| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20250821142946.00110c49@fedora>
Date: Thu, 21 Aug 2025 14:29:46 +0200
From: Boris Brezillon <boris.brezillon@...labora.com>
To: "Danilo Krummrich" <dakr@...nel.org>
Cc: "Caterina Shablia" <caterina.shablia@...labora.com>, "Maarten Lankhorst"
<maarten.lankhorst@...ux.intel.com>, "Maxime Ripard" <mripard@...nel.org>,
"Thomas Zimmermann" <tzimmermann@...e.de>, "David Airlie"
<airlied@...il.com>, "Simona Vetter" <simona@...ll.ch>, "Frank Binns"
<frank.binns@...tec.com>, "Matt Coster" <matt.coster@...tec.com>, "Karol
Herbst" <kherbst@...hat.com>, "Lyude Paul" <lyude@...hat.com>, "Steven
Price" <steven.price@....com>, "Liviu Dudau" <liviu.dudau@....com>, "Lucas
De Marchi" <lucas.demarchi@...el.com>, Thomas Hellström
<thomas.hellstrom@...ux.intel.com>, "Rodrigo Vivi"
<rodrigo.vivi@...el.com>, <dri-devel@...ts.freedesktop.org>,
<linux-kernel@...r.kernel.org>, <nouveau@...ts.freedesktop.org>,
<intel-xe@...ts.freedesktop.org>, <asahi@...ts.linux.dev>, "Asahi Lina"
<lina@...hilina.net>, "Asahi Lina" <lina+kernel@...hilina.net>
Subject: Re: [PATCH v4 6/7] drm/gpuvm: Add DRM_GPUVA_REPEAT flag and logic
On Mon, 07 Jul 2025 21:33:13 +0200
"Danilo Krummrich" <dakr@...nel.org> wrote:
> On Mon Jul 7, 2025 at 7:04 PM CEST, Caterina Shablia wrote:
> > From: Asahi Lina <lina+kernel@...hilina.net>
> >
> > To be able to support "fake sparse" mappings without relying on GPU page
> > fault handling, drivers may need to create large (e.g. 4GiB) mappings of
> > the same page repeatedly (or same range of pages). Doing this through
> > individual mappings would be very wasteful. This can be handled better
> > by using a flag on map creation, but to do it safely, drm_gpuvm needs to
> > be aware of this special case.
> >
> > Add a flag that signals that a given mapping is a page mapping, which is
> > repeated all over the entire requested VA range. This tweaks the
> > sm_map() logic to treat the GEM offsets differently when mappings are
> > a repeated ones so they are not incremented as they would be with regular
> > mappings.
> >
> > The size of the GEM portion to repeat is passed through
> > drm_gpuva::gem::range. Most of the time it will be a page size, but
> > it can be bigger as long as it's less that drm_gpuva::va::range, and
> > drm_gpuva::gem::range is a multiple of drm_gpuva::va::range.
>
> Should be "as long as it's less that drm_gpuva::va::range, and
> drm_gpuva::va::range is a multiple of drm_gpuva::gem::range".
>
> I also think this feature deserves its own section in the global GPUVM
> documentation -- please add a corresponding paragraph.
Sure.
>
> > +static int check_map_req(struct drm_gpuvm *gpuvm,
>
> Let's call this validate_map_request().
I can do that, sure.
>
> > + const struct drm_gpuvm_map_req *req)
> > +{
> > + if (unlikely(!drm_gpuvm_range_valid(gpuvm, req->va.addr, req->va.range)))
> > + return -EINVAL;
> > +
> > + if (req->flags & DRM_GPUVA_REPEAT) {
> > + u64 va_range = req->va.range;
> > +
> > + /* For a repeated mapping, GEM range must be > 0
> > + * and a multiple of the VA range.
> > + */
> > + if (unlikely(!req->gem.range ||
> > + va_range < req->gem.range ||
> > + do_div(va_range, req->gem.range)))
> > + return -EINVAL;
> > + }
> > +
> > + return 0;
> > +}
> > +
> > static int
> > __drm_gpuvm_sm_map(struct drm_gpuvm *gpuvm,
> > const struct drm_gpuvm_ops *ops, void *priv,
> > @@ -2137,6 +2179,7 @@ __drm_gpuvm_sm_map(struct drm_gpuvm *gpuvm,
> > struct drm_gpuva reqva = {
> > .va.addr = req->va.addr,
> > .va.range = req->va.range,
> > + .gem.range = req->gem.range,
> > .gem.offset = req->gem.offset,
> > .gem.obj = req->gem.obj,
> > .flags = req->flags,
> > @@ -2144,7 +2187,8 @@ __drm_gpuvm_sm_map(struct drm_gpuvm *gpuvm,
> > u64 req_end = req->va.addr + req->va.range;
> > int ret;
> >
> > - if (unlikely(!drm_gpuvm_range_valid(gpuvm, req->va.addr, req->va.range)))
> > + ret = check_map_req(gpuvm, req);
> > + if (unlikely(ret))
> > return -EINVAL;
> >
> > drm_gpuvm_for_each_va_range_safe(va, next, gpuvm, req->va.addr, req_end) {
> > @@ -2175,7 +2219,8 @@ __drm_gpuvm_sm_map(struct drm_gpuvm *gpuvm,
> > .va.addr = req_end,
> > .va.range = range - req->va.range,
> > .gem.obj = obj,
> > - .gem.offset = offset + req->va.range,
> > + .gem.range = va->gem.range,
> > + .gem.offset = offset,
>
> Why change this from offset + req->va.range to just offset?
This is conditionally updated if DRM_GPUVA_REPEAT is not set further
down, because we don't want to move the GEM offset if the mapped portion
is repeated.
>
> Same for similar other changes below.
>
> Also it seems that we need to update the documentation which shows all potential
> cases when calling __drm_gpuvm_sm_map() [1].
Yep, will do.
>
> [1] https://docs.kernel.org/gpu/drm-mm.html#split-and-merge
>
> > .flags = va->flags,
> > };
> > struct drm_gpuva_op_unmap u = {
> > @@ -2183,6 +2228,9 @@ __drm_gpuvm_sm_map(struct drm_gpuvm *gpuvm,
> > .keep = merge,
> > };
> >
> > + if (!(va->flags & DRM_GPUVA_REPEAT))
> > + n.gem.offset += req->va.range;
> > +
> > ret = op_remap_cb(ops, priv, NULL, &n, &u);
> > if (ret)
> > return ret;
> > @@ -2194,6 +2242,7 @@ __drm_gpuvm_sm_map(struct drm_gpuvm *gpuvm,
> > .va.addr = addr,
> > .va.range = ls_range,
> > .gem.obj = obj,
> > + .gem.range = va->gem.range,
> > .gem.offset = offset,
> > .flags = va->flags,
> > };
> > @@ -2220,11 +2269,14 @@ __drm_gpuvm_sm_map(struct drm_gpuvm *gpuvm,
> > .va.addr = req_end,
> > .va.range = end - req_end,
> > .gem.obj = obj,
> > - .gem.offset = offset + ls_range +
> > - req->va.range,
> > + .gem.range = va->gem.range,
> > + .gem.offset = offset,
> > .flags = va->flags,
> > };
> >
> > + if (!(va->flags & DRM_GPUVA_REPEAT))
> > + n.gem.offset += ls_range + req->va.range;
> > +
> > ret = op_remap_cb(ops, priv, &p, &n, &u);
> > if (ret)
> > return ret;
> > @@ -2250,7 +2302,8 @@ __drm_gpuvm_sm_map(struct drm_gpuvm *gpuvm,
> > .va.addr = req_end,
> > .va.range = end - req_end,
> > .gem.obj = obj,
> > - .gem.offset = offset + req_end - addr,
> > + .gem.range = va->gem.range,
> > + .gem.offset = offset,
> > .flags = va->flags,
> > };
> > struct drm_gpuva_op_unmap u = {
> > @@ -2258,6 +2311,9 @@ __drm_gpuvm_sm_map(struct drm_gpuvm *gpuvm,
> > .keep = merge,
> > };
> >
> > + if (!(va->flags & DRM_GPUVA_REPEAT))
> > + n.gem.offset += req_end - addr;
> > +
> > ret = op_remap_cb(ops, priv, NULL, &n, &u);
> > if (ret)
> > return ret;
> > @@ -2294,6 +2350,7 @@ __drm_gpuvm_sm_unmap(struct drm_gpuvm *gpuvm,
> > prev.va.addr = addr;
> > prev.va.range = req_addr - addr;
> > prev.gem.obj = obj;
> > + prev.gem.range = va->gem.range;
> > prev.gem.offset = offset;
> > prev.flags = va->flags;
> >
> > @@ -2304,7 +2361,10 @@ __drm_gpuvm_sm_unmap(struct drm_gpuvm *gpuvm,
> > next.va.addr = req_end;
> > next.va.range = end - req_end;
> > next.gem.obj = obj;
> > - next.gem.offset = offset + (req_end - addr);
> > + prev.gem.range = va->gem.range;
> > + next.gem.offset = offset;
> > + if (!(va->flags & DRM_GPUVA_REPEAT))
> > + next.gem.offset += req_end - addr;
> > next.flags = va->flags;
> >
> > next_split = true;
> > diff --git a/include/drm/drm_gpuvm.h b/include/drm/drm_gpuvm.h
> > index f77a89e791f1..629e8508f99f 100644
> > --- a/include/drm/drm_gpuvm.h
> > +++ b/include/drm/drm_gpuvm.h
> > @@ -56,10 +56,19 @@ enum drm_gpuva_flags {
> > */
> > DRM_GPUVA_SPARSE = (1 << 1),
> >
> > + /**
> > + * @DRM_GPUVA_REPEAT:
> > + *
> > + * Flag indicating that the &drm_gpuva is a mapping of a GEM
> > + * portion repeated multiple times to fill the virtual address
>
> "of a GEM object with a certain range that is repeated multiple times to ..."
Powered by blists - more mailing lists