diff options
author | Jason Gunthorpe <jgg@mellanox.com> | 2019-08-21 14:10:36 -0300 |
---|---|---|
committer | Jason Gunthorpe <jgg@mellanox.com> | 2019-08-21 14:10:36 -0300 |
commit | 868df536f5e84672c3e002b949e0e44f97cb0f09 (patch) | |
tree | f76da5f6d06125b6d91c75ebfe4079ec9e2f958a /drivers/misc/habanalabs/irq.c | |
parent | RDMA: Delete DEBUG code (diff) | |
parent | RDMA/mlx5: Use odp instead of mr->umem in pagefault_mr (diff) | |
download | linux-dev-868df536f5e84672c3e002b949e0e44f97cb0f09.tar.xz linux-dev-868df536f5e84672c3e002b949e0e44f97cb0f09.zip |
Merge branch 'odp_fixes' into rdma.git for-next
Jason Gunthorpe says:
====================
This is a collection of general cleanups for ODP to clarify some of the
flows around umem creation and use of the interval tree.
====================
The branch is based on v5.3-rc5 due to dependencies
* odp_fixes:
RDMA/mlx5: Use odp instead of mr->umem in pagefault_mr
RDMA/mlx5: Use ib_umem_start instead of umem.address
RDMA/core: Make invalidate_range a device operation
RDMA/odp: Use kvcalloc for the dma_list and page_list
RDMA/odp: Check for overflow when computing the umem_odp end
RDMA/odp: Provide ib_umem_odp_release() to undo the allocs
RDMA/odp: Split creating a umem_odp from ib_umem_get
RDMA/odp: Make the three ways to create a umem_odp clear
RMDA/odp: Consolidate umem_odp initialization
RDMA/odp: Make it clearer when a umem is an implicit ODP umem
RDMA/odp: Iterate over the whole rbtree directly
RDMA/odp: Use the common interval tree library instead of generic
RDMA/mlx5: Fix MR npages calculation for IB_ACCESS_HUGETLB
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Diffstat (limited to 'drivers/misc/habanalabs/irq.c')
-rw-r--r-- | drivers/misc/habanalabs/irq.c | 27 |
1 files changed, 13 insertions, 14 deletions
diff --git a/drivers/misc/habanalabs/irq.c b/drivers/misc/habanalabs/irq.c index ea9f72ff456c..199791b57caf 100644 --- a/drivers/misc/habanalabs/irq.c +++ b/drivers/misc/habanalabs/irq.c @@ -80,8 +80,7 @@ irqreturn_t hl_irq_handler_cq(int irq, void *arg) struct hl_cs_job *job; bool shadow_index_valid; u16 shadow_index; - u32 *cq_entry; - u32 *cq_base; + struct hl_cq_entry *cq_entry, *cq_base; if (hdev->disabled) { dev_dbg(hdev->dev, @@ -90,29 +89,29 @@ irqreturn_t hl_irq_handler_cq(int irq, void *arg) return IRQ_HANDLED; } - cq_base = (u32 *) (uintptr_t) cq->kernel_address; + cq_base = (struct hl_cq_entry *) (uintptr_t) cq->kernel_address; while (1) { - bool entry_ready = ((cq_base[cq->ci] & CQ_ENTRY_READY_MASK) + bool entry_ready = ((le32_to_cpu(cq_base[cq->ci].data) & + CQ_ENTRY_READY_MASK) >> CQ_ENTRY_READY_SHIFT); if (!entry_ready) break; - cq_entry = (u32 *) &cq_base[cq->ci]; + cq_entry = (struct hl_cq_entry *) &cq_base[cq->ci]; - /* - * Make sure we read CQ entry contents after we've + /* Make sure we read CQ entry contents after we've * checked the ownership bit. */ dma_rmb(); - shadow_index_valid = - ((*cq_entry & CQ_ENTRY_SHADOW_INDEX_VALID_MASK) + shadow_index_valid = ((le32_to_cpu(cq_entry->data) & + CQ_ENTRY_SHADOW_INDEX_VALID_MASK) >> CQ_ENTRY_SHADOW_INDEX_VALID_SHIFT); - shadow_index = (u16) - ((*cq_entry & CQ_ENTRY_SHADOW_INDEX_MASK) + shadow_index = (u16) ((le32_to_cpu(cq_entry->data) & + CQ_ENTRY_SHADOW_INDEX_MASK) >> CQ_ENTRY_SHADOW_INDEX_SHIFT); queue = &hdev->kernel_queues[cq->hw_queue_id]; @@ -122,8 +121,7 @@ irqreturn_t hl_irq_handler_cq(int irq, void *arg) queue_work(hdev->cq_wq, &job->finish_work); } - /* - * Update ci of the context's queue. There is no + /* Update ci of the context's queue. There is no * need to protect it with spinlock because this update is * done only inside IRQ and there is a different IRQ per * queue @@ -131,7 +129,8 @@ irqreturn_t hl_irq_handler_cq(int irq, void *arg) queue->ci = hl_queue_inc_ptr(queue->ci); /* Clear CQ entry ready bit */ - cq_base[cq->ci] &= ~CQ_ENTRY_READY_MASK; + cq_entry->data = cpu_to_le32(le32_to_cpu(cq_entry->data) & + ~CQ_ENTRY_READY_MASK); cq->ci = hl_cq_inc_ptr(cq->ci); |