[Intel-gfx] [PATCH v2] drm/i915: Sysfs interface to get GFX shmem usage stats per process
Daniel Vetter
daniel at ffwll.ch
Wed Sep 3 12:58:03 CEST 2014
On Wed, Sep 03, 2014 at 03:39:55PM +0530, sourab.gupta at intel.com wrote:
> From: Sourab Gupta <sourab.gupta at intel.com>
>
> Currently the Graphics Driver provides an interface through which
> one can get a snapshot of the overall Graphics memory consumption.
> Also there is an interface available, which provides information
> about the several memory related attributes of every single Graphics
> buffer created by the various clients.
>
> There is a requirement of a new interface for achieving below
> functionalities:
> 1) Need to provide Client based detailed information about the
> distribution of Graphics memory
> 2) Need to provide an interface which can provide info about the
> sharing of Graphics buffers between the clients.
>
> The client based interface would also aid in debugging of
> memory usage/consumption by each client & debug memleak related issues.
>
> With this new interface,
> 1) In case of memleak scenarios, we can easily zero in on the culprit
> client which is unexpectedly holding on the Graphics buffers for an
> inordinate amount of time.
> 2) We can get an estimate of the instantaneous memory footprint of
> every Graphics client.
> 3) We can now trace all the processes sharing a particular Graphics buffer.
>
> By means of this patch we try to provide a sysfs interface to achieve
> the mentioned functionalities.
>
> There are two files created in sysfs:
> 'i915_gem_meminfo' will provide summary of the graphics resources used by
> each graphics client.
> 'i915_gem_objinfo' will provide detailed view of each object created by
> individual clients.
>
> v2: Changes made for
> - adding support to report user virtual addresses of mapped buffers
> - replacing pid based reporting with tgid based one
> - checkpatch and other misc cleanup
>
> Signed-off-by: Sourab Gupta <sourab.gupta at intel.com>
> Signed-off-by: Akash Goel <akash.goel at intel.com>
Sorry I didn't spot this the first time around, but I think sysfs is the
wrong place for this.
Generally sysfs is for setting/reading per-object values, and it has the
big rule that there should be only _one_ value per file. The error state
is a bit an exception, but otoh it's also just the full dump as a binary
file (which for historical reasons is printed as ascii).
The other issue is that imo this should be a generic interface, so that we
can write a gpu_top tool for dumping memory consumers which works on all
linux platforms.
To avoid delaying for a long time can we just move ahead by putting this
into debugfs?
Also in debugfs there's already a lot of this stuff around - why is that
not sufficient and could we extend it somehow with the missing bits?
Thanks, Daniel
> ---
> drivers/gpu/drm/i915/i915_dma.c | 1 +
> drivers/gpu/drm/i915/i915_drv.c | 2 +
> drivers/gpu/drm/i915/i915_drv.h | 26 ++
> drivers/gpu/drm/i915/i915_gem.c | 169 ++++++++++-
> drivers/gpu/drm/i915/i915_gem_debug.c | 542 ++++++++++++++++++++++++++++++++++
> drivers/gpu/drm/i915/i915_gpu_error.c | 2 +-
> drivers/gpu/drm/i915/i915_sysfs.c | 83 ++++++
> 7 files changed, 822 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
> index a58fed9..7ea3250 100644
> --- a/drivers/gpu/drm/i915/i915_dma.c
> +++ b/drivers/gpu/drm/i915/i915_dma.c
> @@ -1985,6 +1985,7 @@ void i915_driver_postclose(struct drm_device *dev, struct drm_file *file)
> {
> struct drm_i915_file_private *file_priv = file->driver_priv;
>
> + kfree(file_priv->process_name);
> if (file_priv && file_priv->bsd_ring)
> file_priv->bsd_ring = NULL;
> kfree(file_priv);
> diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
> index 1d6d9ac..9bee20e 100644
> --- a/drivers/gpu/drm/i915/i915_drv.c
> +++ b/drivers/gpu/drm/i915/i915_drv.c
> @@ -1628,6 +1628,8 @@ static struct drm_driver driver = {
> .debugfs_init = i915_debugfs_init,
> .debugfs_cleanup = i915_debugfs_cleanup,
> #endif
> + .gem_open_object = i915_gem_open_object,
> + .gem_close_object = i915_gem_close_object,
> .gem_free_object = i915_gem_free_object,
> .gem_vm_ops = &i915_gem_vm_ops,
>
> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> index 36f3da6..43ba7c4 100644
> --- a/drivers/gpu/drm/i915/i915_drv.h
> +++ b/drivers/gpu/drm/i915/i915_drv.h
> @@ -1765,6 +1765,11 @@ struct drm_i915_gem_object_ops {
> #define INTEL_FRONTBUFFER_ALL_MASK(pipe) \
> (0xf << (INTEL_FRONTBUFFER_BITS_PER_PIPE * (pipe)))
>
> +struct drm_i915_obj_virt_addr {
> + struct list_head head;
> + unsigned long user_virt_addr;
> +};
> +
> struct drm_i915_gem_object {
> struct drm_gem_object base;
>
> @@ -1890,6 +1895,13 @@ struct drm_i915_gem_object {
> struct work_struct *work;
> } userptr;
> };
> +
> +#define MAX_OPEN_HANDLE 20
> + struct {
> + struct list_head virt_addr_head;
> + pid_t pid;
> + int open_handle_count;
> + } pid_array[MAX_OPEN_HANDLE];
> };
> #define to_intel_bo(x) container_of(x, struct drm_i915_gem_object, base)
>
> @@ -1940,6 +1952,8 @@ struct drm_i915_gem_request {
> struct drm_i915_file_private {
> struct drm_i915_private *dev_priv;
> struct drm_file *file;
> + char *process_name;
> + struct pid *tgid;
>
> struct {
> spinlock_t lock;
> @@ -2370,6 +2384,10 @@ void i915_init_vm(struct drm_i915_private *dev_priv,
> struct i915_address_space *vm);
> void i915_gem_free_object(struct drm_gem_object *obj);
> void i915_gem_vma_destroy(struct i915_vma *vma);
> +int i915_gem_open_object(struct drm_gem_object *gem_obj,
> + struct drm_file *file_priv);
> +int i915_gem_close_object(struct drm_gem_object *gem_obj,
> + struct drm_file *file_priv);
>
> #define PIN_MAPPABLE 0x1
> #define PIN_NONBLOCK 0x2
> @@ -2420,6 +2438,8 @@ int i915_gem_dumb_create(struct drm_file *file_priv,
> struct drm_mode_create_dumb *args);
> int i915_gem_mmap_gtt(struct drm_file *file_priv, struct drm_device *dev,
> uint32_t handle, uint64_t *offset);
> +int i915_gem_obj_shmem_pages_alloced(struct drm_i915_gem_object *obj);
> +
> /**
> * Returns true if seq1 is later than seq2.
> */
> @@ -2686,6 +2706,10 @@ int i915_verify_lists(struct drm_device *dev);
> #else
> #define i915_verify_lists(dev) 0
> #endif
> +int i915_get_drm_clients_info(struct drm_i915_error_state_buf *m,
> + struct drm_device *dev);
> +int i915_gem_get_all_obj_info(struct drm_i915_error_state_buf *m,
> + struct drm_device *dev);
>
> /* i915_debugfs.c */
> int i915_debugfs_init(struct drm_minor *minor);
> @@ -2699,6 +2723,8 @@ static inline void intel_display_crc_init(struct drm_device *dev) {}
> /* i915_gpu_error.c */
> __printf(2, 3)
> void i915_error_printf(struct drm_i915_error_state_buf *e, const char *f, ...);
> +void i915_error_puts(struct drm_i915_error_state_buf *e,
> + const char *str);
> int i915_error_state_to_str(struct drm_i915_error_state_buf *estr,
> const struct i915_error_state_file_priv *error);
> int i915_error_state_buf_init(struct drm_i915_error_state_buf *eb,
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index 6c68570..3c36486 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -1461,6 +1461,45 @@ unlock:
> return ret;
> }
>
> +static void
> +i915_gem_obj_insert_virt_addr(struct drm_i915_gem_object *obj,
> + unsigned long addr,
> + bool is_map_gtt)
> +{
> + pid_t current_pid = task_tgid_nr(current);
> + int i, found = 0;
> +
> + if (is_map_gtt)
> + addr |= 1;
> +
> + for (i = 0; i < MAX_OPEN_HANDLE; i++) {
> + if (obj->pid_array[i].pid == current_pid) {
> + struct drm_i915_obj_virt_addr *entry, *new_entry;
> +
> + list_for_each_entry(entry,
> + &obj->pid_array[i].virt_addr_head,
> + head) {
> + if (entry->user_virt_addr == addr) {
> + found = 1;
> + break;
> + }
> + }
> + if (found)
> + break;
> + new_entry = kzalloc
> + (sizeof(struct drm_i915_obj_virt_addr),
> + GFP_KERNEL);
> + new_entry->user_virt_addr = addr;
> + list_add_tail(&new_entry->head,
> + &obj->pid_array[i].virt_addr_head);
> + break;
> + }
> + }
> + if (i == MAX_OPEN_HANDLE)
> + DRM_DEBUG("Couldn't find matching pid %d for obj 0x%x\n",
> + current_pid, (u32) obj);
> +}
> +
> /**
> * Maps the contents of an object, returning the address it is mapped
> * into.
> @@ -1495,6 +1534,7 @@ i915_gem_mmap_ioctl(struct drm_device *dev, void *data,
> if (IS_ERR((void *)addr))
> return addr;
>
> + i915_gem_obj_insert_virt_addr(to_intel_bo(obj), addr, false);
> args->addr_ptr = (uint64_t) addr;
>
> return 0;
> @@ -1585,6 +1625,8 @@ int i915_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
> }
>
> obj->fault_mappable = true;
> + i915_gem_obj_insert_virt_addr(obj,
> + (unsigned long)vma->vm_start, true);
> } else
> ret = vm_insert_pfn(vma,
> (unsigned long)vmf->virtual_address,
> @@ -1830,6 +1872,24 @@ i915_gem_object_is_purgeable(struct drm_i915_gem_object *obj)
> return obj->madv == I915_MADV_DONTNEED;
> }
>
> +int i915_gem_obj_shmem_pages_alloced(struct drm_i915_gem_object *obj)
> +{
> + int ret;
> +
> + if (obj->base.filp) {
> + struct inode *inode = file_inode(obj->base.filp);
> + struct shmem_inode_info *info = SHMEM_I(inode);
> +
> + if (!inode)
> + return 0;
> + spin_lock(&info->lock);
> + ret = inode->i_mapping->nrpages;
> + spin_unlock(&info->lock);
> + return ret;
> + }
> + return 0;
> +}
> +
> /* Immediately discard the backing storage */
> static void
> i915_gem_object_truncate(struct drm_i915_gem_object *obj)
> @@ -4447,6 +4507,79 @@ static bool discard_backing_storage(struct drm_i915_gem_object *obj)
> return atomic_long_read(&obj->base.filp->f_count) == 1;
> }
>
> +int
> +i915_gem_open_object(struct drm_gem_object *gem_obj,
> + struct drm_file *file_priv)
> +{
> + struct drm_i915_gem_object *obj = to_intel_bo(gem_obj);
> + pid_t current_pid = task_tgid_nr(current);
> + int i, ret, free = -1;
> +
> + ret = i915_mutex_lock_interruptible(gem_obj->dev);
> + if (ret)
> + return ret;
> +
> + for (i = 0; i < MAX_OPEN_HANDLE; i++) {
> + if (obj->pid_array[i].pid == current_pid) {
> + obj->pid_array[i].open_handle_count++;
> + break;
> + } else if (obj->pid_array[i].pid == 0)
> + free = i;
> + }
> +
> + if (i == MAX_OPEN_HANDLE) {
> + if (free != -1) {
> + WARN_ON(obj->pid_array[free].open_handle_count);
> + obj->pid_array[free].open_handle_count = 1;
> + obj->pid_array[free].pid = current_pid;
> + INIT_LIST_HEAD(&obj->pid_array[free].virt_addr_head);
> + } else
> + DRM_DEBUG("Max open handle count limit: obj 0x%x\n",
> + (u32) obj);
> + }
> +
> + mutex_unlock(&gem_obj->dev->struct_mutex);
> + return 0;
> +}
> +
> +int
> +i915_gem_close_object(struct drm_gem_object *gem_obj,
> + struct drm_file *file_priv)
> +{
> + struct drm_i915_gem_object *obj = to_intel_bo(gem_obj);
> + pid_t current_pid = task_tgid_nr(current);
> + int i, ret;
> +
> + ret = i915_mutex_lock_interruptible(gem_obj->dev);
> + if (ret)
> + return ret;
> +
> + for (i = 0; i < MAX_OPEN_HANDLE; i++) {
> + if (obj->pid_array[i].pid == current_pid) {
> + obj->pid_array[i].open_handle_count--;
> + if (obj->pid_array[i].open_handle_count == 0) {
> + struct drm_i915_obj_virt_addr *entry, *next;
> +
> + list_for_each_entry_safe(entry, next,
> + &obj->pid_array[i].virt_addr_head,
> + head) {
> + list_del(&entry->head);
> + kfree(entry);
> + }
> + obj->pid_array[i].pid = 0;
> + }
> + break;
> + }
> + }
> + if (i == MAX_OPEN_HANDLE)
> + DRM_DEBUG("Couldn't find matching pid %d for obj 0x%x\n",
> + current_pid, (u32) obj);
> +
> + mutex_unlock(&gem_obj->dev->struct_mutex);
> + return 0;
> +}
> +
> +
> void i915_gem_free_object(struct drm_gem_object *gem_obj)
> {
> struct drm_i915_gem_object *obj = to_intel_bo(gem_obj);
> @@ -5072,13 +5205,37 @@ i915_gem_file_idle_work_handler(struct work_struct *work)
> atomic_set(&file_priv->rps_wait_boost, false);
> }
>
> +static int i915_gem_get_pid_cmdline(struct task_struct *task, char *buffer)
> +{
> + int res = 0;
> + unsigned int len;
> + struct mm_struct *mm = get_task_mm(task);
> +
> + if (!mm)
> + goto out;
> + if (!mm->arg_end)
> + goto out_mm;
> +
> + len = mm->arg_end - mm->arg_start;
> +
> + if (len > PAGE_SIZE)
> + len = PAGE_SIZE;
> +
> + res = access_process_vm(task, mm->arg_start, buffer, len, 0);
> +
> + if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE)
> + buffer[res-1] = '\0';
> +out_mm:
> + mmput(mm);
> +out:
> + return res;
> +}
> +
> int i915_gem_open(struct drm_device *dev, struct drm_file *file)
> {
> struct drm_i915_file_private *file_priv;
> int ret;
>
> - DRM_DEBUG_DRIVER("\n");
> -
> file_priv = kzalloc(sizeof(*file_priv), GFP_KERNEL);
> if (!file_priv)
> return -ENOMEM;
> @@ -5086,6 +5243,14 @@ int i915_gem_open(struct drm_device *dev, struct drm_file *file)
> file->driver_priv = file_priv;
> file_priv->dev_priv = dev->dev_private;
> file_priv->file = file;
> + file_priv->tgid = find_vpid(task_tgid_nr(current));
> + file_priv->process_name = kzalloc(PAGE_SIZE, GFP_ATOMIC);
> + if (!file_priv->process_name) {
> + kfree(file_priv);
> + return -ENOMEM;
> + }
> +
> + ret = i915_gem_get_pid_cmdline(current, file_priv->process_name);
>
> spin_lock_init(&file_priv->mm.lock);
> INIT_LIST_HEAD(&file_priv->mm.request_list);
> diff --git a/drivers/gpu/drm/i915/i915_gem_debug.c b/drivers/gpu/drm/i915/i915_gem_debug.c
> index f462d1b..7a42891 100644
> --- a/drivers/gpu/drm/i915/i915_gem_debug.c
> +++ b/drivers/gpu/drm/i915/i915_gem_debug.c
> @@ -25,6 +25,7 @@
> *
> */
>
> +#include <linux/pid.h>
> #include <drm/drmP.h>
> #include <drm/i915_drm.h>
> #include "i915_drv.h"
> @@ -116,3 +117,544 @@ i915_verify_lists(struct drm_device *dev)
> return warned = err;
> }
> #endif /* WATCH_LIST */
> +
> +struct per_file_obj_mem_info {
> + int num_obj;
> + int num_obj_shared;
> + int num_obj_private;
> + int num_obj_gtt_bound;
> + int num_obj_purged;
> + int num_obj_purgeable;
> + int num_obj_allocated;
> + int num_obj_fault_mappable;
> + int num_obj_stolen;
> + size_t gtt_space_allocated_shared;
> + size_t gtt_space_allocated_priv;
> + size_t phys_space_allocated_shared;
> + size_t phys_space_allocated_priv;
> + size_t phys_space_purgeable;
> + size_t phys_space_shared_proportion;
> + size_t fault_mappable_size;
> + size_t stolen_space_allocated;
> + char *process_name;
> +};
> +
> +struct name_entry {
> + struct list_head head;
> + struct drm_hash_item hash_item;
> +};
> +
> +struct pid_stat_entry {
> + struct list_head head;
> + struct list_head namefree;
> + struct drm_open_hash namelist;
> + struct per_file_obj_mem_info stats;
> + struct pid *pid;
> + int pid_num;
> +};
> +
> +
> +#define err_printf(e, ...) i915_error_printf(e, __VA_ARGS__)
> +#define err_puts(e, s) i915_error_puts(e, s)
> +
> +static const char *get_pin_flag(struct drm_i915_gem_object *obj)
> +{
> + if (obj->user_pin_count > 0)
> + return "P";
> + else if (i915_gem_obj_is_pinned(obj))
> + return "p";
> + return " ";
> +}
> +
> +static const char *get_tiling_flag(struct drm_i915_gem_object *obj)
> +{
> + switch (obj->tiling_mode) {
> + default:
> + case I915_TILING_NONE: return " ";
> + case I915_TILING_X: return "X";
> + case I915_TILING_Y: return "Y";
> + }
> +}
> +
> +static int i915_obj_virt_addr_is_valid(struct drm_gem_object *obj,
> + struct pid *pid, unsigned long addr)
> +{
> + struct task_struct *task;
> + struct mm_struct *mm;
> + struct vm_area_struct *vma;
> + int locked, ret = 0;
> +
> + task = get_pid_task(pid, PIDTYPE_PID);
> + if (task == NULL) {
> + DRM_DEBUG("null task for pid=%d\n", pid_nr(pid));
> + return -EINVAL;
> + }
> +
> + mm = get_task_mm(task);
> + if (mm == NULL) {
> + DRM_DEBUG("null mm for pid=%d\n", pid_nr(pid));
> + return -EINVAL;
> + }
> +
> + locked = down_read_trylock(&mm->mmap_sem);
> +
> + vma = find_vma(mm, addr);
> + if (vma) {
> + if (addr & 1) { /* mmap_gtt case */
> + if (vma->vm_pgoff*PAGE_SIZE == (unsigned long)
> + drm_vma_node_offset_addr(&obj->vma_node))
> + ret = 0;
> + else
> + ret = -EINVAL;
> + } else { /* mmap case */
> + if (vma->vm_file == obj->filp)
> + ret = 0;
> + else
> + ret = -EINVAL;
> + }
> + } else
> + ret = -EINVAL;
> +
> + if (locked)
> + up_read(&mm->mmap_sem);
> +
> + mmput(mm);
> + return ret;
> +}
> +
> +static void i915_obj_pidarray_validate(struct drm_gem_object *gem_obj)
> +{
> + struct drm_i915_gem_object *obj = to_intel_bo(gem_obj);
> + struct drm_device *dev = gem_obj->dev;
> + struct drm_i915_obj_virt_addr *entry, *next;
> + struct drm_file *file;
> + struct drm_i915_file_private *file_priv;
> + struct pid *tgid;
> + int pid_num, i, present;
> +
> + /* Run a sanity check on pid_array. All entries in pid_array should
> + * be subset of the the drm filelist pid entries.
> + */
> + for (i = 0; i < MAX_OPEN_HANDLE; i++) {
> + if (obj->pid_array[i].pid == 0)
> + continue;
> +
> + present = 0;
> + list_for_each_entry(file, &dev->filelist, lhead) {
> + file_priv = file->driver_priv;
> + tgid = file_priv->tgid;
> + pid_num = pid_nr(tgid);
> +
> + if (pid_num == obj->pid_array[i].pid) {
> + present = 1;
> + break;
> + }
> + }
> + if (present == 0) {
> + DRM_DEBUG("stale_pid=%d\n", obj->pid_array[i].pid);
> + list_for_each_entry_safe(entry, next,
> + &obj->pid_array[i].virt_addr_head,
> + head) {
> + list_del(&entry->head);
> + kfree(entry);
> + }
> +
> + obj->pid_array[i].open_handle_count = 0;
> + obj->pid_array[i].pid = 0;
> + } else {
> + /* Validate the virtual address list */
> + struct task_struct *task =
> + get_pid_task(tgid, PIDTYPE_PID);
> + if (task == NULL)
> + continue;
> +
> + list_for_each_entry_safe(entry, next,
> + &obj->pid_array[i].virt_addr_head,
> + head) {
> + if (i915_obj_virt_addr_is_valid(gem_obj, tgid,
> + entry->user_virt_addr)) {
> + DRM_DEBUG("stale_addr=%ld\n",
> + entry->user_virt_addr);
> + list_del(&entry->head);
> + kfree(entry);
> + }
> + }
> + }
> + }
> +}
> +
> +static int
> +i915_describe_obj(struct drm_i915_error_state_buf *m,
> + struct drm_i915_gem_object *obj)
> +{
> + int i;
> + struct i915_vma *vma;
> + struct drm_i915_obj_virt_addr *entry;
> +
> + err_printf(m,
> + "%p: %7zdK %s %s %s %s %s %s %s ",
> + &obj->base,
> + obj->base.size / 1024,
> + get_pin_flag(obj),
> + get_tiling_flag(obj),
> + obj->dirty ? "Y" : "N",
> + obj->base.name ? "Y" : "N",
> + (obj->userptr.mm != 0) ? "Y" : "N",
> + obj->stolen ? "Y" : "N",
> + (obj->pin_mappable || obj->fault_mappable) ? "Y" : "N");
> +
> + if (obj->madv == __I915_MADV_PURGED)
> + err_printf(m, " purged ");
> + else if (obj->madv == I915_MADV_DONTNEED)
> + err_printf(m, " purgeable ");
> + else if (i915_gem_obj_shmem_pages_alloced(obj) != 0)
> + err_printf(m, " allocated ");
> +
> +
> + list_for_each_entry(vma, &obj->vma_list, vma_link) {
> + if (!i915_is_ggtt(vma->vm))
> + err_puts(m, " PP ");
> + else
> + err_puts(m, " G ");
> + err_printf(m, " %08lx ", vma->node.start);
> + }
> +
> + for (i = 0; i < MAX_OPEN_HANDLE; i++) {
> + if (obj->pid_array[i].pid != 0) {
> + err_printf(m, " (%d: %d:",
> + obj->pid_array[i].pid,
> + obj->pid_array[i].open_handle_count);
> + list_for_each_entry(entry,
> + &obj->pid_array[i].virt_addr_head, head) {
> + if (entry->user_virt_addr & 1)
> + err_printf(m, " %p",
> + (void *)(entry->user_virt_addr & ~1));
> + else
> + err_printf(m, " %p*",
> + (void *)entry->user_virt_addr);
> + }
> + err_printf(m, ") ");
> + }
> + }
> +
> + err_printf(m, "\n");
> +
> + if (m->bytes == 0 && m->err)
> + return m->err;
> +
> + return 0;
> +}
> +
> +static int
> +i915_drm_gem_obj_info(int id, void *ptr, void *data)
> +{
> + struct drm_i915_gem_object *obj = ptr;
> + struct drm_i915_error_state_buf *m = data;
> + int ret;
> +
> + i915_obj_pidarray_validate(&obj->base);
> + ret = i915_describe_obj(m, obj);
> +
> + return ret;
> +}
> +
> +static int
> +i915_drm_gem_object_per_file_summary(int id, void *ptr, void *data)
> +{
> + struct pid_stat_entry *pid_entry = data;
> + struct drm_i915_gem_object *obj = ptr;
> + struct per_file_obj_mem_info *stats = &pid_entry->stats;
> + struct drm_hash_item *hash_item;
> + int i, obj_shared_count = 0;
> +
> + i915_obj_pidarray_validate(&obj->base);
> +
> + stats->num_obj++;
> +
> + if (obj->base.name) {
> +
> + if (drm_ht_find_item(&pid_entry->namelist,
> + (unsigned long)obj->base.name, &hash_item)) {
> + struct name_entry *entry =
> + kzalloc(sizeof(struct name_entry), GFP_KERNEL);
> + if (entry == NULL) {
> + DRM_ERROR("alloc failed\n");
> + return -ENOMEM;
> + }
> + entry->hash_item.key = obj->base.name;
> + drm_ht_insert_item(&pid_entry->namelist,
> + &entry->hash_item);
> + list_add_tail(&entry->head, &pid_entry->namefree);
> + } else {
> + DRM_DEBUG("Duplicate obj with name %d for process %s\n",
> + obj->base.name, stats->process_name);
> + return 0;
> + }
> + for (i = 0; i < MAX_OPEN_HANDLE; i++) {
> + if (obj->pid_array[i].pid != 0)
> + obj_shared_count++;
> + }
> + if (WARN_ON(obj_shared_count == 0))
> + return 1;
> +
> + DRM_DEBUG("Obj: %p, shared count =%d\n",
> + &obj->base, obj_shared_count);
> +
> + if (obj_shared_count > 1)
> + stats->num_obj_shared++;
> + else
> + stats->num_obj_private++;
> + } else {
> + obj_shared_count = 1;
> + stats->num_obj_private++;
> + }
> +
> + if (i915_gem_obj_bound_any(obj)) {
> + stats->num_obj_gtt_bound++;
> + if (obj_shared_count > 1)
> + stats->gtt_space_allocated_shared += obj->base.size;
> + else
> + stats->gtt_space_allocated_priv += obj->base.size;
> + }
> +
> + if (obj->stolen) {
> + stats->num_obj_stolen++;
> + stats->stolen_space_allocated += obj->base.size;
> + } else if (obj->madv == __I915_MADV_PURGED) {
> + stats->num_obj_purged++;
> + } else if (obj->madv == I915_MADV_DONTNEED) {
> + stats->num_obj_purgeable++;
> + stats->num_obj_allocated++;
> + if (i915_gem_obj_shmem_pages_alloced(obj) != 0) {
> + stats->phys_space_purgeable += obj->base.size;
> + if (obj_shared_count > 1) {
> + stats->phys_space_allocated_shared +=
> + obj->base.size;
> + stats->phys_space_shared_proportion +=
> + obj->base.size/obj_shared_count;
> + } else
> + stats->phys_space_allocated_priv +=
> + obj->base.size;
> + } else
> + WARN_ON(1);
> + } else if (i915_gem_obj_shmem_pages_alloced(obj) != 0) {
> + stats->num_obj_allocated++;
> + if (obj_shared_count > 1) {
> + stats->phys_space_allocated_shared +=
> + obj->base.size;
> + stats->phys_space_shared_proportion +=
> + obj->base.size/obj_shared_count;
> + }
> + else
> + stats->phys_space_allocated_priv += obj->base.size;
> + }
> + if (obj->fault_mappable) {
> + stats->num_obj_fault_mappable++;
> + stats->fault_mappable_size += obj->base.size;
> + }
> + return 0;
> +}
> +
> +int i915_get_drm_clients_info(struct drm_i915_error_state_buf *m,
> + struct drm_device *dev)
> +{
> + struct drm_file *file;
> + struct drm_i915_private *dev_priv = dev->dev_private;
> +
> + struct name_entry *entry, *next;
> + struct pid_stat_entry *pid_entry, *temp_entry;
> + struct pid_stat_entry *new_pid_entry, *new_temp_entry;
> + struct list_head per_pid_stats, sorted_pid_stats;
> + int ret = 0, total_shared_prop_space = 0, total_priv_space = 0;
> +
> + INIT_LIST_HEAD(&per_pid_stats);
> + INIT_LIST_HEAD(&sorted_pid_stats);
> +
> + err_printf(m,
> + "\n\n pid Total Shared Priv Purgeable Alloced SharedPHYsize SharedPHYprop PrivPHYsize PurgeablePHYsize process\n");
> +
> + /* Protect the access to global drm resources such as filelist. Protect
> + * against their removal under our noses, while in use.
> + */
> + mutex_lock(&drm_global_mutex);
> + ret = i915_mutex_lock_interruptible(dev);
> + if (ret) {
> + mutex_unlock(&drm_global_mutex);
> + return ret;
> + }
> +
> + list_for_each_entry(file, &dev->filelist, lhead) {
> + struct pid *tgid;
> + struct drm_i915_file_private *file_priv = file->driver_priv;
> + int pid_num, found = 0;
> +
> + tgid = file_priv->tgid;
> + pid_num = pid_nr(tgid);
> +
> + list_for_each_entry(pid_entry, &per_pid_stats, head) {
> + if (pid_entry->pid_num == pid_num) {
> + found = 1;
> + break;
> + }
> + }
> +
> + if (!found) {
> + struct pid_stat_entry *new_entry =
> + kzalloc(sizeof(struct pid_stat_entry),
> + GFP_KERNEL);
> + if (new_entry == NULL) {
> + DRM_ERROR("alloc failed\n");
> + ret = -ENOMEM;
> + goto out_unlock;
> + }
> + new_entry->pid = tgid;
> + new_entry->pid_num = pid_num;
> + list_add_tail(&new_entry->head, &per_pid_stats);
> + drm_ht_create(&new_entry->namelist,
> + DRM_MAGIC_HASH_ORDER);
> + INIT_LIST_HEAD(&new_entry->namefree);
> + new_entry->stats.process_name = file_priv->process_name;
> + pid_entry = new_entry;
> + }
> +
> + ret = idr_for_each(&file->object_idr,
> + &i915_drm_gem_object_per_file_summary, pid_entry);
> + if (ret)
> + break;
> + }
> +
> + list_for_each_entry_safe(pid_entry, temp_entry, &per_pid_stats, head) {
> + if (list_empty(&sorted_pid_stats)) {
> + list_del(&pid_entry->head);
> + list_add_tail(&pid_entry->head, &sorted_pid_stats);
> + continue;
> + }
> +
> + list_for_each_entry_safe(new_pid_entry, new_temp_entry,
> + &sorted_pid_stats, head) {
> + int prev_space =
> + pid_entry->stats.phys_space_shared_proportion +
> + pid_entry->stats.phys_space_allocated_priv;
> + int new_space =
> + new_pid_entry->
> + stats.phys_space_shared_proportion +
> + new_pid_entry->stats.phys_space_allocated_priv;
> + if (prev_space > new_space) {
> + list_del(&pid_entry->head);
> + list_add_tail(&pid_entry->head,
> + &new_pid_entry->head);
> + break;
> + }
> + if (list_is_last(&new_pid_entry->head,
> + &sorted_pid_stats)) {
> + list_del(&pid_entry->head);
> + list_add_tail(&pid_entry->head,
> + &sorted_pid_stats);
> + }
> + }
> + }
> +
> + list_for_each_entry_safe(pid_entry, temp_entry,
> + &sorted_pid_stats, head) {
> + struct task_struct *task = get_pid_task(pid_entry->pid,
> + PIDTYPE_PID);
> + err_printf(m,
> + "%5d %6d %6d %6d %9d %8d %14zdK %14zdK %14zdK %14zdK %s",
> + pid_entry->pid_num,
> + pid_entry->stats.num_obj,
> + pid_entry->stats.num_obj_shared,
> + pid_entry->stats.num_obj_private,
> + pid_entry->stats.num_obj_purgeable,
> + pid_entry->stats.num_obj_allocated,
> + pid_entry->stats.phys_space_allocated_shared/1024,
> + pid_entry->stats.phys_space_shared_proportion/1024,
> + pid_entry->stats.phys_space_allocated_priv/1024,
> + pid_entry->stats.phys_space_purgeable/1024,
> + pid_entry->stats.process_name);
> +
> + if (task == NULL)
> + err_printf(m, "*\n");
> + else
> + err_printf(m, "\n");
> +
> + total_shared_prop_space +=
> + pid_entry->stats.phys_space_shared_proportion/1024;
> + total_priv_space +=
> + pid_entry->stats.phys_space_allocated_priv/1024;
> + list_del(&pid_entry->head);
> +
> + list_for_each_entry_safe(entry, next,
> + &pid_entry->namefree, head) {
> + list_del(&entry->head);
> + drm_ht_remove_item(&pid_entry->namelist,
> + &entry->hash_item);
> + kfree(entry);
> + }
> + drm_ht_remove(&pid_entry->namelist);
> + kfree(pid_entry);
> + }
> +
> + err_printf(m,
> + "\t\t\t\t\t\t\t\t--------------\t-------------\t--------\n");
> + err_printf(m,
> + "\t\t\t\t\t\t\t\t%13zdK\t%12zdK\tTotal\n",
> + total_shared_prop_space, total_priv_space);
> +
> +out_unlock:
> + mutex_unlock(&dev->struct_mutex);
> + mutex_unlock(&drm_global_mutex);
> +
> + if (ret)
> + return ret;
> + if (m->bytes == 0 && m->err)
> + return m->err;
> +
> + return 0;
> +}
> +
> +int i915_gem_get_all_obj_info(struct drm_i915_error_state_buf *m,
> + struct drm_device *dev)
> +{
> + struct drm_file *file;
> + int pid_num, ret = 0;
> +
> + /* Protect the access to global drm resources such as filelist. Protect
> + * against their removal under our noses, while in use.
> + */
> + mutex_lock(&drm_global_mutex);
> + ret = i915_mutex_lock_interruptible(dev);
> + if (ret) {
> + mutex_unlock(&drm_global_mutex);
> + return ret;
> + }
> +
> + list_for_each_entry(file, &dev->filelist, lhead) {
> + struct pid *tgid;
> + struct drm_i915_file_private *file_priv = file->driver_priv;
> +
> + tgid = file_priv->tgid;
> + pid_num = pid_nr(tgid);
> +
> + err_printf(m, "\n\n PID process\n");
> +
> + err_printf(m, "%5d %s\n",
> + pid_num, file_priv->process_name);
> +
> + err_printf(m,
> + "\n Obj Identifier Size Pin Tiling Dirty Shared Vmap Stolen Mappable AllocState Global/PP GttOffset (PID: handle count: user virt addrs)\n");
> + ret = idr_for_each(&file->object_idr,
> + &i915_drm_gem_obj_info, m);
> + if (ret)
> + break;
> + }
> + mutex_unlock(&dev->struct_mutex);
> + mutex_unlock(&drm_global_mutex);
> +
> + if (ret)
> + return ret;
> + if (m->bytes == 0 && m->err)
> + return m->err;
> +
> + return 0;
> +}
> +
> diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
> index 2c87a79..089c7df 100644
> --- a/drivers/gpu/drm/i915/i915_gpu_error.c
> +++ b/drivers/gpu/drm/i915/i915_gpu_error.c
> @@ -161,7 +161,7 @@ static void i915_error_vprintf(struct drm_i915_error_state_buf *e,
> __i915_error_advance(e, len);
> }
>
> -static void i915_error_puts(struct drm_i915_error_state_buf *e,
> +void i915_error_puts(struct drm_i915_error_state_buf *e,
> const char *str)
> {
> unsigned len;
> diff --git a/drivers/gpu/drm/i915/i915_sysfs.c b/drivers/gpu/drm/i915/i915_sysfs.c
> index 503847f..b204c92 100644
> --- a/drivers/gpu/drm/i915/i915_sysfs.c
> +++ b/drivers/gpu/drm/i915/i915_sysfs.c
> @@ -582,6 +582,64 @@ static ssize_t error_state_write(struct file *file, struct kobject *kobj,
> return count;
> }
>
> +static ssize_t i915_gem_clients_state_read(struct file *filp,
> + struct kobject *kobj,
> + struct bin_attribute *attr,
> + char *buf, loff_t off, size_t count)
> +{
> + struct device *kdev = container_of(kobj, struct device, kobj);
> + struct drm_minor *minor = dev_to_drm_minor(kdev);
> + struct drm_device *dev = minor->dev;
> + struct drm_i915_error_state_buf error_str;
> + ssize_t ret_count = 0;
> + int ret;
> +
> + ret = i915_error_state_buf_init(&error_str, to_i915(dev), count, off);
> + if (ret)
> + return ret;
> +
> + ret = i915_get_drm_clients_info(&error_str, dev);
> + if (ret)
> + goto out;
> +
> + ret_count = count < error_str.bytes ? count : error_str.bytes;
> +
> + memcpy(buf, error_str.buf, ret_count);
> +out:
> + i915_error_state_buf_release(&error_str);
> +
> + return ret ?: ret_count;
> +}
> +
> +static ssize_t i915_gem_objects_state_read(struct file *filp,
> + struct kobject *kobj,
> + struct bin_attribute *attr,
> + char *buf, loff_t off, size_t count)
> +{
> + struct device *kdev = container_of(kobj, struct device, kobj);
> + struct drm_minor *minor = dev_to_drm_minor(kdev);
> + struct drm_device *dev = minor->dev;
> + struct drm_i915_error_state_buf error_str;
> + ssize_t ret_count = 0;
> + int ret;
> +
> + ret = i915_error_state_buf_init(&error_str, to_i915(dev), count, off);
> + if (ret)
> + return ret;
> +
> + ret = i915_gem_get_all_obj_info(&error_str, dev);
> + if (ret)
> + goto out;
> +
> + ret_count = count < error_str.bytes ? count : error_str.bytes;
> +
> + memcpy(buf, error_str.buf, ret_count);
> +out:
> + i915_error_state_buf_release(&error_str);
> +
> + return ret ?: ret_count;
> +}
> +
> static struct bin_attribute error_state_attr = {
> .attr.name = "error",
> .attr.mode = S_IRUSR | S_IWUSR,
> @@ -590,6 +648,20 @@ static struct bin_attribute error_state_attr = {
> .write = error_state_write,
> };
>
> +static struct bin_attribute i915_gem_client_state_attr = {
> + .attr.name = "i915_gem_meminfo",
> + .attr.mode = S_IRUSR | S_IWUSR,
> + .size = 0,
> + .read = i915_gem_clients_state_read,
> +};
> +
> +static struct bin_attribute i915_gem_objects_state_attr = {
> + .attr.name = "i915_gem_objinfo",
> + .attr.mode = S_IRUSR | S_IWUSR,
> + .size = 0,
> + .read = i915_gem_objects_state_read,
> +};
> +
> void i915_setup_sysfs(struct drm_device *dev)
> {
> int ret;
> @@ -627,6 +699,17 @@ void i915_setup_sysfs(struct drm_device *dev)
> &error_state_attr);
> if (ret)
> DRM_ERROR("error_state sysfs setup failed\n");
> +
> + ret = sysfs_create_bin_file(&dev->primary->kdev->kobj,
> + &i915_gem_client_state_attr);
> + if (ret)
> + DRM_ERROR("i915_gem_client_state sysfs setup failed\n");
> +
> + ret = sysfs_create_bin_file(&dev->primary->kdev->kobj,
> + &i915_gem_objects_state_attr);
> + if (ret)
> + DRM_ERROR("i915_gem_objects_state sysfs setup failed\n");
> +
> }
>
> void i915_teardown_sysfs(struct drm_device *dev)
> --
> 1.8.5.1
>
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx
--
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
More information about the Intel-gfx
mailing list