drm/xe/guc: Add page reclamation interface to GuC

Add page reclamation related changes to GuC interface, handlers, and
senders to support page reclamation.

Currently TLB invalidations will perform an entire PPC flush in order to
prevent stale memory access for noncoherent system memory. Page
reclamation is an extension of the typical TLB invalidation
workflow, allowing disabling of full PPC flush and enable selective PPC
flushing. Selective flushing will be decided by a list of pages whom's
address is passed to GuC at time of action.

Page reclamation interfaces require at least GuC FW ver 70.31.0.

v2:
 - Moved send_page_reclaim to first patch usage.
 - Add comments explaining shared done handler. (Matthew B)
 - Add FW version fallback to disable page reclaim
   on older versions. (Matthew B, Shuicheng)

Signed-off-by: Brian Nguyen <brian3.nguyen@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Shuicheng Lin <shuicheng.lin@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patch.msgid.link/20251212213225.3564537-16-brian3.nguyen@intel.com

diff --git a/drivers/gpu/drm/xe/abi/guc_actions_abi.h b/drivers/gpu/drm/xe/abi/guc_actions_abi.h
index 8af3691626bfd8b00907db2e298677a2176dc8fd..83a6e7794982e0c9d0a9ae23a1895ddf99a6cc61 100644 (file)
--- a/drivers/gpu/drm/xe/abi/guc_actions_abi.h
+++ b/drivers/gpu/drm/xe/abi/guc_actions_abi.h
@@ -155,6 +155,8 @@ enum xe_guc_action {
        XE_GUC_ACTION_TLB_INVALIDATION = 0x7000,
        XE_GUC_ACTION_TLB_INVALIDATION_DONE = 0x7001,
        XE_GUC_ACTION_TLB_INVALIDATION_ALL = 0x7002,
+       XE_GUC_ACTION_PAGE_RECLAMATION = 0x7003,
+       XE_GUC_ACTION_PAGE_RECLAMATION_DONE = 0x7004,
        XE_GUC_ACTION_STATE_CAPTURE_NOTIFICATION = 0x8002,
        XE_GUC_ACTION_NOTIFY_FLUSH_LOG_BUFFER_TO_FILE = 0x8003,
        XE_GUC_ACTION_NOTIFY_CRASH_DUMP_POSTED = 0x8004,

diff --git a/drivers/gpu/drm/xe/xe_guc.c b/drivers/gpu/drm/xe/xe_guc.c
index f0407bab9a0c3126f44f1d9f9be6a88fa6bf247c..7daae32946659ecc8ae2b1393db07b972c910ca4 100644 (file)
--- a/drivers/gpu/drm/xe/xe_guc.c
+++ b/drivers/gpu/drm/xe/xe_guc.c
@@ -767,6 +767,10 @@ int xe_guc_init(struct xe_guc *guc)
        if (!xe_uc_fw_is_enabled(&guc->fw))
                return 0;
 
+       /* Disable page reclaim if GuC FW does not support */
+       if (GUC_FIRMWARE_VER(guc) < MAKE_GUC_VER(70, 31, 0))
+               xe->info.has_page_reclaim_hw_assist = false;
+
        if (IS_SRIOV_VF(xe)) {
                ret = xe_guc_ct_init(&guc->ct);
                if (ret)

diff --git a/drivers/gpu/drm/xe/xe_guc_ct.c b/drivers/gpu/drm/xe/xe_guc_ct.c
index 3e49e7fd00311f8256dfb6c24ed854f364a57c49..c3df9b3f1b4de5f462dd61c0d4c9c9487204b13f 100644 (file)
--- a/drivers/gpu/drm/xe/xe_guc_ct.c
+++ b/drivers/gpu/drm/xe/xe_guc_ct.c
@@ -1406,6 +1406,7 @@ static int parse_g2h_event(struct xe_guc_ct *ct, u32 *msg, u32 len)
        case XE_GUC_ACTION_DEREGISTER_CONTEXT_DONE:
        case XE_GUC_ACTION_SCHED_ENGINE_MODE_DONE:
        case XE_GUC_ACTION_TLB_INVALIDATION_DONE:
+       case XE_GUC_ACTION_PAGE_RECLAMATION_DONE:
                g2h_release_space(ct, len);
        }
 
@@ -1592,6 +1593,15 @@ static int process_g2h_msg(struct xe_guc_ct *ct, u32 *msg, u32 len)
                ret = xe_guc_pagefault_handler(guc, payload, adj_len);
                break;
        case XE_GUC_ACTION_TLB_INVALIDATION_DONE:
+       case XE_GUC_ACTION_PAGE_RECLAMATION_DONE:
+               /*
+                * Page reclamation is an extension of TLB invalidation. Both
+                * operations share the same seqno and fence. When either
+                * action completes, we need to signal the corresponding
+                * fence. Since the handling logic (lookup fence by seqno,
+                * fence signalling) is identical, we use the same handler
+                * for both G2H events.
+                */
                ret = xe_guc_tlb_inval_done_handler(guc, payload, adj_len);
                break;
        case XE_GUC_ACTION_GUC2PF_RELAY_FROM_VF:
@@ -1764,6 +1774,7 @@ static int g2h_read(struct xe_guc_ct *ct, u32 *msg, bool fast_path)
                switch (action) {
                case XE_GUC_ACTION_REPORT_PAGE_FAULT_REQ_DESC:
                case XE_GUC_ACTION_TLB_INVALIDATION_DONE:
+               case XE_GUC_ACTION_PAGE_RECLAMATION_DONE:
                        break;  /* Process these in fast-path */
                default:
                        return 0;
@@ -1800,6 +1811,12 @@ static void g2h_fast_path(struct xe_guc_ct *ct, u32 *msg, u32 len)
                ret = xe_guc_pagefault_handler(guc, payload, adj_len);
                break;
        case XE_GUC_ACTION_TLB_INVALIDATION_DONE:
+       case XE_GUC_ACTION_PAGE_RECLAMATION_DONE:
+               /*
+                * Seqno and fence handling of page reclamation and TLB
+                * invalidation is identical, so we can use the same handler
+                * for both actions.
+                */
                __g2h_release_space(ct, len);
                ret = xe_guc_tlb_inval_done_handler(guc, payload, adj_len);
                break;

diff --git a/drivers/gpu/drm/xe/xe_guc_fwif.h b/drivers/gpu/drm/xe/xe_guc_fwif.h
index e27f0088f24fe4591e4893f16f0ac65fc18414a8..a04faec477ae9996ffb3ba0683ba2bd8528e4837 100644 (file)
--- a/drivers/gpu/drm/xe/xe_guc_fwif.h
+++ b/drivers/gpu/drm/xe/xe_guc_fwif.h
@@ -17,6 +17,7 @@
 #define G2H_LEN_DW_TLB_INVALIDATE              3
 #define G2H_LEN_DW_G2G_NOTIFY_MIN              3
 #define G2H_LEN_DW_MULTI_QUEUE_CONTEXT         3
+#define G2H_LEN_DW_PAGE_RECLAMATION            3
 
 #define GUC_ID_MAX                     65535
 #define GUC_ID_UNKNOWN                 0xffffffff