Re: [PATCH 0/2] drm/bridge: start moving towards struct drm_edid
On Fri, 22 Dec 2023, Jani Nikula wrote: > Okay, I rebased and pushed [1]. Probably doesn't make sense to send a > patch bomb like that right now... And here are the patches: https://patchwork.freedesktop.org/series/128149/ BR, Jani. > > BR, > Jani. > > > [1] https://gitlab.freedesktop.org/jani/linux/-/commits/drm-edid-bridge -- Jani Nikula, Intel
Re: [PATCH 0/2] drm/bridge: start moving towards struct drm_edid
On Fri, 22 Dec 2023, Jani Nikula wrote: > On Fri, 22 Dec 2023, Neil Armstrong wrote: >> On 19/12/2023 13:15, Jani Nikula wrote: >>> On Tue, 14 Nov 2023, Jani Nikula wrote: On Thu, 26 Oct 2023, Jani Nikula wrote: > This is just the first two patches of a lengthy series that I'm not > really sure how to proceed with. Basically the series converts all of > drm/bridge to the new struct drm_edid infrastructure. It's safer than > struct edid, because it contains meta information about the allocated > size of the EDID, instead of relying on the size (number of extensions) > originating from outside of the kernel. > > The rest is at [1]. The commit messages are lacking, and I don't really > have the toolchain to even build test most of it. But I think this is > where drm/bridge should go. Among all of drm, I think bridge has the > most uses of struct edid that do not originate from the drm_get_edid() > family of functions, which means the validity checks are somewhat > inconsistent, and having the meta information is more crucial. > > Bridge maintainers, please instruct how to best proceed with this. Ping. >>> >>> Ping. >> >> Sorry for the delay, I would have preferred changing the get_edid to return >> a drm_edid, but I understand the task is too high, could you instead use >> @get_drm_edid instead of @edid_read ? > > edid_read matches the changes in drm_edid.c, going from drm_get_edid() > to drm_edid_read(). > > There's a nice symmetry when ->get_edid() hooks using drm_get_edid() are > converted to ->edid_read() hooks using drm_edid_read(). > >> And perhaps convert one very common bridge to this so we can validate >> the change in CI. > > So I did convert all of bridge over a few months back, and pushed the > branch to [1]. Should I brush that up and send the entire series? I > don't really know what's common and what's not. Okay, I rebased and pushed [1]. Probably doesn't make sense to send a patch bomb like that right now... BR, Jani. [1] https://gitlab.freedesktop.org/jani/linux/-/commits/drm-edid-bridge -- Jani Nikula, Intel
Re: [PATCH 0/2] drm/bridge: start moving towards struct drm_edid
On Fri, 22 Dec 2023, Neil Armstrong wrote: > On 19/12/2023 13:15, Jani Nikula wrote: >> On Tue, 14 Nov 2023, Jani Nikula wrote: >>> On Thu, 26 Oct 2023, Jani Nikula wrote: This is just the first two patches of a lengthy series that I'm not really sure how to proceed with. Basically the series converts all of drm/bridge to the new struct drm_edid infrastructure. It's safer than struct edid, because it contains meta information about the allocated size of the EDID, instead of relying on the size (number of extensions) originating from outside of the kernel. The rest is at [1]. The commit messages are lacking, and I don't really have the toolchain to even build test most of it. But I think this is where drm/bridge should go. Among all of drm, I think bridge has the most uses of struct edid that do not originate from the drm_get_edid() family of functions, which means the validity checks are somewhat inconsistent, and having the meta information is more crucial. Bridge maintainers, please instruct how to best proceed with this. >>> >>> Ping. >> >> Ping. > > Sorry for the delay, I would have preferred changing the get_edid to return > a drm_edid, but I understand the task is too high, could you instead use > @get_drm_edid instead of @edid_read ? edid_read matches the changes in drm_edid.c, going from drm_get_edid() to drm_edid_read(). There's a nice symmetry when ->get_edid() hooks using drm_get_edid() are converted to ->edid_read() hooks using drm_edid_read(). > And perhaps convert one very common bridge to this so we can validate > the change in CI. So I did convert all of bridge over a few months back, and pushed the branch to [1]. Should I brush that up and send the entire series? I don't really know what's common and what's not. BR, Jani. > > Neil > >> >>> >>> The two patches posted here could be merged, to add something to build >>> the later commits on gradually. >>> >>> BR, >>> Jani. >>> Thanks, Jani. [1] https://gitlab.freedesktop.org/jani/linux/-/commits/drm-edid-bridge Jani Nikula (2): drm/bridge: add ->edid_read hook and drm_bridge_edid_read() drm/bridge: switch to drm_bridge_read_edid() drivers/gpu/drm/drm_bridge.c | 46 +- drivers/gpu/drm/drm_bridge_connector.c | 16 - include/drm/drm_bridge.h | 33 ++ 3 files changed, 86 insertions(+), 9 deletions(-) >> > -- Jani Nikula, Intel
Re: [PATCH 0/2] drm/bridge: start moving towards struct drm_edid
On 19/12/2023 13:15, Jani Nikula wrote: On Tue, 14 Nov 2023, Jani Nikula wrote: On Thu, 26 Oct 2023, Jani Nikula wrote: This is just the first two patches of a lengthy series that I'm not really sure how to proceed with. Basically the series converts all of drm/bridge to the new struct drm_edid infrastructure. It's safer than struct edid, because it contains meta information about the allocated size of the EDID, instead of relying on the size (number of extensions) originating from outside of the kernel. The rest is at [1]. The commit messages are lacking, and I don't really have the toolchain to even build test most of it. But I think this is where drm/bridge should go. Among all of drm, I think bridge has the most uses of struct edid that do not originate from the drm_get_edid() family of functions, which means the validity checks are somewhat inconsistent, and having the meta information is more crucial. Bridge maintainers, please instruct how to best proceed with this. Ping. Ping. Sorry for the delay, I would have preferred changing the get_edid to return a drm_edid, but I understand the task is too high, could you instead use @get_drm_edid instead of @edid_read ? And perhaps convert one very common bridge to this so we can validate the change in CI. Neil The two patches posted here could be merged, to add something to build the later commits on gradually. BR, Jani. Thanks, Jani. [1] https://gitlab.freedesktop.org/jani/linux/-/commits/drm-edid-bridge Jani Nikula (2): drm/bridge: add ->edid_read hook and drm_bridge_edid_read() drm/bridge: switch to drm_bridge_read_edid() drivers/gpu/drm/drm_bridge.c | 46 +- drivers/gpu/drm/drm_bridge_connector.c | 16 - include/drm/drm_bridge.h | 33 ++ 3 files changed, 86 insertions(+), 9 deletions(-)
Re: [PATCH 0/2] drm/bridge: start moving towards struct drm_edid
On Tue, 14 Nov 2023, Jani Nikula wrote: > On Thu, 26 Oct 2023, Jani Nikula wrote: >> This is just the first two patches of a lengthy series that I'm not >> really sure how to proceed with. Basically the series converts all of >> drm/bridge to the new struct drm_edid infrastructure. It's safer than >> struct edid, because it contains meta information about the allocated >> size of the EDID, instead of relying on the size (number of extensions) >> originating from outside of the kernel. >> >> The rest is at [1]. The commit messages are lacking, and I don't really >> have the toolchain to even build test most of it. But I think this is >> where drm/bridge should go. Among all of drm, I think bridge has the >> most uses of struct edid that do not originate from the drm_get_edid() >> family of functions, which means the validity checks are somewhat >> inconsistent, and having the meta information is more crucial. >> >> Bridge maintainers, please instruct how to best proceed with this. > > Ping. Ping. > > The two patches posted here could be merged, to add something to build > the later commits on gradually. > > BR, > Jani. > >> >> >> Thanks, >> Jani. >> >> >> >> [1] https://gitlab.freedesktop.org/jani/linux/-/commits/drm-edid-bridge >> >> >> >> Jani Nikula (2): >> drm/bridge: add ->edid_read hook and drm_bridge_edid_read() >> drm/bridge: switch to drm_bridge_read_edid() >> >> drivers/gpu/drm/drm_bridge.c | 46 +- >> drivers/gpu/drm/drm_bridge_connector.c | 16 - >> include/drm/drm_bridge.h | 33 ++ >> 3 files changed, 86 insertions(+), 9 deletions(-) -- Jani Nikula, Intel
Re: [PATCH 0/2] drm/bridge: start moving towards struct drm_edid
On Thu, 26 Oct 2023, Jani Nikula wrote: > This is just the first two patches of a lengthy series that I'm not > really sure how to proceed with. Basically the series converts all of > drm/bridge to the new struct drm_edid infrastructure. It's safer than > struct edid, because it contains meta information about the allocated > size of the EDID, instead of relying on the size (number of extensions) > originating from outside of the kernel. > > The rest is at [1]. The commit messages are lacking, and I don't really > have the toolchain to even build test most of it. But I think this is > where drm/bridge should go. Among all of drm, I think bridge has the > most uses of struct edid that do not originate from the drm_get_edid() > family of functions, which means the validity checks are somewhat > inconsistent, and having the meta information is more crucial. > > Bridge maintainers, please instruct how to best proceed with this. Ping. The two patches posted here could be merged, to add something to build the later commits on gradually. BR, Jani. > > > Thanks, > Jani. > > > > [1] https://gitlab.freedesktop.org/jani/linux/-/commits/drm-edid-bridge > > > > Jani Nikula (2): > drm/bridge: add ->edid_read hook and drm_bridge_edid_read() > drm/bridge: switch to drm_bridge_read_edid() > > drivers/gpu/drm/drm_bridge.c | 46 +- > drivers/gpu/drm/drm_bridge_connector.c | 16 - > include/drm/drm_bridge.h | 33 ++ > 3 files changed, 86 insertions(+), 9 deletions(-) -- Jani Nikula, Intel
[PATCH 0/2] drm/bridge: start moving towards struct drm_edid
This is just the first two patches of a lengthy series that I'm not really sure how to proceed with. Basically the series converts all of drm/bridge to the new struct drm_edid infrastructure. It's safer than struct edid, because it contains meta information about the allocated size of the EDID, instead of relying on the size (number of extensions) originating from outside of the kernel. The rest is at [1]. The commit messages are lacking, and I don't really have the toolchain to even build test most of it. But I think this is where drm/bridge should go. Among all of drm, I think bridge has the most uses of struct edid that do not originate from the drm_get_edid() family of functions, which means the validity checks are somewhat inconsistent, and having the meta information is more crucial. Bridge maintainers, please instruct how to best proceed with this. Thanks, Jani. [1] https://gitlab.freedesktop.org/jani/linux/-/commits/drm-edid-bridge Jani Nikula (2): drm/bridge: add ->edid_read hook and drm_bridge_edid_read() drm/bridge: switch to drm_bridge_read_edid() drivers/gpu/drm/drm_bridge.c | 46 +- drivers/gpu/drm/drm_bridge_connector.c | 16 - include/drm/drm_bridge.h | 33 ++ 3 files changed, 86 insertions(+), 9 deletions(-) -- 2.39.2