AMD-GFX Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Alex Deucher <alexander.deucher@amd.com>
To: <amd-gfx@lists.freedesktop.org>
Cc: Mukul Joshi <mukul.joshi@amd.com>,
	Felix Kuehling <felix.kuehling@amd.com>,
	Alex Deucher <alexander.deucher@amd.com>
Subject: [PATCH 57/95] drm/amdgpu: Send TLB shootdown on exported memory unmap
Date: Fri, 21 Aug 2026 15:34:20 -0400	[thread overview]
Message-ID: <20260821193458.808626-58-alexander.deucher@amd.com> (raw)
In-Reply-To: <20260821193458.808626-1-alexander.deucher@amd.com>

From: Mukul Joshi <mukul.joshi@amd.com>

Replace the stub implementation of TLB shootdown message with
the actual implementation.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.c | 66 ++++++++++++++++------
 drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.h |  3 +
 2 files changed, 52 insertions(+), 17 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.c
index 2f7df10f39423..57332243ea027 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.c
@@ -47,6 +47,9 @@ static void amdgpu_ualink_handle_connection_reset(struct amdgpu_device *adev,
 						  u32 remote_accel_id, u32 state,
 						  u32 generation_count);
 static void amdgpu_ualink_invalidate_import_mappings(struct amdgpu_bo *bo);
+static int amdgpu_ualink_remote_shootdown(struct amdgpu_device *adev,
+					  u32 remote_accel_id, u64 addr,
+					  u32 size_in_pages, u32 flush_type);
 
 #define STRIP_NPA(addr)						\
 	(((u64)(addr) & ~AMDGPU_UALINK_NPA_ADDR_GPUID_MASK))
@@ -1401,12 +1404,6 @@ static int amdgpu_ualink_send_npa_req_msg(struct amdgpu_device *adev,
 					      dw2, dw3);
 }
 
-static int amdgpu_ualink_send_tlb_shootdown(struct amdgpu_device *adev,
-					    u32 remote_acc_id)
-{
-	return 0;
-}
-
 static u64 amdgpu_ualink_get_export_pte_flags(struct amdgpu_device *adev,
 				       struct amdgpu_bo *bo,
 				       u64 mapping_flags)
@@ -2005,6 +2002,51 @@ static void amdgpu_ualink_force_retry_rpcs(struct amdgpu_device *adev,
 	amdgpu_ualink_flush_tlb(adev, TLB_FLUSH_HEAVYWEIGHT);
 }
 
+static void amdgpu_ualink_send_tlb_shootdown(struct amdgpu_device *adev,
+				struct amdgpu_ualink_exp_xa_node *exp_xa_node)
+{
+	u32 addr_mode = adev->ualink.info->vpod.addr_mode;
+	u32 accel_id = adev->ualink.info->ppod.accel_id;
+	struct amdgpu_ualink_importer_entry *imp_entry;
+	u64 npa_addr, size;
+	u32 remote_acc_id;
+	int r;
+
+	size = amdgpu_bo_ngpu_pages(exp_xa_node->bo);
+
+	for_each_set_bit(remote_acc_id, exp_xa_node->importers_bitmap,
+			 AMDGPU_UALINK_ACCEL_MAX) {
+		/*
+		 * Replace the GPU-id bits with local accel_id
+		 */
+		if (addr_mode == AMDGPU_UALINK_ADDR_MODE_SOURCE_ALIAS)
+			imp_entry = &exp_xa_node->importer_entries[0];
+		else
+			imp_entry = &exp_xa_node->importer_entries[remote_acc_id];
+
+		npa_addr = STRIP_NPA(imp_entry->npa_addr);
+		npa_addr = GENERATE_NPA(npa_addr, accel_id);
+
+		if (!amdgpu_ualink_check_conn_ready(adev, remote_acc_id,
+					imp_entry->generation_count)) {
+			clear_bit(remote_acc_id,
+				  exp_xa_node->importers_bitmap);
+			continue;
+		}
+
+		dev_dbg(adev->dev,
+			"EXP-CLEANUP: Sending TLB-shootdown to remote:%u\n",
+			remote_acc_id);
+		r = amdgpu_ualink_remote_shootdown(adev, remote_acc_id,
+					npa_addr, size,
+					AMDGPU_UALINK_HEAVYWEIGHT_TLB_SHOOTDOWN);
+		if (r)
+			dev_err(adev->dev,
+				"EXP-CLEANUP: TLB shootdown send failed to remote:%u\n",
+				remote_acc_id);
+	}
+}
+
 /* Unmap all NPA addresses associated with a BO (UALink handle). This function is used
  * only in Source Identification mode.
  */
@@ -2163,17 +2205,7 @@ static void amdgpu_ualink_exp_cleanup_worker(struct work_struct *work)
 	amdgpu_ualink_force_retry_rpcs(adev, exp_xa_node);
 
 	/* Send TLB-shootdown to all importer GPUs */
-	for_each_set_bit(remote_acc_id, exp_xa_node->importers_bitmap,
-			 AMDGPU_UALINK_ACCEL_MAX) {
-		dev_dbg(adev->dev,
-			"EXP-CLEANUP: Sending TLB-shootdown to remote:%u\n",
-			remote_acc_id);
-		r = amdgpu_ualink_send_tlb_shootdown(adev, remote_acc_id);
-		if (r)
-			dev_err(adev->dev,
-				"EXP-CLEANUP: TLB shootdown send failed to remote:%u\n",
-				remote_acc_id);
-	}
+	amdgpu_ualink_send_tlb_shootdown(adev, exp_xa_node);
 
 	/* Unmap all NPA addresses for this BO from NPA VM */
 	amdgpu_ualink_unmap_all_npa_addr(adev, exp_xa_node);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.h
index bd8af7596a6ef..a1fd5fe953115 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.h
@@ -78,6 +78,9 @@ enum AMDGPU_UALINK_PROTOCOL_MESSAGES {
 	AMDGPU_UALINK_MAX_PROTOCOL_MSG
 };
 
+#define AMDGPU_UALINK_LIGHTWEIGHT_TLB_SHOOTDOWN		0x1
+#define AMDGPU_UALINK_HEAVYWEIGHT_TLB_SHOOTDOWN		0x2
+
 enum amdgpu_ualink_conn_state {
 	AMDGPU_UALINK_CONN_NOT_READY			= 0,
 	AMDGPU_UALINK_CONN_IN_PROGRESS			= 1,
-- 
2.55.0


  parent reply	other threads:[~2026-08-21 20:00 UTC|newest]

Thread overview: 98+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-08-21 19:33 [PATCH 00/95] Add UALink instrastructure series 1 Alex Deucher
2026-08-21 19:33 ` [PATCH 01/95] drm/amdgpu: Add psp ualink command interfaces Alex Deucher
2026-08-21 19:33 ` [PATCH 02/95] drm/amdgpu: Fetch asp ualink interface version Alex Deucher
2026-08-21 19:33 ` [PATCH 03/95] drm/amdgpu: Add sysfs API for UALink information Alex Deucher
2026-08-21 19:33 ` [PATCH 04/95] drm/amdgpu: Add sysfs API for UALink physical pod setup Alex Deucher
2026-08-21 19:33 ` [PATCH 05/95] drm/amdgpu: Add sysfs API for UALink virtual pod config Alex Deucher
2026-08-21 19:33 ` [PATCH 06/95] drm/amdgpu: Add sysfs API for UALink station configuration Alex Deucher
2026-08-21 19:33 ` [PATCH 07/95] drm/amdgpu: Add UALink manager core infrastructure Alex Deucher
2026-08-21 19:33 ` [PATCH 08/95] drm/amdgpu: Implement PSP cmd UAL_GET_CONFIG Alex Deucher
2026-08-21 19:33 ` [PATCH 09/95] drm/amdgpu: Query initial UALink config from PSP Alex Deucher
2026-08-21 19:33 ` [PATCH 10/95] drm/amdgpu: Implement PSP cmd UAL_SET_PPOD_CONFIG Alex Deucher
2026-08-21 19:33 ` [PATCH 11/95] drm/amdgpu: Set physical pod configuration to PSP Alex Deucher
2026-08-21 19:33 ` [PATCH 12/95] drm/amdgpu: Implement PSP cmd UAL_SET_VPOD_CONFIG Alex Deucher
2026-08-21 19:33 ` [PATCH 13/95] drm/amdgpu: Set virtual pod configuration to PSP Alex Deucher
2026-08-21 19:33 ` [PATCH 14/95] drm/amdgpu: Implement PSP cmd UAL_SET_STATION_CONFIG Alex Deucher
2026-08-21 19:33 ` [PATCH 15/95] drm/amdgpu: Set UALink station config to PSP Alex Deucher
2026-08-21 19:33 ` [PATCH 16/95] drm/amdgpu: Implement PSP cmd UAL_SET_NPA_CONFIG Alex Deucher
2026-08-21 19:33 ` [PATCH 17/95] drm/amdgpu: Enable/disable NPA address translation using PSP Alex Deucher
2026-08-21 19:33 ` [PATCH 18/95] drm/amdgpu: Add helper function to check psp xgmi ta Alex Deucher
2026-08-21 19:33 ` [PATCH 19/95] drm/amdgpu: add handler for nHT error Alex Deucher
2026-08-21 19:33 ` [PATCH 20/95] drm/amdgpu: Add ual_config_state to ual_get_config Alex Deucher
2026-08-21 19:33 ` [PATCH 21/95] drm/amdgpu: extend PSP command polling sleep range Alex Deucher
2026-08-21 19:33 ` [PATCH 22/95] drm/amdgpu: Fix NULL pointer issue during ualink init Alex Deucher
2026-08-21 19:33 ` [PATCH 23/95] drm/amdgpu: Add a new NPA Address space Alex Deucher
2026-08-21 19:33 ` [PATCH 24/95] drm/amdgpu: Add address allocator for NPA addresses Alex Deucher
2026-08-21 19:33 ` [PATCH 25/95] drm/amdgpu: Initialize VM for NPA addr management Alex Deucher
2026-08-21 19:33 ` [PATCH 26/95] drm/amdgpu: Rework VMID reservation logic Alex Deucher
2026-08-21 19:33 ` [PATCH 27/95] drm/amdgpu: Reserve VMID for NPA VM Alex Deucher
2026-08-21 19:33 ` [PATCH 28/95] drm/amdgpu: Use reserved " Alex Deucher
2026-08-21 19:33 ` [PATCH 29/95] drm/amdgpu: Enable UALink Manager when pod becomes active Alex Deucher
2026-08-21 19:33 ` [PATCH 30/95] drm/amdgpu: Fix UALink vPod double-activation Alex Deucher
2026-08-21 19:33 ` [PATCH 31/95] drm/amdgpu: Add UALink remote state structures and API declarations Alex Deucher
2026-08-21 19:33 ` [PATCH 32/95] drm/amdgpu: Add UALink NPA address layout helpers Alex Deucher
2026-08-21 19:33 ` [PATCH 33/95] drm/amdgpu: Add UALink NPA address computation for ring buffers Alex Deucher
2026-08-21 19:33 ` [PATCH 34/95] drm/amdgpu: Add UALink NPA VM mapping " Alex Deucher
2026-08-21 19:33 ` [PATCH 35/95] drm/amdgpu: Add UALink SDMA scheduler entities Alex Deucher
2026-08-21 19:33 ` [PATCH 36/95] drm/amdgpu: Add UALink GART helpers for NPA address access Alex Deucher
2026-08-21 19:34 ` [PATCH 37/95] drm/amdgpu: Add UALink ring buffer allocation and firmware init Alex Deucher
2026-08-21 19:34 ` [PATCH 38/95] drm/amdgpu: Add UALink remote command packets and SDMA dispatch Alex Deucher
2026-08-21 19:34 ` [PATCH 39/95] drm/amdgpu: Add UALink firmware writeback address configuration Alex Deucher
2026-08-21 19:34 ` [PATCH 40/95] drm/amdgpu: Add UALink cross-GPU TLB shootdown and remote interrupt Alex Deucher
2026-08-21 19:34 ` [PATCH 41/95] drm/amdgpu: Add UALink software init, teardown, and reset Alex Deucher
2026-08-21 19:34 ` [PATCH 42/95] drm/amdgpu: Add UALink IH ring and enable interrupt Alex Deucher
2026-08-21 19:34 ` [PATCH 43/95] drm/amdgpu: UALink use LSDMA to send remote interrupt command Alex Deucher
2026-08-21 19:34 ` [PATCH 44/95] drm/amdgpu: Create a drm client for UALink NPA BOs Alex Deucher
2026-08-21 19:34 ` [PATCH 45/95] drm/amdgpu: Control NPA DMA-buf importing Alex Deucher
2026-08-21 19:34 ` [PATCH 46/95] drm/amdgpu: Add ualink handle to BOs Alex Deucher
2026-08-21 19:34 ` [PATCH 47/95] drm/amdgpu: Implement UALink handle export Alex Deucher
2026-08-21 19:34 ` [PATCH 48/95] drm/amdgpu: Add connection state management Alex Deucher
2026-08-21 19:34 ` [PATCH 49/95] drm/amdgpu: Implement UALink handle import ioctl Alex Deucher
2026-08-21 19:34 ` [PATCH 50/95] drm/amdgpu: Implement mechanism to revoke exported memory Alex Deucher
2026-08-21 19:34 ` [PATCH 51/95] drm/amdgpu: lock UALink import invalidation via drm_exec Alex Deucher
2026-08-21 19:34 ` [PATCH 52/95] drm/amdgpu: Cleanup exported UALink handles Alex Deucher
2026-08-21 19:34 ` [PATCH 53/95] drm/amdgpu: Cleanup imported " Alex Deucher
2026-08-21 19:34 ` [PATCH 54/95] drm/amdgpu: Handle connection reset Alex Deucher
2026-08-21 19:34 ` [PATCH 55/95] drm/amdgpu: Setup PTE mappings for NPA addresses Alex Deucher
2026-08-21 19:34 ` [PATCH 56/95] drm/amdgpu: Add handling for remote interrupts Alex Deucher
2026-08-21 19:34 ` Alex Deucher [this message]
2026-08-21 19:34 ` [PATCH 58/95] drm/amdgpu: Handle local GPUs in UALink import Alex Deucher
2026-08-21 19:34 ` [PATCH 59/95] drm/amdgpu: Add debugfs to drop UALink protocol messages Alex Deucher
2026-08-21 19:34 ` [PATCH 60/95] drm/amdgpu: Temporarily disable sending remote TLB shootdowns Alex Deucher
2026-08-21 19:34 ` [PATCH 61/95] drm/amdgpu: Temporarily Flush TLB on NPA mapping always Alex Deucher
2026-08-21 19:34 ` [PATCH 62/95] drm/amdgpu: log remote memory MTYPE for GC 12.1.0 Alex Deucher
2026-08-21 19:34 ` [PATCH 63/95] drm/amdgpu: Prevent double-free of drm_exec Alex Deucher
2026-08-21 19:34 ` [PATCH 64/95] drm/amdgpu: fix NPA-RELEASE race in UALink exporter cleanup Alex Deucher
2026-08-21 19:34 ` [PATCH 65/95] drm/amdgpu: initialize UALink importer node list head Alex Deucher
2026-08-21 19:34 ` [PATCH 66/95] drm/amdgpu: Fix initialization flags for UALink XAs Alex Deucher
2026-08-21 19:34 ` [PATCH 67/95] drm/amdgpu: fix dma_buf leak in UALink exporter cleanup Alex Deucher
2026-08-21 19:34 ` [PATCH 68/95] drm/amdgpu: Increase UALink soft ring size Alex Deucher
2026-08-21 19:34 ` [PATCH 69/95] drm/amdgpu: Fix uninitialized fence in UALink NPA unmap Alex Deucher
2026-08-21 19:34 ` [PATCH 70/95] drm/amdgpu: Use vm->last_update fence in UALink NPA unmap paths Alex Deucher
2026-08-21 19:34 ` [PATCH 71/95] drm/amdgpu: Pin page tables in NPA VMs Alex Deucher
2026-08-21 19:34 ` [PATCH 72/95] drm/amdgpu: Initialize NPA PT/PDs to noretry Alex Deucher
2026-08-21 19:34 ` [PATCH 73/95] drm/amdgpu: always use MTYPE_UC for remote memory on GFX 12.1 Alex Deucher
2026-08-21 19:34 ` [PATCH 74/95] drm/amdkfd: program compute MQD coherent_aql_mtype " Alex Deucher
2026-08-21 19:34 ` [PATCH 75/95] drm/amdgpu: Separate out ualink init sequences Alex Deucher
2026-08-21 19:34 ` [PATCH 76/95] drm/amdgpu: Add ualink as separate ip block Alex Deucher
2026-08-21 19:34 ` [PATCH 77/95] drm/admgpu: Seggregate ualink nht messaging Alex Deucher
2026-08-21 19:34 ` [PATCH 78/95] drm/amdgpu: Assign accel state based on ASP config Alex Deucher
2026-08-21 19:34 ` [PATCH 79/95] drm/amdgpu: Drop duplicate vpod check functions Alex Deucher
2026-08-21 19:34 ` [PATCH 80/95] drm/amdgpu: Add support to send ASP completion Alex Deucher
2026-08-21 19:34 ` [PATCH 81/95] drm/amdgpu: Add handlers for ualink notifications Alex Deucher
2026-08-21 19:34 ` [PATCH 82/95] drm/amdgpu: Improve ualink state transitions Alex Deucher
2026-08-21 19:34 ` [PATCH 83/95] drm/amdgpu: Use uniform logic for inband/sideband Alex Deucher
2026-08-21 19:34 ` [PATCH 84/95] drm/amdgpu: Fix GART and SDMA entity leak on vPod reconfiguration Alex Deucher
2026-08-21 19:34 ` [PATCH 85/95] drm/amdgpu: Add UALink diagnostic logging for vpod commit/activation Alex Deucher
2026-08-21 19:34 ` [PATCH 86/95] drm/amdgpu: Handle UALink vPod reconfiguration while ACTIVE Alex Deucher
2026-08-21 19:34 ` [PATCH 87/95] drm/amdgpu: Add name for ualink ip block Alex Deucher
2026-08-21 19:34 ` [PATCH 88/95] drm/amdgpu: Cleanup UALink XA entries on manager stop Alex Deucher
2026-08-21 19:34 ` [PATCH 89/95] drm/amdgpu: Move ualink ip version related changes Alex Deucher
2026-08-21 19:34 ` [PATCH 90/95] drm/amdgpu: Add hw_fini for ualink Alex Deucher
2026-08-21 19:34 ` [PATCH 91/95] drm/amdgpu: Expose ualink info under each xcp Alex Deucher
2026-08-21 19:34 ` [PATCH 92/95] drm/amdgpu: Handle concurrent UALINK handle import race Alex Deucher
2026-08-21 19:34 ` [PATCH 93/95] drm/amdgpu: create UALink NPA import BO directly in the NPA domain Alex Deucher
2026-08-21 19:34 ` [PATCH 94/95] drm/amdgpu: add mtype_remote module parameter Alex Deucher
2026-08-21 19:34 ` [PATCH 95/95] drm/amdgpu: Honor mtype overrides for NPA remote memory Alex Deucher
2026-08-25 14:45 ` [PATCH 00/95] Add UALink instrastructure series 1 Philip Yang
  -- strict thread matches above, loose matches on Subject: below --
2026-08-31 18:26 [PATCH V2 00/95] Add UALink infrastructure " Alex Deucher
2026-08-31 18:27 ` [PATCH 57/95] drm/amdgpu: Send TLB shootdown on exported memory unmap Alex Deucher

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20260821193458.808626-58-alexander.deucher@amd.com \
    --to=alexander.deucher@amd.com \
    --cc=amd-gfx@lists.freedesktop.org \
    --cc=felix.kuehling@amd.com \
    --cc=mukul.joshi@amd.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox