CONNECT_REQUET because only
it can return an error and not log information about it.
Signed-off-by: Chesnokov Gleb <Chesnokov.G@raidix.com>
[bvanassche: modified patch title]
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@9486 d57e44dd-8a1f-0410-8b47-8ef2f437770f
During processing RDMA_CM_EVENT_ADDR_CHANGE event rdma_bind_addr in isert_setup_id function from isert_cm_evt_listener_handler returns error -98 [EADDRINUSE].
In principle, it is logical, because at that time the socket address was still bound to the old cma_id which will be destroyed via rdma_destroy_id only after processing the RDMA_CM_EVENT_ADDR_CHANGE event.
Move the creation of the cma_id in workqueue context and delete old cma_id directly, not through returning the error code to the upper level.
Signed-off-by: Chesnokov Gleb <Chesnokov.G@raidix.com>
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@9484 d57e44dd-8a1f-0410-8b47-8ef2f437770f
Introduce the function isert_setup_id(). This function creates and sets up
an RDMA CM ID. Move the isert_portal_listen() call into isert_portal_create().
Signed-off-by: Chesnokov Gleb <Chesnokov.G@raidix.com>
[ bvanassche: edited patch description and dropped support for older kernel
versions in the PRINT_INFO() statement ]
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@9481 d57e44dd-8a1f-0410-8b47-8ef2f437770f
When the low level driver exercises the hot unplug they would call
rdma_cm cma_remove_one which would fire DEVICE_REMOVAL event to all cma
consumers. Now, if consumer doesn't make sure they destroy all IB
objects created on that IB device instance prior to finalizing all
processing of DEVICE_REMOVAL callback, rdma_cm will let the lld to
de-register with IB core and destroy the IB device instance. And if the
consumer calls (say) ib_dereg_mr(), it will crash since that dev object
is NULL.
In the current implementation, iser-target just initiates the cleanup
and returns from DEVICE_REMOVAL callback. This deferred work creates a
race between iser-target cleaning IB objects(say MR) and lld destroying
IB device instance.
This patch includes the following fixes
-> make sure that consumer frees all IB objects associated with device
instance
-> return non-zero from the callback to destroy the rdma_cm id
Signed-off-by: Raju Rangoju <rajur@chelsio.com>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Doug Ledford <dledford@redhat.com>
See also upstream commit 63b268d232b8 ("IB/isert: Properly release
resources on DEVICE_REMOVAL")
Signed-off-by: Chesnokov Gleb <Chesnokov.G@raidix.com>
[ bvanassche: edited patch description and moved a break statement into a code
block ]
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@9480 d57e44dd-8a1f-0410-8b47-8ef2f437770f
Minimize the diffs with the upstream code base by introducing the function
isert_conn_init().
Signed-off-by: Chesnokov Gleb <Chesnokov.G@raidix.com>
[ bvanassche: edited patch description ]
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@9478 d57e44dd-8a1f-0410-8b47-8ef2f437770f
Move the isert_cm_disconnect_handler() definition, pass the event type to
isert_cm_evt_listener_handler() instead of the struct rdma_cm_event pointer
and change the return value of isert_cm_evt_listener_handler() from 0 /
-EINVAL into -1.
Signed-off-by: Chesnokov Gleb <Chesnokov.G@raidix.com>
[ bvanassche: made the patch description more detailed ]
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@9473 d57e44dd-8a1f-0410-8b47-8ef2f437770f
This patch does not change any functionality but removes multiple __force
__user casts.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@9197 d57e44dd-8a1f-0410-8b47-8ef2f437770f
In Linux kernel v5.9 use of the KERNEL_SOCKPTR() function is mandatory
when calling setsockopt(). Implement that function for kernels before v5.9.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@9149 d57e44dd-8a1f-0410-8b47-8ef2f437770f
