Message-ID: <ZDcEGoSPGr/oRLas@phenom.ffwll.local>
Date:   Wed, 12 Apr 2023 21:18:50 +0200
From:   Daniel Vetter <daniel@...ll.ch>
To:     Rob Clark <robdclark@...il.com>
Cc:     Tvrtko Ursulin <tvrtko.ursulin@...ux.intel.com>,
        dri-devel@...ts.freedesktop.org, linux-arm-msm@...r.kernel.org,
        freedreno@...ts.freedesktop.org,
        Boris Brezillon <boris.brezillon@...labora.com>,
        Christopher Healy <healych@...zon.com>,
        Emil Velikov <emil.l.velikov@...il.com>,
        Rob Clark <robdclark@...omium.org>,
        David Airlie <airlied@...il.com>,
        Maarten Lankhorst <maarten.lankhorst@...ux.intel.com>,
        Maxime Ripard <mripard@...nel.org>,
        Thomas Zimmermann <tzimmermann@...e.de>,
        Jonathan Corbet <corbet@....net>,
        "open list:DOCUMENTATION" <linux-doc@...r.kernel.org>,
        open list <linux-kernel@...r.kernel.org>,
        Daniel Vetter <daniel@...ll.ch>
Subject: Re: [PATCH v3 6/7] drm: Add fdinfo memory stats

On Wed, Apr 12, 2023 at 11:42:07AM -0700, Rob Clark wrote:
> On Wed, Apr 12, 2023 at 11:17 AM Daniel Vetter <daniel@...ll.ch> wrote:
> >
> > On Wed, Apr 12, 2023 at 10:59:54AM -0700, Rob Clark wrote:
> > > On Wed, Apr 12, 2023 at 7:42 AM Tvrtko Ursulin
> > > <tvrtko.ursulin@...ux.intel.com> wrote:
> > > >
> > > >
> > > > On 11/04/2023 23:56, Rob Clark wrote:
> > > > > From: Rob Clark <robdclark@...omium.org>
> > > > >
> > > > > Add support to dump GEM stats to fdinfo.
> > > > >
> > > > > v2: Fix typos, change size units to match docs, use div_u64
> > > > > v3: Do it in core
> > > > >
> > > > > Signed-off-by: Rob Clark <robdclark@...omium.org>
> > > > > Reviewed-by: Emil Velikov <emil.l.velikov@...il.com>
> > > > > ---
> > > > >   Documentation/gpu/drm-usage-stats.rst | 21 ++++++++
> > > > >   drivers/gpu/drm/drm_file.c            | 76 +++++++++++++++++++++++++++
> > > > >   include/drm/drm_file.h                |  1 +
> > > > >   include/drm/drm_gem.h                 | 19 +++++++
> > > > >   4 files changed, 117 insertions(+)
> > > > >
> > > > > diff --git a/Documentation/gpu/drm-usage-stats.rst b/Documentation/gpu/drm-usage-stats.rst
> > > > > index b46327356e80..b5e7802532ed 100644
> > > > > --- a/Documentation/gpu/drm-usage-stats.rst
> > > > > +++ b/Documentation/gpu/drm-usage-stats.rst
> > > > > @@ -105,6 +105,27 @@ object belong to this client, in the respective memory region.
> > > > >   Default unit shall be bytes with optional unit specifiers of 'KiB' or 'MiB'
> > > > >   indicating kibi- or mebi-bytes.
> > > > >
> > > > > +- drm-shared-memory: <uint> [KiB|MiB]
> > > > > +
> > > > > +The total size of buffers that are shared with another file (ie. have more
> > > > > +than a single handle).
> > > > > +
> > > > > +- drm-private-memory: <uint> [KiB|MiB]
> > > > > +
> > > > > +The total size of buffers that are not shared with another file.
> > > > > +
> > > > > +- drm-resident-memory: <uint> [KiB|MiB]
> > > > > +
> > > > > +The total size of buffers that are resident in system memory.
> > > >
> > > > I think this naming maybe does not work best with the existing
> > > > drm-memory-<region> keys.
> > >
> > > Actually, it was very deliberate not to conflict with the existing
> > > drm-memory-<region> keys ;-)
> > >
> > > I wouldn't have preferred drm-memory-{active,resident,...} but it
> > > could be mis-parsed by existing userspace so my hands were a bit tied.
> > >
> > > > How about introduce the concept of a memory region from the start and
> > > > use naming similar like we do for engines?
> > > >
> > > > drm-memory-$CATEGORY-$REGION: ...
> > > >
> > > > Then we document a bunch of categories and their semantics, for instance:
> > > >
> > > > 'size' - All reachable objects
> > > > 'shared' - Subset of 'size' with handle_count > 1
> > > > 'resident' - Objects with backing store
> > > > 'active' - Objects in use, subset of resident
> > > > 'purgeable' - Or inactive? Subset of resident.
> > > >
> > > > We keep the same semantics as with process memory accounting (if I got
> > > > it right) which could be desirable for a simplified mental model.
> > > >
> > > > (AMD needs to remind me of their 'drm-memory-...' keys semantics. If we
> > > > correctly captured this in the first round it should be equivalent to
> > > > 'resident' above. In any case we can document no category is equal to
> > > > which category, and at most one of the two must be output.)
> > > >
> > > > Region names we at most partially standardize. Like we could say
> > > > 'system' is to be used where backing store is system RAM and others are
> > > > driver defined.
> > > >
> > > > Then discrete GPUs could emit N sets of key-values, one for each memory
> > > > region they support.
> > > >
> > > > I think this all also works for objects which can be migrated between
> > > > memory regions. 'Size' accounts them against all regions while for
> > > > 'resident' they only appear in the region of their current placement, etc.
> > >
> > > I'm not too sure how to rectify different memory regions with this,
> > > since drm core doesn't really know about the driver's memory regions.
> > > Perhaps we can go back to this being a helper and drivers with vram
> > > just don't use the helper?  Or??
> >
> > I think if you flip it around to drm-$CATEGORY-memory{-$REGION}: then it
> > all works out reasonably consistently?
> 
> That is basically what we have now.  I could append -system to each to
> make things easier to add vram/etc (from a uabi standpoint)..

What you have isn't really -system, but everything. So doesn't really make
sense to me to mark this -system, it's only really true for integrated (if
they don't have stolen or something like that).

Also my comment was more in reply to Tvrtko's suggestion.
-Daniel


> 
> BR,
> -R
> 
> > And ttm could/should perhaps provide a helper to dump the region specific
> > version of this. Or we lift the concept of regions out of ttm a bit
> > higher, that's kinda needed for cgroups eventually anyway I think.
> > -Daniel
> >
> > >
> > > BR,
> > > -R
> > >
> > > > Userspace can aggregate if it wishes to do so but kernel side should not.
> > > >
> > > > > +
> > > > > +- drm-purgeable-memory: <uint> [KiB|MiB]
> > > > > +
> > > > > +The total size of buffers that are purgeable.
> > > > > +
> > > > > +- drm-active-memory: <uint> [KiB|MiB]
> > > > > +
> > > > > +The total size of buffers that are active on one or more rings.
> > > > > +
> > > > >   - drm-cycles-<str> <uint>
> > > > >
> > > > >   Engine identifier string must be the same as the one specified in the
> > > > > diff --git a/drivers/gpu/drm/drm_file.c b/drivers/gpu/drm/drm_file.c
> > > > > index 37dfaa6be560..46fdd843bb3a 100644
> > > > > --- a/drivers/gpu/drm/drm_file.c
> > > > > +++ b/drivers/gpu/drm/drm_file.c
> > > > > @@ -42,6 +42,7 @@
> > > > >   #include <drm/drm_client.h>
> > > > >   #include <drm/drm_drv.h>
> > > > >   #include <drm/drm_file.h>
> > > > > +#include <drm/drm_gem.h>
> > > > >   #include <drm/drm_print.h>
> > > > >
> > > > >   #include "drm_crtc_internal.h"
> > > > > @@ -871,6 +872,79 @@ void drm_send_event(struct drm_device *dev, struct drm_pending_event *e)
> > > > >   }
> > > > >   EXPORT_SYMBOL(drm_send_event);
> > > > >
> > > > > +static void print_size(struct drm_printer *p, const char *stat, size_t sz)
> > > > > +{
> > > > > +     const char *units[] = {"", " KiB", " MiB"};
> > > > > +     unsigned u;
> > > > > +
> > > > > +     for (u = 0; u < ARRAY_SIZE(units) - 1; u++) {
> > > > > +             if (sz < SZ_1K)
> > > > > +                     break;
> > > > > +             sz = div_u64(sz, SZ_1K);
> > > > > +     }
> > > > > +
> > > > > +     drm_printf(p, "%s:\t%zu%s\n", stat, sz, units[u]);
> > > > > +}
> > > > > +
> > > > > +static void print_memory_stats(struct drm_printer *p, struct drm_file *file)
> > > > > +{
> > > > > +     struct drm_gem_object *obj;
> > > > > +     struct {
> > > > > +             size_t shared;
> > > > > +             size_t private;
> > > > > +             size_t resident;
> > > > > +             size_t purgeable;
> > > > > +             size_t active;
> > > > > +     } size = {0};
> > > > > +     bool has_status = false;
> > > > > +     int id;
> > > > > +
> > > > > +     spin_lock(&file->table_lock);
> > > > > +     idr_for_each_entry (&file->object_idr, obj, id) {
> > > > > +             enum drm_gem_object_status s = 0;
> > > > > +
> > > > > +             if (obj->funcs && obj->funcs->status) {
> > > > > +                     s = obj->funcs->status(obj);
> > > > > +                     has_status = true;
> > > > > +             }
> > > > > +
> > > > > +             if (obj->handle_count > 1) {
> > > > > +                     size.shared += obj->size;
> > > > > +             } else {
> > > > > +                     size.private += obj->size;
> > > > > +             }
> > > > > +
> > > > > +             if (s & DRM_GEM_OBJECT_RESIDENT) {
> > > > > +                     size.resident += obj->size;
> > > > > +             } else {
> > > > > +                     /* If already purged or not yet backed by pages, don't
> > > > > +                      * count it as purgeable:
> > > > > +                      */
> > > > > +                     s &= ~DRM_GEM_OBJECT_PURGEABLE;
> > > >
> > > > Side question - why couldn't resident buffers be purgeable? Did you mean
> > > > for the if branch check to be active here? But then it wouldn't make
> > > > sense for a driver to report active _and_ purgeable..
> > > >
> > > > > +             }
> > > > > +
> > > > > +             if (!dma_resv_test_signaled(obj->resv, dma_resv_usage_rw(true))) {
> > > > > +                     size.active += obj->size;
> > > > > +
> > > > > +                     /* If still active, don't count as purgeable: */
> > > > > +                     s &= ~DRM_GEM_OBJECT_PURGEABLE;
> > > >
> > > > Another side question - I guess this tidies a race in reporting? If so
> > > > not sure it matters given the stats are all rather approximate.
> > > >
> > > > > +             }
> > > > > +
> > > > > +             if (s & DRM_GEM_OBJECT_PURGEABLE)
> > > > > +                     size.purgeable += obj->size;
> > > > > +     }
> > > >
> > > > One concern I have here is that it is all based on obj->size. That is,
> > > > there is no provision for drivers to implement page level granularity.
> > > > So correct reporting in use cases such as VM BIND in the future wouldn't
> > > > work unless it was a driver hook to get almost all of the info above. At
> > > > which point common code is just a loop. TBF I don't know if any drivers
> > > > do sub obj->size backing store granularity today, but I think it is
> > > > sometimes to be sure of before proceeding.
> > > >
> > > > Second concern is what I touched upon in the first reply block - if the
> > > > common code blindly loops over all objects then on discrete GPUs it
> > > > seems we get an 'aggregate' value here which is not what I think we
> > > > want. We rather want to have the ability for drivers to list stats per
> > > > individual memory region.
> > > >
> > > > > +     spin_unlock(&file->table_lock);
> > > > > +
> > > > > +     print_size(p, "drm-shared-memory", size.shared);
> > > > > +     print_size(p, "drm-private-memory", size.private);
> > > > > +     print_size(p, "drm-active-memory", size.active);
> > > > > +
> > > > > +     if (has_status) {
> > > > > +             print_size(p, "drm-resident-memory", size.resident);
> > > > > +             print_size(p, "drm-purgeable-memory", size.purgeable);
> > > > > +     }
> > > > > +}
> > > > > +
> > > > >   /**
> > > > >    * drm_fop_show_fdinfo - helper for drm file fops
> > > > >    * @seq_file: output stream
> > > > > @@ -904,6 +978,8 @@ void drm_fop_show_fdinfo(struct seq_file *m, struct file *f)
> > > > >
> > > > >       if (dev->driver->show_fdinfo)
> > > > >               dev->driver->show_fdinfo(&p, file);
> > > > > +
> > > > > +     print_memory_stats(&p, file);
> > > > >   }
> > > > >   EXPORT_SYMBOL(drm_fop_show_fdinfo);
> > > > >
> > > > > diff --git a/include/drm/drm_file.h b/include/drm/drm_file.h
> > > > > index dfa995b787e1..e5b40084538f 100644
> > > > > --- a/include/drm/drm_file.h
> > > > > +++ b/include/drm/drm_file.h
> > > > > @@ -41,6 +41,7 @@
> > > > >   struct dma_fence;
> > > > >   struct drm_file;
> > > > >   struct drm_device;
> > > > > +struct drm_printer;
> > > > >   struct device;
> > > > >   struct file;
> > > > >
> > > > > diff --git a/include/drm/drm_gem.h b/include/drm/drm_gem.h
> > > > > index 189fd618ca65..213917bb6b11 100644
> > > > > --- a/include/drm/drm_gem.h
> > > > > +++ b/include/drm/drm_gem.h
> > > > > @@ -42,6 +42,14 @@
> > > > >   struct iosys_map;
> > > > >   struct drm_gem_object;
> > > > >
> > > > > +/**
> > > > > + * enum drm_gem_object_status - bitmask of object state for fdinfo reporting
> > > > > + */
> > > > > +enum drm_gem_object_status {
> > > > > +     DRM_GEM_OBJECT_RESIDENT  = BIT(0),
> > > > > +     DRM_GEM_OBJECT_PURGEABLE = BIT(1),
> > > > > +};
> > > > > +
> > > > >   /**
> > > > >    * struct drm_gem_object_funcs - GEM object functions
> > > > >    */
> > > > > @@ -174,6 +182,17 @@ struct drm_gem_object_funcs {
> > > > >        */
> > > > >       int (*evict)(struct drm_gem_object *obj);
> > > > >
> > > > > +     /**
> > > > > +      * @status:
> > > > > +      *
> > > > > +      * The optional status callback can return additional object state
> > > > > +      * which determines which stats the object is counted against.  The
> > > > > +      * callback is called under table_lock.  Racing against object status
> > > > > +      * change is "harmless", and the callback can expect to not race
> > > > > +      * against object destruction.
> > > > > +      */
> > > > > +     enum drm_gem_object_status (*status)(struct drm_gem_object *obj);
> > > >
> > > > Does this needs to be in object funcs and couldn't be consolidated to
> > > > driver level?
> > > >
> > > > Regards,
> > > >
> > > > Tvrtko
> > > >
> > > > > +
> > > > >       /**
> > > > >        * @vm_ops:
> > > > >        *
> >
> > --
> > Daniel Vetter
> > Software Engineer, Intel Corporation
> > http://blog.ffwll.ch

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch

Powered by blists - more mailing lists