From 016c290447c72a72eaac8337cbf09b32f87e9093 Mon Sep 17 00:00:00 2001
From: Arthur Hanson Rasmusson <arthur@vgpu.io>
Date: Sat, 24 Oct 2026 13:03:56 -0400
Subject: vgsp-rm r623: oserrlog - deliver a synthetic OS_ERROR_LOG (0x1006)
 engine-fault report to the guest's RM queue (debug param; payload mirrors the
 GSP ABI decode: exceptType/runlistId/chid/errString/prevXid) - the suspected
 missing recovery half after a synth-fault (post-fault clients stick
 pre-submission)

---
 kernel/main.c | 79 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 79 insertions(+)

(limited to 'kernel/main.c')

diff --git a/kernel/main.c b/kernel/main.c
index 9a5d912..dd11d12 100644
--- a/kernel/main.c
+++ b/kernel/main.c
@@ -3952,5 +3952,84 @@ static const struct kernel_param_ops nv_vgpu_faultpoke_ops = {
 module_param_cb(grfaultpoke, &nv_vgpu_faultpoke_ops, NULL, 0200);
 MODULE_PARM_DESC(grfaultpoke, "Write a channel error notifier + post its event (r616 experiment)");
 
+/*
+ * Debug r623: deliver a synthetic OS_ERROR_LOG (0x1006) engine-fault report
+ * to the guest's RM queue:
+ *   echo 1            > .../parameters/oserrlog
+ *   echo 0x1,0,0      > .../parameters/oserrlog   (exceptType,runlistId,chid)
+ * Experiment: after a synth-fault, the next client sticks pre-submission; if
+ * the guest's fault handling completes only when the 0x1006 report arrives,
+ * this record is the missing recovery half.
+ */
+struct nv_vgpu_oserrlog_req {
+	u32 except_type, runlist_id, chid;
+	int rc;
+	struct completion done;
+};
+
+static int nv_vgpu_oserrlog_thread(void *arg)
+{
+	struct nv_vgpu_oserrlog_req *req = arg;
+	struct nv_vgpu_device *vgpu;
+
+	mutex_lock(&nv_vgpu_driver.lock);
+	list_for_each_entry(vgpu, &nv_vgpu_driver.devices, list) {
+		if (atomic_read(&vgpu->vfio_dev_open_count) <= 0)
+			continue;
+		req->rc = gsp_rm_send_os_error_log(vgpu, req->except_type,
+						   req->runlist_id, req->chid,
+						   "GR");
+		pr_info("vgsp: oserrlog except=0x%x rl=%u chid=%u -> %d\n",
+			req->except_type, req->runlist_id, req->chid, req->rc);
+	}
+	mutex_unlock(&nv_vgpu_driver.lock);
+
+	complete(&req->done);
+	return 0;
+}
+
+static int nv_vgpu_oserrlog_set(const char *val, const struct kernel_param *kp)
+{
+	static struct nv_vgpu_oserrlog_req req;
+	struct task_struct *task;
+	char buf[64];
+	char *c1, *c2;
+
+	memset(&req, 0, sizeof(req));
+	req.rc = -ENODEV;
+	strscpy(buf, val, sizeof(buf));
+	req.except_type = 1;	/* generic engine exception, as the synth-fault */
+	{
+		c1 = strchr(buf, ',');
+		if (c1) {
+			*c1++ = '\0';
+			c2 = strchr(c1, ',');
+			if (c2) {
+				*c2++ = '\0';
+				if (kstrtou32(c2, 0, &req.chid))
+					return -EINVAL;
+			}
+			if (kstrtou32(c1, 0, &req.runlist_id))
+				return -EINVAL;
+		}
+		if (kstrtou32(buf, 0, &req.except_type))
+			return -EINVAL;
+	}
+
+	init_completion(&req.done);
+	task = kthread_run(nv_vgpu_oserrlog_thread, &req, "vgpu-oserrlog");
+	if (IS_ERR(task))
+		return PTR_ERR(task);
+	wait_for_completion(&req.done);
+	return 0;
+}
+
+static const struct kernel_param_ops nv_vgpu_oserrlog_ops = {
+	.set = nv_vgpu_oserrlog_set,
+};
+
+module_param_cb(oserrlog, &nv_vgpu_oserrlog_ops, NULL, 0200);
+MODULE_PARM_DESC(oserrlog, "Send a synthetic OS_ERROR_LOG (0x1006) RM record to the guest (r623 experiment)");
+
 module_init(nv_vgpu_init);
 module_exit(nv_vgpu_exit);
-- 
cgit v1.2.3