kernel/drivers/net/ethernet/cavium/liquidio/request_manager.c, branch linux-6.18.y

net: WQ_PERCPU added to alloc_workqueue users

2025-09-23T00:40:30Z

Currently if a user enqueue a work item using schedule_delayed_work() the used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to schedule_work() that is using system_wq and queue_work(), that makes use again of WORK_CPU_UNBOUND. This lack of consistentcy cannot be addressed without refactoring the API. alloc_workqueue() treats all queues as per-CPU by default, while unbound workqueues must opt-in via WQ_UNBOUND. This default is suboptimal: most workloads benefit from unbound queues, allowing the scheduler to place worker threads where they’re needed and reducing noise when CPUs are isolated. This change adds a new WQ_PERCPU flag at the network subsystem, to explicitly request the use of the per-CPU behavior. Both flags coexist for one release cycle to allow callers to transition their calls. Once migration is complete, WQ_UNBOUND can be removed and unbound will become the implicit default. With the introduction of the WQ_PERCPU flag (equivalent to !WQ_UNBOUND), any alloc_workqueue() caller that doesn’t explicitly specify WQ_UNBOUND must now use WQ_PERCPU. All existing users have been updated accordingly. Suggested-by: Tejun Heo Signed-off-by: Marco Crivellari Link: https://patch.msgid.link/20250918142427.309519-4-marco.crivellari@suse.com Signed-off-by: Jakub Kicinski

net: liquidio: fix overflow in octeon_init_instr_queue()

2025-09-18T14:47:17Z

The expression `(conf->instr_type == 64) << iq_no` can overflow because `iq_no` may be as high as 64 (`CN23XX_MAX_RINGS_PER_PF`). Casting the operand to `u64` ensures correct 64-bit arithmetic. Fixes: f21fb3ed364b ("Add support of Cavium Liquidio ethernet adapters") Signed-off-by: Alexey Nepomnyashih Signed-off-by: Jakub Kicinski

net: liquidio: fix mixed module-builtin object

2023-06-07T11:22:44Z

With CONFIG_LIQUIDIO=m and CONFIG_LIQUIDIO_VF=y (or vice versa), $(common-objs) are linked to a module and also to vmlinux even though the expected CFLAGS are different between builtins and modules. This is the same situation as fixed by commit 637a642f5ca5 ("zstd: Fixing mixed module-builtin objects"). Introduce the new module, liquidio-core, to provide the common functions to liquidio and liquidio-vf. Signed-off-by: Masahiro Yamada Reviewed-by: Simon Horman Signed-off-by: David S. Miller

liquidio: remove unused IQ_INSTR_MODE_64B function

2023-03-23T05:05:01Z

clang with W=1 reports drivers/net/ethernet/cavium/liquidio/request_manager.c:43:19: error: unused function 'IQ_INSTR_MODE_64B' [-Werror,-Wunused-function] static inline int IQ_INSTR_MODE_64B(struct octeon_device *oct, int iq_no) ^ This function and its macro wrapper are not used, so remove them. Signed-off-by: Tom Rix Reviewed-by: Simon Horman Link: https://lore.kernel.org/r/20230321184811.1827306-1-trix@redhat.com Signed-off-by: Jakub Kicinski

liquidio: Replace vmalloc_node + memset with vzalloc_node and use array_size

2020-06-17T22:04:03Z

Use vzalloc/vzalloc_node instead of the vmalloc/vzalloc_node and memset. Also, notice that vzalloc_node() function has no 2-factor argument form to calculate the size for the allocation, so multiplication factors need to be wrapped in array_size(). This issue was found with the help of Coccinelle and, audited and fixed manually. Addresses-KSPP-ID: https://github.com/KSPP/linux/issues/83 Signed-off-by: Gustavo A. R. Silva Signed-off-by: David S. Miller

liquidio: add cleanup in octeon_setup_iq()

2019-08-15T19:37:37Z

If oct->fn_list.enable_io_queues() fails, no cleanup is executed, leading to memory/resource leaks. To fix this issue, invoke octeon_delete_instr_queue() before returning from the function. Signed-off-by: Wenwen Wang Signed-off-by: David S. Miller

liquidio: Replace vmalloc + memset with vzalloc

2019-07-18T18:54:29Z

Use vzalloc and vzalloc_node instead of using vmalloc and vmalloc_node and then zeroing the allocated memory by memset 0. This simplifies the code. Signed-off-by: Chuhong Yuan Signed-off-by: David S. Miller

drivers: Remove explicit invocations of mmiowb()

2019-04-08T11:01:02Z

mmiowb() is now implied by spin_unlock() on architectures that require it, so there is no reason to call it from driver code. This patch was generated using coccinelle: @mmiowb@ @@ - mmiowb(); and invoked as: $ for d in drivers include/linux/qed sound; do \ spatch --include-headers --sp-file mmiowb.cocci --dir $d --in-place; done NOTE: mmiowb() has only ever guaranteed ordering in conjunction with spin_unlock(). However, pairing each mmiowb() removal in this patch with the corresponding call to spin_unlock() is not at all trivial, so there is a small chance that this change may regress any drivers incorrectly relying on mmiowb() to order MMIO writes between CPUs using lock-free synchronisation. If you've ended up bisecting to this commit, you can reintroduce the mmiowb() calls using wmb() instead, which should restore the old behaviour on all architectures other than some esoteric ia64 systems. Acked-by: Linus Torvalds Signed-off-by: Will Deacon

liquidio: remove set but not used variable 'irh'

2018-09-02T05:24:52Z

Fixes gcc '-Wunused-but-set-variable' warning: drivers/net/ethernet/cavium/liquidio/request_manager.c: In function 'lio_process_iq_request_list': drivers/net/ethernet/cavium/liquidio/request_manager.c:383:27: warning: variable 'irh' set but not used [-Wunused-but-set-variable] Signed-off-by: YueHaibing Acked-by: Felix Manlunas Signed-off-by: David S. Miller

liquidio: improve soft command handling

2018-08-30T03:07:41Z

1. Set LIO_SC_MAX_TMO_MS as the maximum timeout value for a soft command (sc). All sc's use this value as a hard timeout value. Add expiry_time in struct octeon_soft_command to keep the hard timeout value. The field wait_time and timeout in struct octeon_soft_command will be obsoleted in the last patch of this patch series. 2. Add processing a synchronous sc in sc response thread lio_process_ordered_list. The memory allocated for a synchronous sc will be freed by lio_process_ordered_list() to the sc pool. 3. Add two response lists for lio_process_ordered_list to process the storage allocated for sc's: OCTEON_DONE_SC_LIST response list keeps all sc's which will be freed to the pool after their requestors have finished processing the responses. OCTEON_ZOMBIE_SC_LIST response list keeps all sc's which have got LIO_SC_MAX_TMO_MS timeout. When an sc gets a hard timeout, lio_process_order_list() will recheck its status 1 ms later. If the status has not updated by the firmware at that time, the sc will be removed from OCTEON_DONE_SC_LIST response list to OCTEON_ZOMBIE_SC_LIST response list. The sc's in the OCTEON_ZOMBIE_SC_LIST response list will be freed when the driver is unloaded. Signed-off-by: Weilin Chang Signed-off-by: Felix Manlunas Signed-off-by: David S. Miller