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 56/95] drm/amdgpu: Add handling for remote interrupts
Date: Fri, 21 Aug 2026 15:34:19 -0400 [thread overview]
Message-ID: <20260821193458.808626-57-alexander.deucher@amd.com> (raw)
In-Reply-To: <20260821193458.808626-1-alexander.deucher@amd.com>
From: Mukul Joshi <mukul.joshi@amd.com>
Add interrupt handling to process the NPA remote interrupts.
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 | 114 ++++++++++++++++++++-
drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.h | 3 +
2 files changed, 113 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.c
index c393633e69443..2f7df10f39423 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.c
@@ -5189,10 +5189,31 @@ void amdgpu_ualink_sw_fini(struct amdgpu_device *adev)
adev->ualink.remote = NULL;
}
+int ualink_send_hello(struct amdgpu_device *adev, u32 remote_accel_id)
+{
+ return amdgpu_ualink_send_hello_msg(adev, remote_accel_id);
+}
+
+/*
+ * The low bits of handle_lo carry the message-header field on the wire.
+ * Strip those bits and splice in @acc_id to recover the
+ * fully-qualified handle_lo value.
+ */
+static inline u64 amdgpu_ualink_reassemble_handle(u64 handle_lo, u32 acc_id)
+{
+ return (handle_lo & ~AMDGPU_UALINK_MESSAGE_HEADER_MASK) |
+ (acc_id & AMDGPU_UALINK_HANDLE_ACCID_MASK);
+}
+
static int amdgpu_ualink_process_irq(struct amdgpu_device *adev,
struct amdgpu_irq_src *source,
struct amdgpu_iv_entry *entry)
{
+ u32 sender_acc_id, receiver_acc_id, msg_type, src_acc_id;
+ struct amdgpu_ualink_handle handle;
+ u32 size, npa_addr, fail_reason;
+ u32 dw0, dw1, dw2, dw3;
+ u32 local_acc_id;
int handled = 1;
dev_dbg(adev->dev, "%s client_id 0x%x src_id 0x%x ih\n",
@@ -5206,10 +5227,95 @@ static int amdgpu_ualink_process_irq(struct amdgpu_device *adev,
return handled;
}
- /*
- * Call amdgpu_ualink_interrupt handler
- * amdgpu_ualink_interrupt(adev, entry);
- */
+ /* ContextID 4 dwords */
+ src_acc_id = entry->pasid;
+ dw0 = entry->src_data[0];
+ dw1 = entry->src_data[1];
+ dw2 = entry->src_data[2];
+ dw3 = entry->src_data[3];
+
+ dev_dbg(adev->dev, "src accel_id %u context id 0x%x 0x%x 0x%x 0x%x\n",
+ src_acc_id, dw0, dw1, dw2, dw3);
+
+ msg_type = dw0 & AMDGPU_UALINK_MESSAGE_HEADER_MASK;
+ local_acc_id = adev->ualink.info->ppod.accel_id;
+ dev_dbg(adev->dev, "Got MSG: remote acc_id %u msg_type %u\n",
+ src_acc_id, msg_type);
+
+ switch (msg_type) {
+ case AMDGPU_UALINK_HELLO_MSG:
+ receiver_acc_id = (dw0 >> AMDGPU_UALINK_HELLO_MSG_RECV_ACCID_SHIFT) &
+ AMDGPU_UALINK_HELLO_MSG_ACCID_MASK;
+ sender_acc_id = (dw0 >> AMDGPU_UALINK_HELLO_MSG_SENDER_ACCID_SHIFT) &
+ AMDGPU_UALINK_HELLO_MSG_ACCID_MASK;
+ dev_dbg(adev->dev,
+ "Got HELLO MSG: src acc_id %u receiver_acc_id %u sender_acc_id %u\n",
+ src_acc_id, receiver_acc_id, sender_acc_id);
+ amdgpu_ualink_process_hello_msg(adev, receiver_acc_id, sender_acc_id,
+ src_acc_id);
+ break;
+ case AMDGPU_UALINK_HELLO_ACK_MSG:
+ dev_dbg(adev->dev,
+ "Got HELLO-ACK MSG: remote acc_id %u\n", src_acc_id);
+ amdgpu_ualink_process_hello_ack_msg(adev, src_acc_id);
+ break;
+ case AMDGPU_UALINK_NPA_REQ_MSG:
+ memcpy(&handle, entry->src_data, sizeof(struct amdgpu_ualink_handle));
+ handle.handle_lo = amdgpu_ualink_reassemble_handle(handle.handle_lo,
+ local_acc_id);
+ dev_dbg(adev->dev,
+ "Got NPA-REQ MSG: remote acc_id %u handle %llx:%llx\n",
+ src_acc_id, handle.handle_hi, handle.handle_lo);
+ amdgpu_ualink_process_npa_req_msg(adev, src_acc_id, handle);
+ break;
+ case AMDGPU_UALINK_NPA_RSP_MSG:
+ memcpy(&handle.handle_lo, entry->src_data, sizeof(handle.handle_lo));
+ handle.handle_lo = amdgpu_ualink_reassemble_handle(handle.handle_lo,
+ src_acc_id);
+ size = entry->src_data[2];
+ npa_addr = entry->src_data[3];
+ dev_dbg(adev->dev,
+ "Got NPA-RSP MSG: remote acc_id %u handle_lo %llx size 0x%x npa_addr 0x%x\n",
+ src_acc_id, handle.handle_lo, size, npa_addr);
+ amdgpu_ualink_process_npa_rsp_msg(adev, src_acc_id,
+ handle.handle_lo,
+ npa_addr, size);
+ break;
+ case AMDGPU_UALINK_NPA_FAIL_MSG:
+ memcpy(&handle.handle_lo, entry->src_data, sizeof(handle.handle_lo));
+ handle.handle_lo = amdgpu_ualink_reassemble_handle(handle.handle_lo,
+ src_acc_id);
+ fail_reason = entry->src_data[2] &
+ AMDGPU_UALINK_NPA_FAIL_MSG_FAIL_REASON_MASK;
+ dev_dbg(adev->dev,
+ "Got NPA-FAIL MSG: remote acc_id %u handle_lo %llx fail_reason %u\n",
+ src_acc_id, handle.handle_lo, fail_reason);
+ amdgpu_ualink_process_npa_fail_msg(adev, src_acc_id,
+ handle.handle_lo,
+ fail_reason);
+ break;
+ case AMDGPU_UALINK_NPA_REVOKE_MSG:
+ memcpy(&handle, entry->src_data, sizeof(struct amdgpu_ualink_handle));
+ handle.handle_lo = amdgpu_ualink_reassemble_handle(handle.handle_lo,
+ src_acc_id);
+ dev_dbg(adev->dev,
+ "Got NPA-REVOKE MSG: remote acc_id %u handle %llx:%llx\n",
+ src_acc_id, handle.handle_hi, handle.handle_lo);
+ amdgpu_ualink_process_npa_revoke_msg(adev, src_acc_id, handle);
+ break;
+ case AMDGPU_UALINK_NPA_RELEASE_MSG:
+ memcpy(&handle, entry->src_data, sizeof(struct amdgpu_ualink_handle));
+ handle.handle_lo = amdgpu_ualink_reassemble_handle(handle.handle_lo,
+ local_acc_id);
+ dev_dbg(adev->dev,
+ "Got NPA-RELEASE MSG: remote acc_id %u handle %llx:%llx\n",
+ src_acc_id, handle.handle_hi, handle.handle_lo);
+ amdgpu_ualink_process_npa_release_msg(adev, src_acc_id, handle);
+ break;
+ default:
+ dev_err(adev->dev, "Unknown message type (%u)\n", msg_type);
+ break;
+ }
return handled;
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.h
index 0c2c5fad867af..bd8af7596a6ef 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ualink.h
@@ -35,6 +35,7 @@
#define AMDGPU_UALINK_HANDLE_ACCID_MASK GENMASK_ULL(9, 0)
#define AMDGPU_UALINK_MESSAGE_HEADER_MASK GENMASK_ULL(9, 0)
+#define AMDGPU_UALINK_HELLO_MSG_ACCID_MASK GENMASK_U32(9, 0)
#define AMDGPU_UALINK_HELLO_MSG_RECV_ACCID_SHIFT 10
#define AMDGPU_UALINK_HELLO_MSG_SENDER_ACCID_SHIFT 20
#define AMDGPU_UALINK_NPA_FAIL_MSG_FAIL_REASON_MASK GENMASK_U32(7, 0)
@@ -340,6 +341,8 @@ int amdgpu_ualink_init_interrupt(struct amdgpu_device *adev);
int amdgpu_ualink_sw_init(struct amdgpu_device *adev);
void amdgpu_ualink_sw_fini(struct amdgpu_device *adev);
+int ualink_send_hello(struct amdgpu_device *adev, u32 remote_accel_id);
+
int amdgpu_ualink_sysfs_init(struct amdgpu_device *adev);
void amdgpu_ualink_sysfs_fini(struct amdgpu_device *adev);
int amdgpu_ualink_init(struct amdgpu_device *adev);
--
2.55.0
next prev parent reply other threads:[~2026-08-21 19:53 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 ` Alex Deucher [this message]
2026-08-21 19:34 ` [PATCH 57/95] drm/amdgpu: Send TLB shootdown on exported memory unmap Alex Deucher
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 56/95] drm/amdgpu: Add handling for remote interrupts 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=20260821193458.808626-57-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;
as well as URLs for NNTP newsgroup(s).