[Intel-xe] [RFC 17/25] drm/xe/eudebug: Add per process coredumps
Mika Kuoppala
mika.kuoppala at linux.intel.com
Mon Nov 6 11:18:37 UTC 2023
We gather state of execution errors for particular pid for
retrospection. State includes vmas, contents of vmas
that were marked for capture and vm and vma based metadata.
TODO: Quotas
Signed-off-by: Mika Kuoppala <mika.kuoppala at linux.intel.com>
---
drivers/gpu/drm/xe/xe_device.c | 3 +
drivers/gpu/drm/xe/xe_device_types.h | 3 +
drivers/gpu/drm/xe/xe_usercoredump.c | 182 ++++++++++++++++++++-
drivers/gpu/drm/xe/xe_usercoredump.h | 7 +
drivers/gpu/drm/xe/xe_usercoredump_types.h | 23 +++
5 files changed, 217 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/xe/xe_device.c b/drivers/gpu/drm/xe/xe_device.c
index 9644387cd11f..bc8032266ef0 100644
--- a/drivers/gpu/drm/xe/xe_device.c
+++ b/drivers/gpu/drm/xe/xe_device.c
@@ -41,6 +41,7 @@
#include "xe_wait_user_fence.h"
#include "xe_hwmon.h"
#include "xe_eudebug.h"
+#include "xe_usercoredump.h"
#ifdef CONFIG_LOCKDEP
struct lockdep_map xe_device_mem_access_lockdep_map = {
@@ -456,6 +457,7 @@ int xe_device_probe(struct xe_device *xe)
xe_display_register(xe);
xe_debugfs_register(xe);
+ xe_usercoredumps_init(xe);
xe_pmu_register(&xe->pmu);
@@ -479,6 +481,7 @@ int xe_device_probe(struct xe_device *xe)
static void xe_device_remove_display(struct xe_device *xe)
{
+ xe_usercoredumps_fini(xe);
xe_display_unregister(xe);
drm_dev_unplug(&xe->drm);
diff --git a/drivers/gpu/drm/xe/xe_device_types.h b/drivers/gpu/drm/xe/xe_device_types.h
index c025ecd1c321..57ec132286d7 100644
--- a/drivers/gpu/drm/xe/xe_device_types.h
+++ b/drivers/gpu/drm/xe/xe_device_types.h
@@ -203,6 +203,9 @@ struct xe_device {
/** @devcoredump: device coredump */
struct xe_devcoredump devcoredump;
+ /** @usercoredumps: User state coredump repository */
+ struct xe_usercoredumps usercoredumps;
+
/** @info: device info */
struct intel_device_info {
/** @graphics_name: graphics IP name */
diff --git a/drivers/gpu/drm/xe/xe_usercoredump.c b/drivers/gpu/drm/xe/xe_usercoredump.c
index 2ebaf15b563c..2d3fc5a24c48 100644
--- a/drivers/gpu/drm/xe/xe_usercoredump.c
+++ b/drivers/gpu/drm/xe/xe_usercoredump.c
@@ -89,7 +89,7 @@ vma_snapshot_create(struct xe_vma *vma, u64 addr, bool is_vram)
{
struct xe_vma_snapshot *s;
- s = kzalloc(sizeof(*s), GFP_ATOMIC);
+ s = kzalloc(sizeof(*s), GFP_NOWAIT | __GFP_NOWARN);
if (!s)
return NULL;
@@ -254,6 +254,7 @@ void xe_user_state_snapshot_print(struct xe_user_state_snapshot *s,
mutex_lock(&s->vmas.lock);
list_for_each_entry(v, &s->vmas.list, link)
vma_snapshot_print(p, v);
+
mutex_unlock(&s->vmas.lock);
}
@@ -290,3 +291,182 @@ xe_user_state_snapshot_capture(struct xe_exec_queue *q)
return s;
}
+static int create_pidroot_debugfs(struct xe_usercoredumps *store,
+ struct xe_pidroot *r)
+{
+ struct dentry *root = store->debugfs_root;
+ struct dentry *d;
+ char pid_str[8];
+
+ if (!root) {
+ drm_warn(&store->xe->drm, "no debugfs root for usercoredumps\n");
+ return -ENOENT;
+ }
+
+ sprintf(pid_str, "%u", r->pid_nr);
+ d = debugfs_create_dir(pid_str, root);
+ if (IS_ERR(d))
+ return PTR_ERR(d);
+
+ r->dentry = d;
+
+ return 0;
+}
+
+static struct xe_pidroot *create_pidroot(struct xe_usercoredumps *store,
+ struct xe_user_state_snapshot *s)
+{
+ struct xe_pidroot *r, *old;
+ int ret;
+
+ r = kzalloc(sizeof(*r), GFP_KERNEL);
+ if (!r)
+ return ERR_PTR(-ENOMEM);
+
+ r->pid_nr = pid_vnr(s->pid);
+ INIT_LIST_HEAD(&r->snapshots);
+ old = xa_store(&store->pidroots, r->pid_nr, r, GFP_KERNEL);
+ if (xa_is_err(old)) {
+ kfree(r);
+ return ERR_PTR(xa_err(old));
+ }
+
+ ret = create_pidroot_debugfs(store, r);
+ if (ret) {
+ kfree(r);
+ return ERR_PTR(ret);
+ }
+
+ r->store = store;
+
+ return r;
+}
+
+static int snapshot_show(struct seq_file *m, void *data)
+{
+ struct xe_user_state_snapshot *s = m->private;
+ struct drm_printer p = drm_seq_file_printer(m);
+
+ xe_user_state_snapshot_print(s, &p);
+
+ return 0;
+}
+
+static ssize_t snapshot_write(struct file *file,
+ const char __user *ubuf,
+ size_t len, loff_t *offp)
+{
+ /* struct seq_file *m = file->private_data; */
+ /* struct xe_user_state_snapshot *s = m->private; */
+
+ return len;
+}
+
+static int snapshot_open(struct inode *inode, struct file *file)
+{
+ return single_open(file, snapshot_show, inode->i_private);
+}
+
+static const struct file_operations snapshot_fops = {
+ .owner = THIS_MODULE,
+ .open = snapshot_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = single_release,
+ .write = snapshot_write
+};
+
+static int create_snapshot_debugfs(const struct xe_pidroot *r,
+ struct xe_user_state_snapshot *s)
+{
+ struct dentry *d;
+ char name[32];
+
+ sprintf(name, "%lu", r->seqno);
+
+ d = debugfs_create_file(name, S_IRUSR | S_IWUSR,
+ r->dentry,
+ s, &snapshot_fops);
+ if (IS_ERR(d))
+ return PTR_ERR(d);
+
+ s->dentry = d;
+
+ return 0;
+}
+
+int xe_usercoredumps_add(struct xe_device *xe,
+ struct xe_user_state_snapshot *s)
+{
+ struct xe_usercoredumps *store = &xe->usercoredumps;
+ struct xe_pidroot *pr;
+ int ret;
+
+ ret = 0;
+ mutex_lock(&store->lock);
+ pr = xa_load(&store->pidroots, pid_vnr(s->pid));
+ if (!pr) {
+ pr = create_pidroot(store, s);
+ if (IS_ERR(pr)) {
+ ret = PTR_ERR(pr);
+ pr = NULL;
+ }
+ }
+
+ if (pr) {
+ s->pidroot = pr;
+ pr->seqno++;
+ list_add_tail(&s->link, &pr->snapshots);
+ create_snapshot_debugfs(pr, s);
+ }
+
+ mutex_unlock(&store->lock);
+
+ return ret;
+}
+
+void xe_usercoredumps_init(struct xe_device *xe)
+{
+ struct drm_minor *minor = xe->drm.primary;
+ struct dentry *root = minor->debugfs_root;
+ struct xe_usercoredumps *ucd = &xe->usercoredumps;
+
+ if (XE_WARN_ON(ucd->debugfs_root))
+ return;
+
+ mutex_init(&ucd->lock);
+ xa_init_flags(&ucd->pidroots, GFP_KERNEL);
+
+ ucd->debugfs_root = debugfs_create_dir("usercoredumps", root);
+ if (IS_ERR(ucd->debugfs_root)) {
+ drm_warn(&xe->drm, "Create usercoredumps directory failed %ld",
+ PTR_ERR(ucd->debugfs_root));
+ ucd->debugfs_root = NULL;
+ }
+
+ ucd->xe = xe;
+}
+
+void xe_usercoredumps_fini(struct xe_device *xe)
+{
+ struct xe_usercoredumps *ucd = &xe->usercoredumps;
+ struct xe_pidroot *pr;
+ long unsigned int pid;
+
+ if (XE_WARN_ON(!ucd->debugfs_root))
+ return;
+
+ xa_for_each(&ucd->pidroots, pid, pr) {
+ struct xe_user_state_snapshot *s, *tmp;
+
+ list_for_each_entry_safe(s, tmp, &pr->snapshots, link)
+ xe_user_state_snapshot_release(s);
+
+ kfree(pr);
+ }
+
+ debugfs_remove_recursive(ucd->debugfs_root);
+
+ xa_destroy(&ucd->pidroots);
+ mutex_destroy(&ucd->lock);
+}
diff --git a/drivers/gpu/drm/xe/xe_usercoredump.h b/drivers/gpu/drm/xe/xe_usercoredump.h
index cc94e85996e9..f7bc5e0f7607 100644
--- a/drivers/gpu/drm/xe/xe_usercoredump.h
+++ b/drivers/gpu/drm/xe/xe_usercoredump.h
@@ -16,4 +16,11 @@ xe_user_state_snapshot_capture(struct xe_exec_queue *eq);
void xe_user_state_snapshot_release(struct xe_user_state_snapshot *s);
void xe_user_state_snapshot_print(struct xe_user_state_snapshot *s,
struct drm_printer *p);
+
+void xe_usercoredumps_init(struct xe_device *xe);
+void xe_usercoredumps_fini(struct xe_device *xe);
+int xe_usercoredumps_add(struct xe_device *xe,
+ struct xe_user_state_snapshot *s);
+int xe_usercoredumps_remove(struct xe_device *xe,
+ struct xe_user_state_snapshot *s);
#endif
diff --git a/drivers/gpu/drm/xe/xe_usercoredump_types.h b/drivers/gpu/drm/xe/xe_usercoredump_types.h
index 7d80f412edaf..635656e830cf 100644
--- a/drivers/gpu/drm/xe/xe_usercoredump_types.h
+++ b/drivers/gpu/drm/xe/xe_usercoredump_types.h
@@ -9,8 +9,10 @@
#include <linux/ktime.h>
#include <linux/mutex.h>
#include <linux/sched.h>
+#include <linux/xarray.h>
struct xe_device;
+struct xe_pidroot;
struct xe_vma_snapshot {
struct list_head link;
@@ -34,6 +36,7 @@ struct xe_vma_snapshot {
* shows the state of the GPU of when the issue has happened.
*/
struct xe_user_state_snapshot {
+ struct list_head link;
struct {
struct mutex lock;
@@ -43,6 +46,26 @@ struct xe_user_state_snapshot {
struct pid *pid;
u64 client_id; /* drm client id */
char comm[TASK_COMM_LEN];
+ u64 count;
+
+ struct xe_pidroot *pidroot;
+ struct dentry *dentry;
+};
+
+struct xe_pidroot {
+ struct xe_usercoredumps *store;
+ int pid_nr;
+ unsigned long seqno;
+ struct list_head snapshots;
+ struct dentry *dentry;
+};
+
+struct xe_usercoredumps {
+ struct xe_device *xe;
+ struct mutex lock;
+ struct xarray pidroots;
+
+ struct dentry *debugfs_root;
};
#endif
--
2.34.1
More information about the Intel-xe
mailing list