linux-kernel - [PATCH 2/7] habanalabs: perform context switch flow only if needed

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-Id: <20220829092254.930753-2-ogabbay@kernel.org>
Date:   Mon, 29 Aug 2022 12:22:49 +0300
From:   Oded Gabbay <ogabbay@...nel.org>
To:     linux-kernel@...r.kernel.org
Cc:     Ofir Bitton <obitton@...ana.ai>
Subject: [PATCH 2/7] habanalabs: perform context switch flow only if needed

From: Ofir Bitton <obitton@...ana.ai>

Except Goya, none of our ASICs require context switch flow, hence we
enable this flow only where it is needed.

Signed-off-by: Ofir Bitton <obitton@...ana.ai>
Reviewed-by: Oded Gabbay <ogabbay@...nel.org>
Signed-off-by: Oded Gabbay <ogabbay@...nel.org>
---
 drivers/misc/habanalabs/common/command_submission.c | 10 ++++++----
 drivers/misc/habanalabs/common/habanalabs.h         |  2 ++
 drivers/misc/habanalabs/goya/goya.c                 |  1 +
 3 files changed, 9 insertions(+), 4 deletions(-)

diff --git a/drivers/misc/habanalabs/common/command_submission.c b/drivers/misc/habanalabs/common/command_submission.c
index cf4118515678..746b688d34cf 100644
--- a/drivers/misc/habanalabs/common/command_submission.c
+++ b/drivers/misc/habanalabs/common/command_submission.c
@@ -1590,13 +1590,14 @@ static int hl_cs_ctx_switch(struct hl_fpriv *hpriv, union hl_cs_args *args,
 	struct hl_device *hdev = hpriv->hdev;
 	struct hl_ctx *ctx = hpriv->ctx;
 	bool need_soft_reset = false;
-	int rc = 0, do_ctx_switch;
+	int rc = 0, do_ctx_switch = 0;
 	void __user *chunks;
 	u32 num_chunks, tmp;
 	u16 sob_count;
 	int ret;
 
-	do_ctx_switch = atomic_cmpxchg(&ctx->thread_ctx_switch_token, 1, 0);
+	if (hdev->supports_ctx_switch)
+		do_ctx_switch = atomic_cmpxchg(&ctx->thread_ctx_switch_token, 1, 0);
 
 	if (do_ctx_switch || (args->in.cs_flags & HL_CS_FLAGS_FORCE_RESTORE)) {
 		mutex_lock(&hpriv->restore_phase_mutex);
@@ -1667,9 +1668,10 @@ static int hl_cs_ctx_switch(struct hl_fpriv *hpriv, union hl_cs_args *args,
 			}
 		}
 
-		ctx->thread_ctx_switch_wait_token = 1;
+		if (hdev->supports_ctx_switch)
+			ctx->thread_ctx_switch_wait_token = 1;
 
-	} else if (!ctx->thread_ctx_switch_wait_token) {
+	} else if (hdev->supports_ctx_switch && !ctx->thread_ctx_switch_wait_token) {
 		rc = hl_poll_timeout_memory(hdev,
 			&ctx->thread_ctx_switch_wait_token, tmp, (tmp == 1),
 			100, jiffies_to_usecs(hdev->timeout_jiffies), false);
diff --git a/drivers/misc/habanalabs/common/habanalabs.h b/drivers/misc/habanalabs/common/habanalabs.h
index d7fd4f57abf3..33c6476b60a9 100644
--- a/drivers/misc/habanalabs/common/habanalabs.h
+++ b/drivers/misc/habanalabs/common/habanalabs.h
@@ -3199,6 +3199,7 @@ struct hl_reset_info {
  *                         Used only for testing.
  * @heartbeat: Controls if we want to enable the heartbeat mechanism vs. the f/w, which verifies
  *             that the f/w is always alive. Used only for testing.
+ * @supports_ctx_switch: true if a ctx switch is required upon first submission.
  */
 struct hl_device {
 	struct pci_dev			*pdev;
@@ -3335,6 +3336,7 @@ struct hl_device {
 	u8				compute_ctx_in_release;
 	u8				supports_mmu_prefetch;
 	u8				reset_upon_device_release;
+	u8				supports_ctx_switch;
 
 	/* Parameters for bring-up */
 	u64				nic_ports_mask;
diff --git a/drivers/misc/habanalabs/goya/goya.c b/drivers/misc/habanalabs/goya/goya.c
index 87465a28af0d..d8fb91d257b9 100644
--- a/drivers/misc/habanalabs/goya/goya.c
+++ b/drivers/misc/habanalabs/goya/goya.c
@@ -1025,6 +1025,7 @@ static int goya_sw_init(struct hl_device *hdev)
 	hdev->asic_prop.supports_compute_reset = true;
 	hdev->asic_prop.allow_inference_soft_reset = true;
 	hdev->supports_wait_for_multi_cs = false;
+	hdev->supports_ctx_switch = true;
 
 	hdev->asic_funcs->set_pci_memory_regions(hdev);
 
-- 
2.25.1