* [PATCH 1/2] habanalabs: remove call to cs_parser()
@ 2019-05-01 9:58 Oded Gabbay
2019-05-01 9:58 ` [PATCH 2/2] habanalabs: rename functions to improve code readability Oded Gabbay
0 siblings, 1 reply; 2+ messages in thread
From: Oded Gabbay @ 2019-05-01 9:58 UTC (permalink / raw)
To: linux-kernel; +Cc: gregkh, Dalit Ben Zoor
From: Dalit Ben Zoor <dbenzoor@habana.ai>
There is no need to parse the command submission when doing memset
of the device memory using the DMA engine because only the driver calls
the memset function and therefore, the CS is trusted and doesn't require
validation and patching.
Signed-off-by: Dalit Ben Zoor <dbenzoor@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
---
drivers/misc/habanalabs/goya/goya.c | 27 +++------------------------
1 file changed, 3 insertions(+), 24 deletions(-)
diff --git a/drivers/misc/habanalabs/goya/goya.c b/drivers/misc/habanalabs/goya/goya.c
index 04e4ed8a0be6..9fc8b6e1369d 100644
--- a/drivers/misc/habanalabs/goya/goya.c
+++ b/drivers/misc/habanalabs/goya/goya.c
@@ -4398,7 +4398,6 @@ static int goya_memset_device_memory(struct hl_device *hdev, u64 addr, u32 size,
u64 val, bool is_dram)
{
struct packet_lin_dma *lin_dma_pkt;
- struct hl_cs_parser parser;
struct hl_cs_job *job;
u32 cb_size, ctl;
struct hl_cb *cb;
@@ -4438,36 +4437,16 @@ static int goya_memset_device_memory(struct hl_device *hdev, u64 addr, u32 size,
job->user_cb->cs_cnt++;
job->user_cb_size = cb_size;
job->hw_queue_id = GOYA_QUEUE_ID_DMA_0;
+ job->patched_cb = job->user_cb;
+ job->job_cb_size = job->user_cb_size +
+ sizeof(struct packet_msg_prot) * 2;
hl_debugfs_add_job(hdev, job);
- parser.ctx_id = HL_KERNEL_ASID_ID;
- parser.cs_sequence = 0;
- parser.job_id = job->id;
- parser.hw_queue_id = job->hw_queue_id;
- parser.job_userptr_list = &job->userptr_list;
- parser.user_cb = job->user_cb;
- parser.user_cb_size = job->user_cb_size;
- parser.ext_queue = job->ext_queue;
- parser.use_virt_addr = hdev->mmu_enable;
-
- rc = hdev->asic_funcs->cs_parser(hdev, &parser);
- if (rc) {
- dev_err(hdev->dev, "Failed to parse kernel CB\n");
- goto free_job;
- }
-
- job->patched_cb = parser.patched_cb;
- job->job_cb_size = parser.patched_cb_size;
- job->patched_cb->cs_cnt++;
-
rc = goya_send_job_on_qman0(hdev, job);
- job->patched_cb->cs_cnt--;
hl_cb_put(job->patched_cb);
-free_job:
- hl_userptr_delete_list(hdev, &job->userptr_list);
hl_debugfs_remove_job(hdev, job);
kfree(job);
cb->cs_cnt--;
--
2.17.1
^ permalink raw reply related [flat|nested] 2+ messages in thread
* [PATCH 2/2] habanalabs: rename functions to improve code readability
2019-05-01 9:58 [PATCH 1/2] habanalabs: remove call to cs_parser() Oded Gabbay
@ 2019-05-01 9:58 ` Oded Gabbay
0 siblings, 0 replies; 2+ messages in thread
From: Oded Gabbay @ 2019-05-01 9:58 UTC (permalink / raw)
To: linux-kernel; +Cc: gregkh
This patch renames four functions in the ASIC-specific functions section,
so it will be easier to differentiate them from the generic kernel
functions with the same name.
This will help in future code reviews, to make sure we don't use the
kernel functions directly.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
---
drivers/misc/habanalabs/command_buffer.c | 6 ++---
drivers/misc/habanalabs/goya/goya.c | 28 ++++++++++++----------
drivers/misc/habanalabs/habanalabs.h | 30 ++++++++++++------------
drivers/misc/habanalabs/hw_queue.c | 6 ++---
drivers/misc/habanalabs/irq.c | 4 ++--
5 files changed, 38 insertions(+), 36 deletions(-)
diff --git a/drivers/misc/habanalabs/command_buffer.c b/drivers/misc/habanalabs/command_buffer.c
index b1ffca47d748..e495f44064fa 100644
--- a/drivers/misc/habanalabs/command_buffer.c
+++ b/drivers/misc/habanalabs/command_buffer.c
@@ -13,7 +13,7 @@
static void cb_fini(struct hl_device *hdev, struct hl_cb *cb)
{
- hdev->asic_funcs->dma_free_coherent(hdev, cb->size,
+ hdev->asic_funcs->asic_dma_free_coherent(hdev, cb->size,
(void *) (uintptr_t) cb->kernel_address,
cb->bus_address);
kfree(cb);
@@ -66,10 +66,10 @@ static struct hl_cb *hl_cb_alloc(struct hl_device *hdev, u32 cb_size,
return NULL;
if (ctx_id == HL_KERNEL_ASID_ID)
- p = hdev->asic_funcs->dma_alloc_coherent(hdev, cb_size,
+ p = hdev->asic_funcs->asic_dma_alloc_coherent(hdev, cb_size,
&cb->bus_address, GFP_ATOMIC);
else
- p = hdev->asic_funcs->dma_alloc_coherent(hdev, cb_size,
+ p = hdev->asic_funcs->asic_dma_alloc_coherent(hdev, cb_size,
&cb->bus_address,
GFP_USER | __GFP_ZERO);
if (!p) {
diff --git a/drivers/misc/habanalabs/goya/goya.c b/drivers/misc/habanalabs/goya/goya.c
index 9fc8b6e1369d..8e18c80a22e7 100644
--- a/drivers/misc/habanalabs/goya/goya.c
+++ b/drivers/misc/habanalabs/goya/goya.c
@@ -646,7 +646,7 @@ static int goya_sw_init(struct hl_device *hdev)
}
hdev->cpu_accessible_dma_mem =
- hdev->asic_funcs->dma_alloc_coherent(hdev,
+ hdev->asic_funcs->asic_dma_alloc_coherent(hdev,
HL_CPU_ACCESSIBLE_MEM_SIZE,
&hdev->cpu_accessible_dma_address,
GFP_KERNEL | __GFP_ZERO);
@@ -681,7 +681,8 @@ static int goya_sw_init(struct hl_device *hdev)
free_cpu_pq_pool:
gen_pool_destroy(hdev->cpu_accessible_dma_pool);
free_cpu_pq_dma_mem:
- hdev->asic_funcs->dma_free_coherent(hdev, HL_CPU_ACCESSIBLE_MEM_SIZE,
+ hdev->asic_funcs->asic_dma_free_coherent(hdev,
+ HL_CPU_ACCESSIBLE_MEM_SIZE,
hdev->cpu_accessible_dma_mem,
hdev->cpu_accessible_dma_address);
free_dma_pool:
@@ -704,7 +705,8 @@ static int goya_sw_fini(struct hl_device *hdev)
gen_pool_destroy(hdev->cpu_accessible_dma_pool);
- hdev->asic_funcs->dma_free_coherent(hdev, HL_CPU_ACCESSIBLE_MEM_SIZE,
+ hdev->asic_funcs->asic_dma_free_coherent(hdev,
+ HL_CPU_ACCESSIBLE_MEM_SIZE,
hdev->cpu_accessible_dma_mem,
hdev->cpu_accessible_dma_address);
@@ -2818,7 +2820,7 @@ static int goya_send_job_on_qman0(struct hl_device *hdev, struct hl_cs_job *job)
return -EBUSY;
}
- fence_ptr = hdev->asic_funcs->dma_pool_zalloc(hdev, 4, GFP_KERNEL,
+ fence_ptr = hdev->asic_funcs->asic_dma_pool_zalloc(hdev, 4, GFP_KERNEL,
&fence_dma_addr);
if (!fence_ptr) {
dev_err(hdev->dev,
@@ -2867,7 +2869,7 @@ static int goya_send_job_on_qman0(struct hl_device *hdev, struct hl_cs_job *job)
}
free_fence_ptr:
- hdev->asic_funcs->dma_pool_free(hdev, (void *) fence_ptr,
+ hdev->asic_funcs->asic_dma_pool_free(hdev, (void *) fence_ptr,
fence_dma_addr);
goya_qman0_set_security(hdev, false);
@@ -2901,7 +2903,7 @@ int goya_test_queue(struct hl_device *hdev, u32 hw_queue_id)
fence_val = GOYA_QMAN0_FENCE_VAL;
- fence_ptr = hdev->asic_funcs->dma_pool_zalloc(hdev, 4, GFP_KERNEL,
+ fence_ptr = hdev->asic_funcs->asic_dma_pool_zalloc(hdev, 4, GFP_KERNEL,
&fence_dma_addr);
if (!fence_ptr) {
dev_err(hdev->dev,
@@ -2911,7 +2913,7 @@ int goya_test_queue(struct hl_device *hdev, u32 hw_queue_id)
*fence_ptr = 0;
- fence_pkt = hdev->asic_funcs->dma_pool_zalloc(hdev,
+ fence_pkt = hdev->asic_funcs->asic_dma_pool_zalloc(hdev,
sizeof(struct packet_msg_prot),
GFP_KERNEL, &pkt_dma_addr);
if (!fence_pkt) {
@@ -2955,10 +2957,10 @@ int goya_test_queue(struct hl_device *hdev, u32 hw_queue_id)
}
free_pkt:
- hdev->asic_funcs->dma_pool_free(hdev, (void *) fence_pkt,
+ hdev->asic_funcs->asic_dma_pool_free(hdev, (void *) fence_pkt,
pkt_dma_addr);
free_fence_ptr:
- hdev->asic_funcs->dma_pool_free(hdev, (void *) fence_ptr,
+ hdev->asic_funcs->asic_dma_pool_free(hdev, (void *) fence_ptr,
fence_dma_addr);
return rc;
}
@@ -4755,12 +4757,12 @@ static const struct hl_asic_funcs goya_funcs = {
.cb_mmap = goya_cb_mmap,
.ring_doorbell = goya_ring_doorbell,
.flush_pq_write = goya_flush_pq_write,
- .dma_alloc_coherent = goya_dma_alloc_coherent,
- .dma_free_coherent = goya_dma_free_coherent,
+ .asic_dma_alloc_coherent = goya_dma_alloc_coherent,
+ .asic_dma_free_coherent = goya_dma_free_coherent,
.get_int_queue_base = goya_get_int_queue_base,
.test_queues = goya_test_queues,
- .dma_pool_zalloc = goya_dma_pool_zalloc,
- .dma_pool_free = goya_dma_pool_free,
+ .asic_dma_pool_zalloc = goya_dma_pool_zalloc,
+ .asic_dma_pool_free = goya_dma_pool_free,
.cpu_accessible_dma_pool_alloc = goya_cpu_accessible_dma_pool_alloc,
.cpu_accessible_dma_pool_free = goya_cpu_accessible_dma_pool_free,
.hl_dma_unmap_sg = goya_dma_unmap_sg,
diff --git a/drivers/misc/habanalabs/habanalabs.h b/drivers/misc/habanalabs/habanalabs.h
index 687651db614c..b64594be6dbd 100644
--- a/drivers/misc/habanalabs/habanalabs.h
+++ b/drivers/misc/habanalabs/habanalabs.h
@@ -453,19 +453,19 @@ enum hl_pll_frequency {
* @cb_mmap: maps a CB.
* @ring_doorbell: increment PI on a given QMAN.
* @flush_pq_write: flush PQ entry write if necessary, WARN if flushing failed.
- * @dma_alloc_coherent: Allocate coherent DMA memory by calling
- * dma_alloc_coherent(). This is ASIC function because its
- * implementation is not trivial when the driver is loaded
- * in simulation mode (not upstreamed).
- * @dma_free_coherent: Free coherent DMA memory by calling dma_free_coherent().
- * This is ASIC function because its implementation is not
- * trivial when the driver is loaded in simulation mode
- * (not upstreamed).
+ * @asic_dma_alloc_coherent: Allocate coherent DMA memory by calling
+ * dma_alloc_coherent(). This is ASIC function because
+ * its implementation is not trivial when the driver
+ * is loaded in simulation mode (not upstreamed).
+ * @asic_dma_free_coherent: Free coherent DMA memory by calling
+ * dma_free_coherent(). This is ASIC function because
+ * its implementation is not trivial when the driver
+ * is loaded in simulation mode (not upstreamed).
* @get_int_queue_base: get the internal queue base address.
* @test_queues: run simple test on all queues for sanity check.
- * @dma_pool_zalloc: small DMA allocation of coherent memory from DMA pool.
- * size of allocation is HL_DMA_POOL_BLK_SIZE.
- * @dma_pool_free: free small DMA allocation from pool.
+ * @asic_dma_pool_zalloc: small DMA allocation of coherent memory from DMA pool.
+ * size of allocation is HL_DMA_POOL_BLK_SIZE.
+ * @asic_dma_pool_free: free small DMA allocation from pool.
* @cpu_accessible_dma_pool_alloc: allocate CPU PQ packet from DMA pool.
* @cpu_accessible_dma_pool_free: free CPU PQ packet from DMA pool.
* @hl_dma_unmap_sg: DMA unmap scatter-gather list.
@@ -521,16 +521,16 @@ struct hl_asic_funcs {
u64 kaddress, phys_addr_t paddress, u32 size);
void (*ring_doorbell)(struct hl_device *hdev, u32 hw_queue_id, u32 pi);
void (*flush_pq_write)(struct hl_device *hdev, u64 *pq, u64 exp_val);
- void* (*dma_alloc_coherent)(struct hl_device *hdev, size_t size,
+ void* (*asic_dma_alloc_coherent)(struct hl_device *hdev, size_t size,
dma_addr_t *dma_handle, gfp_t flag);
- void (*dma_free_coherent)(struct hl_device *hdev, size_t size,
+ void (*asic_dma_free_coherent)(struct hl_device *hdev, size_t size,
void *cpu_addr, dma_addr_t dma_handle);
void* (*get_int_queue_base)(struct hl_device *hdev, u32 queue_id,
dma_addr_t *dma_handle, u16 *queue_len);
int (*test_queues)(struct hl_device *hdev);
- void* (*dma_pool_zalloc)(struct hl_device *hdev, size_t size,
+ void* (*asic_dma_pool_zalloc)(struct hl_device *hdev, size_t size,
gfp_t mem_flags, dma_addr_t *dma_handle);
- void (*dma_pool_free)(struct hl_device *hdev, void *vaddr,
+ void (*asic_dma_pool_free)(struct hl_device *hdev, void *vaddr,
dma_addr_t dma_addr);
void* (*cpu_accessible_dma_pool_alloc)(struct hl_device *hdev,
size_t size, dma_addr_t *dma_handle);
diff --git a/drivers/misc/habanalabs/hw_queue.c b/drivers/misc/habanalabs/hw_queue.c
index a1ee52cfd505..6cdaa117fc40 100644
--- a/drivers/misc/habanalabs/hw_queue.c
+++ b/drivers/misc/habanalabs/hw_queue.c
@@ -425,7 +425,7 @@ static int ext_and_cpu_hw_queue_init(struct hl_device *hdev,
HL_QUEUE_SIZE_IN_BYTES,
&q->bus_address);
else
- p = hdev->asic_funcs->dma_alloc_coherent(hdev,
+ p = hdev->asic_funcs->asic_dma_alloc_coherent(hdev,
HL_QUEUE_SIZE_IN_BYTES,
&q->bus_address,
GFP_KERNEL | __GFP_ZERO);
@@ -457,7 +457,7 @@ static int ext_and_cpu_hw_queue_init(struct hl_device *hdev,
HL_QUEUE_SIZE_IN_BYTES,
(void *) (uintptr_t) q->kernel_address);
else
- hdev->asic_funcs->dma_free_coherent(hdev,
+ hdev->asic_funcs->asic_dma_free_coherent(hdev,
HL_QUEUE_SIZE_IN_BYTES,
(void *) (uintptr_t) q->kernel_address,
q->bus_address);
@@ -587,7 +587,7 @@ static void hw_queue_fini(struct hl_device *hdev, struct hl_hw_queue *q)
HL_QUEUE_SIZE_IN_BYTES,
(void *) (uintptr_t) q->kernel_address);
else
- hdev->asic_funcs->dma_free_coherent(hdev,
+ hdev->asic_funcs->asic_dma_free_coherent(hdev,
HL_QUEUE_SIZE_IN_BYTES,
(void *) (uintptr_t) q->kernel_address,
q->bus_address);
diff --git a/drivers/misc/habanalabs/irq.c b/drivers/misc/habanalabs/irq.c
index 86a8ad57f1ca..ea9f72ff456c 100644
--- a/drivers/misc/habanalabs/irq.c
+++ b/drivers/misc/habanalabs/irq.c
@@ -222,7 +222,7 @@ int hl_cq_init(struct hl_device *hdev, struct hl_cq *q, u32 hw_queue_id)
BUILD_BUG_ON(HL_CQ_SIZE_IN_BYTES > HL_PAGE_SIZE);
- p = hdev->asic_funcs->dma_alloc_coherent(hdev, HL_CQ_SIZE_IN_BYTES,
+ p = hdev->asic_funcs->asic_dma_alloc_coherent(hdev, HL_CQ_SIZE_IN_BYTES,
&q->bus_address, GFP_KERNEL | __GFP_ZERO);
if (!p)
return -ENOMEM;
@@ -248,7 +248,7 @@ int hl_cq_init(struct hl_device *hdev, struct hl_cq *q, u32 hw_queue_id)
*/
void hl_cq_fini(struct hl_device *hdev, struct hl_cq *q)
{
- hdev->asic_funcs->dma_free_coherent(hdev, HL_CQ_SIZE_IN_BYTES,
+ hdev->asic_funcs->asic_dma_free_coherent(hdev, HL_CQ_SIZE_IN_BYTES,
(void *) (uintptr_t) q->kernel_address, q->bus_address);
}
--
2.17.1
^ permalink raw reply related [flat|nested] 2+ messages in thread
end of thread, other threads:[~2019-05-01 9:58 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-05-01 9:58 [PATCH 1/2] habanalabs: remove call to cs_parser() Oded Gabbay
2019-05-01 9:58 ` [PATCH 2/2] habanalabs: rename functions to improve code readability Oded Gabbay
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).