| Message ID | 20220119110839.33187-3-deller@gmx.de (mailing list archive) |
|---|---|
| State | Superseded, archived |
| Headers | show |
| Series | Fix regression introduced by disabling accelerated scrolling in fbcon | expand |
On Wed, Jan 19, 2022 at 12:08:39PM +0100, Helge Deller wrote: > This reverts commit 39aead8373b3c20bb5965c024dfb51a94e526151. > > Revert this patch. This patch started to introduce the regression that > all hardware acceleration of more than 35 existing fbdev drivers were > bypassed and thus fbcon console output for those was dramatically slowed > down by factor of 10 and more. > > Reverting this commit has no impact on DRM, since none of the DRM drivers are > tagged with the acceleration flags FBINFO_HWACCEL_COPYAREA, > FBINFO_HWACCEL_FILLRECT or others. > > Signed-off-by: Helge Deller <deller@gmx.de> > Cc: stable@vger.kernel.org # v5.16 Why just 5.16? This commit came in on 5.11 and was backported to 5.10.5. As for "why", I think there was a number of private bugs that were reported in this code, which is why it was removed. I do not think it can be safely added back in without addressing them first. Let me go dig through my email to see if I can find them... thanks, greg k-h
On Wed, Jan 19, 2022 at 12:22:55PM +0100, Greg Kroah-Hartman wrote: > On Wed, Jan 19, 2022 at 12:08:39PM +0100, Helge Deller wrote: > > This reverts commit 39aead8373b3c20bb5965c024dfb51a94e526151. > > > > Revert this patch. This patch started to introduce the regression that > > all hardware acceleration of more than 35 existing fbdev drivers were > > bypassed and thus fbcon console output for those was dramatically slowed > > down by factor of 10 and more. > > > > Reverting this commit has no impact on DRM, since none of the DRM drivers are > > tagged with the acceleration flags FBINFO_HWACCEL_COPYAREA, > > FBINFO_HWACCEL_FILLRECT or others. > > > > Signed-off-by: Helge Deller <deller@gmx.de> > > Cc: stable@vger.kernel.org # v5.16 > > Why just 5.16? This commit came in on 5.11 and was backported to > 5.10.5. > > As for "why", I think there was a number of private bugs that were > reported in this code, which is why it was removed. I do not think it > can be safely added back in without addressing them first. Let me go > dig through my email to see if I can find them... Ah, no, that was just the soft scrollback code I was thinking of, which was a different revert and is still gone, thankfully :) This one was just removed because Daniel noticed that only 3 drivers used this (nouveau, omapdrm, and gma600), so this shouldn't have caused any regressions in any other drivers like you are reporting here. So perhaps this regression is caused by something else? thanks, greg k-h
Hi Greg, On Wed, Jan 19, 2022 at 12:28 PM Greg Kroah-Hartman <gregkh@linuxfoundation.org> wrote: > On Wed, Jan 19, 2022 at 12:22:55PM +0100, Greg Kroah-Hartman wrote: > > On Wed, Jan 19, 2022 at 12:08:39PM +0100, Helge Deller wrote: > > > This reverts commit 39aead8373b3c20bb5965c024dfb51a94e526151. > > > > > > Revert this patch. This patch started to introduce the regression that > > > all hardware acceleration of more than 35 existing fbdev drivers were > > > bypassed and thus fbcon console output for those was dramatically slowed > > > down by factor of 10 and more. > > > > > > Reverting this commit has no impact on DRM, since none of the DRM drivers are > > > tagged with the acceleration flags FBINFO_HWACCEL_COPYAREA, > > > FBINFO_HWACCEL_FILLRECT or others. > > > > > > Signed-off-by: Helge Deller <deller@gmx.de> > > As for "why", I think there was a number of private bugs that were > > reported in this code, which is why it was removed. I do not think it > > can be safely added back in without addressing them first. Let me go > > dig through my email to see if I can find them... > > Ah, no, that was just the soft scrollback code I was thinking of, which So the bugs argument is moot. > was a different revert and is still gone, thankfully :) FTR, not everyone else was thankful about that one... > This one was just removed because Daniel noticed that only 3 drivers > used this (nouveau, omapdrm, and gma600), so this shouldn't have caused > any regressions in any other drivers like you are reporting here. > > So perhaps this regression is caused by something else? 1. Daniel's patch was not CCed to linux-fbdev, 2. When I discovered the patch, I pointed out that the premise of 3 drivers was not true, and that it affects 32 more fbdev drivers[1] . The patch was applied regardless. 3. When the patch was suggested for backporting, I pointed out the same[2]. The patch was backported regardless. [1] https://lore.kernel.org/r/alpine.DEB.2.22.394.2010311116530.379363@ramsan.of.borg/ [2] https://lore.kernel.org/r/CAMuHMdXRgam2zahPEGcw8+76Xm-0AO-Ci9-YmVa5JpTKVHphRw@mail.gmail.com/ Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds
Hello Greg, On 1/19/22 12:47, Geert Uytterhoeven wrote: > On Wed, Jan 19, 2022 at 12:28 PM Greg Kroah-Hartman > <gregkh@linuxfoundation.org> wrote: >> On Wed, Jan 19, 2022 at 12:22:55PM +0100, Greg Kroah-Hartman wrote: >>> On Wed, Jan 19, 2022 at 12:08:39PM +0100, Helge Deller wrote: >>>> This reverts commit 39aead8373b3c20bb5965c024dfb51a94e526151. >>>> >>>> Revert this patch. This patch started to introduce the regression that >>>> all hardware acceleration of more than 35 existing fbdev drivers were >>>> bypassed and thus fbcon console output for those was dramatically slowed >>>> down by factor of 10 and more. >>>> >>>> Reverting this commit has no impact on DRM, since none of the DRM drivers are >>>> tagged with the acceleration flags FBINFO_HWACCEL_COPYAREA, >>>> FBINFO_HWACCEL_FILLRECT or others. >>>> >>>> Signed-off-by: Helge Deller <deller@gmx.de> > >>> As for "why", I think there was a number of private bugs that were >>> reported in this code, which is why it was removed. I do not think it >>> can be safely added back in without addressing them first. Let me go >>> dig through my email to see if I can find them... >> >> Ah, no, that was just the soft scrollback code I was thinking of, which Right. That was commit 973c096f6a85 and it was about vgacon, not fbcon. I did mentioned it in my cover letter, together with my analysis of the reported bugs. Maybe I should have put all the information from the cover letter into the patch here as well. If you haven't read the cover letter yet, please do. Helge > So the bugs argument is moot. > >> was a different revert and is still gone, thankfully :) > > FTR, not everyone else was thankful about that one... > >> This one was just removed because Daniel noticed that only 3 drivers >> used this (nouveau, omapdrm, and gma600), so this shouldn't have caused >> any regressions in any other drivers like you are reporting here. >> >> So perhaps this regression is caused by something else? > > 1. Daniel's patch was not CCed to linux-fbdev, > 2. When I discovered the patch, I pointed out that the premise of 3 > drivers was not true, and that it affects 32 more fbdev drivers[1] . > The patch was applied regardless. > 3. When the patch was suggested for backporting, I pointed out the > same[2]. > The patch was backported regardless. > > [1] https://lore.kernel.org/r/alpine.DEB.2.22.394.2010311116530.379363@ramsan.of.borg/ > [2] https://lore.kernel.org/r/CAMuHMdXRgam2zahPEGcw8+76Xm-0AO-Ci9-YmVa5JpTKVHphRw@mail.gmail.com/ > > Gr{oetje,eeting}s, > > Geert
Hi Greg, Greg Kroah-Hartman <gregkh@linuxfoundation.org> writes: > On Wed, Jan 19, 2022 at 12:22:55PM +0100, Greg Kroah-Hartman wrote: >> On Wed, Jan 19, 2022 at 12:08:39PM +0100, Helge Deller wrote: >> > This reverts commit 39aead8373b3c20bb5965c024dfb51a94e526151. >> > >> > Revert this patch. This patch started to introduce the regression that >> > all hardware acceleration of more than 35 existing fbdev drivers were >> > bypassed and thus fbcon console output for those was dramatically slowed >> > down by factor of 10 and more. >> > >> > Reverting this commit has no impact on DRM, since none of the DRM drivers are >> > tagged with the acceleration flags FBINFO_HWACCEL_COPYAREA, >> > FBINFO_HWACCEL_FILLRECT or others. >> > >> > Signed-off-by: Helge Deller <deller@gmx.de> >> > Cc: stable@vger.kernel.org # v5.16 >> >> Why just 5.16? This commit came in on 5.11 and was backported to >> 5.10.5. >> >> As for "why", I think there was a number of private bugs that were >> reported in this code, which is why it was removed. I do not think it >> can be safely added back in without addressing them first. Let me go >> dig through my email to see if I can find them... > > Ah, no, that was just the soft scrollback code I was thinking of, which > was a different revert and is still gone, thankfully :) > > This one was just removed because Daniel noticed that only 3 drivers > used this (nouveau, omapdrm, and gma600), so this shouldn't have caused > any regressions in any other drivers like you are reporting here. I'm counting more than 3 drivers using this. I think one of the reasons why it was reverted was that no one is actively maintaining fbdev. With Helge now volunteering i don't see a reason why it should stay reverted. If there are issues coming up i'm pretty sure Helge would care, and i would probably also take a look. /Sven
On Wed, Jan 19, 2022 at 02:01:44PM +0100, Sven Schnelle wrote: > Hi Greg, > > Greg Kroah-Hartman <gregkh@linuxfoundation.org> writes: > > > On Wed, Jan 19, 2022 at 12:22:55PM +0100, Greg Kroah-Hartman wrote: > >> On Wed, Jan 19, 2022 at 12:08:39PM +0100, Helge Deller wrote: > >> > This reverts commit 39aead8373b3c20bb5965c024dfb51a94e526151. > >> > > >> > Revert this patch. This patch started to introduce the regression that > >> > all hardware acceleration of more than 35 existing fbdev drivers were > >> > bypassed and thus fbcon console output for those was dramatically slowed > >> > down by factor of 10 and more. > >> > > >> > Reverting this commit has no impact on DRM, since none of the DRM drivers are > >> > tagged with the acceleration flags FBINFO_HWACCEL_COPYAREA, > >> > FBINFO_HWACCEL_FILLRECT or others. > >> > > >> > Signed-off-by: Helge Deller <deller@gmx.de> > >> > Cc: stable@vger.kernel.org # v5.16 > >> > >> Why just 5.16? This commit came in on 5.11 and was backported to > >> 5.10.5. > >> > >> As for "why", I think there was a number of private bugs that were > >> reported in this code, which is why it was removed. I do not think it > >> can be safely added back in without addressing them first. Let me go > >> dig through my email to see if I can find them... > > > > Ah, no, that was just the soft scrollback code I was thinking of, which > > was a different revert and is still gone, thankfully :) > > > > This one was just removed because Daniel noticed that only 3 drivers > > used this (nouveau, omapdrm, and gma600), so this shouldn't have caused > > any regressions in any other drivers like you are reporting here. > > I'm counting more than 3 drivers using this. I think one of the reasons > why it was reverted was that no one is actively maintaining fbdev. With > Helge now volunteering i don't see a reason why it should stay reverted. > If there are issues coming up i'm pretty sure Helge would care, and i > would probably also take a look. Ok, no objection from me, but I think Daniel should weigh in as it is his commit that is being reverted here. thanks, greg k-h
On Wed, Jan 19, 2022 at 2:29 PM Helge Deller <deller@gmx.de> wrote: > > >> > >> Ah, no, that was just the soft scrollback code I was thinking of, which > > Right. > That was commit 973c096f6a85 and it was about vgacon, not fbcon. No, fbcon had some bug too, although I've paged out the details. See commit 50145474f6ef ("fbcon: remove soft scrollback code"). If I remember correctly (and it's entirely possible that I don't), the whole "softback_lines" logic had serious problems with resizing the console (or maybe changing the font size). There may have been some other bad interaction with foreground/background consoles too, I forget. Linus
On 1/19/22 15:01, Linus Torvalds wrote: > On Wed, Jan 19, 2022 at 2:29 PM Helge Deller <deller@gmx.de> wrote: >> >>>> >>>> Ah, no, that was just the soft scrollback code I was thinking of, which >> >> Right. >> That was commit 973c096f6a85 and it was about vgacon, not fbcon. > > No, fbcon had some bug too, although I've paged out the details. See > commit 50145474f6ef ("fbcon: remove soft scrollback code"). Yes, but those bugs didn't affected > If I remember correctly (and it's entirely possible that I don't), the > whole "softback_lines" logic had serious problems with resizing the > console (or maybe changing the font size). Right. I'm not asking to revert any soft scrollback functions. It's about if you allow to use the fbdev-drivers hardware acceleration to move parts of the screen or if you are stuck to software memcpy() and repainting instead. None of the bug reports touched that part. > There may have been some other bad interaction with > foreground/background consoles too, I forget. I think that is independend if you use hardware acceleration or not. Helge
On Wed, Jan 19, 2022 at 3:01 PM Linus Torvalds <torvalds@linux-foundation.org> wrote: > > On Wed, Jan 19, 2022 at 2:29 PM Helge Deller <deller@gmx.de> wrote: > > > > >> > > >> Ah, no, that was just the soft scrollback code I was thinking of, which > > > > Right. > > That was commit 973c096f6a85 and it was about vgacon, not fbcon. > > No, fbcon had some bug too, although I've paged out the details. See > commit 50145474f6ef ("fbcon: remove soft scrollback code"). tbh I've paged it all out too. > If I remember correctly (and it's entirely possible that I don't), the > whole "softback_lines" logic had serious problems with resizing the > console (or maybe changing the font size). Yeah that pile of reverts was my motiviation to look into this and see what else we could rip out most likely and still have an fbcon that works as well as it does right now for almost all users (which is not so great, but oh well). > There may have been some other bad interaction with > foreground/background consoles too, I forget. Irrespective of this code being buggy or not buggy I think the bigger pictures, and really the reason I want to see as much code ditched from the fbdev/fbcon stack as we possible can, are very clear: - it's full of bugs - there's no test coverage/CI to speak of - it's very arcane code which is damn hard to understand and fix issues within - the locking is busted (largely thanks to console_lock, and the effort to make that reasonable from -rt folks has been slowly creeping forward for years). Iow this subsystem is firmly stuck in the 90s, and I think it's best to just leave it there. There's also not been anyone actually capable and willing to put in the work to change this (pretty much all actual changes/fixes have been done by drm folks anyway, like me having a small pet project to make the fbdev vs fbcon locking slightly less busted). The other side is that being a maintainer is about collaboration, and this entire fbdev maintainership takeover has been a demonstration of anything but that. MAINTAINERS entry was a bit confusing since defacto drm has been maintaining it for years, but for the above reasons we've done that by just aggressively deleting stuff that isn't absolutely needed - hence why I figured "orphaned" is a reasonable description of the state of things. This entire affair of rushing in a maintainer change over the w/e and then being greeted by a lot of wtf mails next Monday does leave a rather sour aftertaste. Plus that thread shows a lot of misunderstandings of what's all been going on and what drm can and cannot do by Helge, which doesn't improve the entire "we need fbdev back" argument. But if the overall consensus is that that fbdev needs to be brought back to it's full 90s glory then I think we need a copy of that code for drm drivers (should work out if we intercept fb_open() and put our own file_ops in there, maybe some more fun with fbcon), so that at least for anything modern using drm driver we can keep on maintaining that compat support code. And with maintaining here I don't mean build a museum around it, but actually try to keep/move the thing towards a state where we can still tell distros that enabling it is an ok thing to do and not just a CVE subscription (well it is that too right now, but at least we can fix a lot of them by just deleting code). I think until that mess is sorted out resurrecting code that's not strictly needed is just not a bright idea. Also wrt the issue at hand of "fbcon scrolling": The way to actually do that with some speed is to render into a fully cached shadow buffer and upload changed areas with a timer. Not with hw accelerated scrolling, at least not if we just don't have full scale development teams for each driver because creating 2d accel that doesn't suck is really hard. drm fbdev compat helpers give you that shadow buffer for free (well you got to set some options). Unfortunately just ditching fbdev/fbcon compat is not an option for many distros still, althought things are very slowly moving towards that. Until we've arrived there I can't just pretend to not care about what's going on in drivers/video. -Daniel
Daniel Vetter <daniel@ffwll.ch> writes: > On Wed, Jan 19, 2022 at 3:01 PM Linus Torvalds > <torvalds@linux-foundation.org> wrote: > Irrespective of this code being buggy or not buggy I think the bigger > pictures, and really the reason I want to see as much code ditched > from the fbdev/fbcon stack as we possible can, are very clear: > > - it's full of bugs > - there's no test coverage/CI to speak of > - it's very arcane code which is damn hard to understand and fix issues within > - the locking is busted (largely thanks to console_lock, and the > effort to make that reasonable from -rt folks has been slowly creeping > forward for years). > > Iow this subsystem is firmly stuck in the 90s, and I think it's best > to just leave it there. There's also not been anyone actually capable > and willing to put in the work to change this (pretty much all actual > changes/fixes have been done by drm folks anyway, like me having a > small pet project to make the fbdev vs fbcon locking slightly less > busted). Saying it's stuck in the 90ies, and actively trying to prevent Helge from taking over maintainership at the same time looks odd. I think Helge should at least get a chance to fix the issues. If the state is still the same in a year or so it should be discussed again. > The other side is that being a maintainer is about collaboration, and > this entire fbdev maintainership takeover has been a demonstration of > anything but that. MAINTAINERS entry was a bit confusing since defacto > drm has been maintaining it for years. It was marked as 'Orphaned'. Anyone is free to send a Patch/PR to take over maintainership. If you have strong opinions about that code (And you obviously have reading your mail, set it to 'maintained' and care about it. Everything else is just wrong in my opinion. /Sven
On 1/19/22 15:34, Daniel Vetter wrote: > On Wed, Jan 19, 2022 at 3:01 PM Linus Torvalds > <torvalds@linux-foundation.org> wrote: >> >> On Wed, Jan 19, 2022 at 2:29 PM Helge Deller <deller@gmx.de> wrote: >>> >>>>> >>>>> Ah, no, that was just the soft scrollback code I was thinking of, which >>> >>> Right. >>> That was commit 973c096f6a85 and it was about vgacon, not fbcon. >> >> No, fbcon had some bug too, although I've paged out the details. See >> commit 50145474f6ef ("fbcon: remove soft scrollback code"). > > tbh I've paged it all out too. > >> If I remember correctly (and it's entirely possible that I don't), the >> whole "softback_lines" logic had serious problems with resizing the >> console (or maybe changing the font size). > > Yeah that pile of reverts was my motiviation to look into this and see > what else we could rip out most likely and still have an fbcon that > works as well as it does right now for almost all users (which is not > so great, but oh well). > >> There may have been some other bad interaction with >> foreground/background consoles too, I forget. > > Irrespective of this code being buggy or not buggy I think the bigger > pictures, and really the reason I want to see as much code ditched > from the fbdev/fbcon stack as we possible can, are very clear: > > - it's full of bugs I'm sure that there are bugs. Not just in fbcon/fbdev. Other than that, if there are bugs I'm sure they are independend of the question if you use hardware acceleration or not. > - there's no test coverage/CI to speak of > - it's very arcane code which is damn hard to understand and fix issues within > - the locking is busted (largely thanks to console_lock, and the > effort to make that reasonable from -rt folks has been slowly creeping > forward for years). > > Iow this subsystem is firmly stuck in the 90s, and I think it's best > to just leave it there. There's also not been anyone actually capable > and willing to put in the work to change this (pretty much all actual > changes/fixes have been done by drm folks anyway, like me having a > small pet project to make the fbdev vs fbcon locking slightly less > busted). Yes, drm folks fixed a lot of bugs in the generic fbcon code. I think everyone is thankful for this. > The other side is that being a maintainer is about collaboration, and > this entire fbdev maintainership takeover has been a demonstration of > anything but that. MAINTAINERS entry was a bit confusing since defacto > drm has been maintaining it for years, but for the above reasons we've > done that by just aggressively deleting stuff that isn't absolutely > needed - hence why I figured "orphaned" is a reasonable description of > the state of things. This entire affair of rushing in a maintainer > change over the w/e and then being greeted by a lot of wtf mails next > Monday does leave a rather sour aftertaste. Plus that thread shows a > lot of misunderstandings of what's all been going on and what drm can > and cannot do by Helge, which doesn't improve the entire "we need > fbdev back" argument. I'm happy to *really* maintain fbdev code & drivers. Up to now only those parts which were still needed by drm (like fbcon) were fixed & "maintained" by drm folks. Nearly all other patches sent to the fbdev list were ignored and even new submissions for fbdev drivers were denied. Now in next step really important infrastructure for fbdev-drivers was ripped out of fbcon, like suddenly denying fbdev-drivers to use hardware acceleration. According to the docs the next step would have been to drop even more code from the fbdev drivers. This is not what "maintain" really is about. > But if the overall consensus is that that fbdev needs to be brought > back to it's full 90s glory then I think we need a copy of that code > for drm drivers (should work out if we intercept fb_open() and put our > own file_ops in there, maybe some more fun with fbcon), so that at > least for anything modern using drm driver we can keep on maintaining > that compat support code. It's not about to keep something alive or to stop future developments. It's about fairness and not actively breaking other parts of the kernel for no good reason. > And with maintaining here I don't mean build a museum around it, but > actually try to keep/move the thing towards a state where we can still > tell distros that enabling it is an ok thing to do and not just a CVE > subscription (well it is that too right now, but at least we can fix a > lot of them by just deleting code). > > I think until that mess is sorted out resurrecting code that's not > strictly needed is just not a bright idea. That's wrong. It's strictly needed by more than 35 fbdev drivers and as such you introduced a regression for those. > Also wrt the issue at hand of "fbcon scrolling": The way to actually > do that with some speed is to render into a fully cached shadow buffer > and upload changed areas with a timer. Not with hw accelerated > scrolling, at least not if we just don't have full scale development > teams for each driver because creating 2d accel that doesn't suck is > really hard. drm fbdev compat helpers give you that shadow buffer for > free (well you got to set some options). > > Unfortunately just ditching fbdev/fbcon compat is not an option for > many distros still, althought things are very slowly moving towards > that. Until we've arrived there I can't just pretend to not care about > what's going on in drivers/video. I'm happy to take care about it. That's why I stepped up as maintainer. Helge
Hi Am 19.01.22 um 16:05 schrieb Sven Schnelle: > Daniel Vetter <daniel@ffwll.ch> writes: > >> On Wed, Jan 19, 2022 at 3:01 PM Linus Torvalds >> <torvalds@linux-foundation.org> wrote: >> Irrespective of this code being buggy or not buggy I think the bigger >> pictures, and really the reason I want to see as much code ditched >> from the fbdev/fbcon stack as we possible can, are very clear: >> >> - it's full of bugs >> - there's no test coverage/CI to speak of >> - it's very arcane code which is damn hard to understand and fix issues within >> - the locking is busted (largely thanks to console_lock, and the >> effort to make that reasonable from -rt folks has been slowly creeping >> forward for years). >> >> Iow this subsystem is firmly stuck in the 90s, and I think it's best >> to just leave it there. There's also not been anyone actually capable >> and willing to put in the work to change this (pretty much all actual >> changes/fixes have been done by drm folks anyway, like me having a >> small pet project to make the fbdev vs fbcon locking slightly less >> busted). > > Saying it's stuck in the 90ies, and actively trying to prevent > Helge from taking over maintainership at the same time looks odd. The issues are in the design itself. It's impossible to model today's hardware and constraints with fbdev. It's impossible to change configuration in a reliable way (i.e., what DRM calls atomic). Fbdev mmaps plain video ram to userspace, which is one of the reasons why DRM's fbdev support is hard to improve. > I think Helge should at least get a chance to fix the issues. If the > state is still the same in a year or so it should be discussed again. You cannot fix that in 10yrs. > >> The other side is that being a maintainer is about collaboration, and >> this entire fbdev maintainership takeover has been a demonstration of >> anything but that. MAINTAINERS entry was a bit confusing since defacto >> drm has been maintaining it for years. > > It was marked as 'Orphaned'. Anyone is free to send a Patch/PR to take > over maintainership. If you have strong opinions about that code (And you > obviously have reading your mail, set it to 'maintained' and care about > it. Everything else is just wrong in my opinion. No, it's not wrong. Helge takes fbdev over the weekend, without noteworthy experience, and ignores advice from the people that have kept it alive over the past years. This isn't going to work in the long term. Best regards Thomas > > /Sven
On Wed, Jan 19, 2022 at 4:06 PM Sven Schnelle <svens@stackframe.org> wrote: > > Daniel Vetter <daniel@ffwll.ch> writes: > > > On Wed, Jan 19, 2022 at 3:01 PM Linus Torvalds > > <torvalds@linux-foundation.org> wrote: > > Irrespective of this code being buggy or not buggy I think the bigger > > pictures, and really the reason I want to see as much code ditched > > from the fbdev/fbcon stack as we possible can, are very clear: > > > > - it's full of bugs > > - there's no test coverage/CI to speak of > > - it's very arcane code which is damn hard to understand and fix issues within > > - the locking is busted (largely thanks to console_lock, and the > > effort to make that reasonable from -rt folks has been slowly creeping > > forward for years). > > > > Iow this subsystem is firmly stuck in the 90s, and I think it's best > > to just leave it there. There's also not been anyone actually capable > > and willing to put in the work to change this (pretty much all actual > > changes/fixes have been done by drm folks anyway, like me having a > > small pet project to make the fbdev vs fbcon locking slightly less > > busted). > > Saying it's stuck in the 90ies, and actively trying to prevent > Helge from taking over maintainership at the same time looks odd. > I think Helge should at least get a chance to fix the issues. If the > state is still the same in a year or so it should be discussed again. You don't need maintainership to fix issues. You need to submit patches. If otoh you get the maintainership first to be able to cram in reverts without discussions, then it's very backwards. > > The other side is that being a maintainer is about collaboration, and > > this entire fbdev maintainership takeover has been a demonstration of > > anything but that. MAINTAINERS entry was a bit confusing since defacto > > drm has been maintaining it for years. > > It was marked as 'Orphaned'. Anyone is free to send a Patch/PR to take > over maintainership. If you have strong opinions about that code (And you > obviously have reading your mail, set it to 'maintained' and care about > it. Everything else is just wrong in my opinion. I already added dri-devel so anything we drastically change can be discussed first. If that's indeed not strong enough then yes I can whack in full maintainer entry with a bugfix-only status. But really I try to not create facts with just editing MAINTAINERS first and ask questions later, that's just not a great way to collaborate. -Daniel
On 1/19/22 16:42, Daniel Vetter wrote: > On Wed, Jan 19, 2022 at 4:06 PM Sven Schnelle <svens@stackframe.org> wrote: >> >> Daniel Vetter <daniel@ffwll.ch> writes: >> >>> On Wed, Jan 19, 2022 at 3:01 PM Linus Torvalds >>> <torvalds@linux-foundation.org> wrote: >>> Irrespective of this code being buggy or not buggy I think the bigger >>> pictures, and really the reason I want to see as much code ditched >>> from the fbdev/fbcon stack as we possible can, are very clear: >>> >>> - it's full of bugs >>> - there's no test coverage/CI to speak of >>> - it's very arcane code which is damn hard to understand and fix issues within >>> - the locking is busted (largely thanks to console_lock, and the >>> effort to make that reasonable from -rt folks has been slowly creeping >>> forward for years). >>> >>> Iow this subsystem is firmly stuck in the 90s, and I think it's best >>> to just leave it there. There's also not been anyone actually capable >>> and willing to put in the work to change this (pretty much all actual >>> changes/fixes have been done by drm folks anyway, like me having a >>> small pet project to make the fbdev vs fbcon locking slightly less >>> busted). >> >> Saying it's stuck in the 90ies, and actively trying to prevent >> Helge from taking over maintainership at the same time looks odd. >> I think Helge should at least get a chance to fix the issues. If the >> state is still the same in a year or so it should be discussed again. > > You don't need maintainership to fix issues. You need to submit patches. The very first email of this thread is my patch. And you just added your comments to this patch. > If otoh you get the maintainership first to be able to cram in reverts > without discussions, then it's very backwards. I'm working on the Linux kernel since at least 23 years and am a maintainer of parts of it. I know that and would never push something which is controversal without discussions. >>> The other side is that being a maintainer is about collaboration, and >>> this entire fbdev maintainership takeover has been a demonstration of >>> anything but that. MAINTAINERS entry was a bit confusing since defacto >>> drm has been maintaining it for years. >> >> It was marked as 'Orphaned'. Anyone is free to send a Patch/PR to take >> over maintainership. If you have strong opinions about that code (And you >> obviously have reading your mail, set it to 'maintained' and care about >> it. Everything else is just wrong in my opinion. > > I already added dri-devel so anything we drastically change can be > discussed first. If that's indeed not strong enough then yes I can > whack in full maintainer entry with a bugfix-only status. > > But really I try to not create facts with just editing MAINTAINERS > first and ask questions later, that's just not a great way to > collaborate. Helge
On Wed, 19 Jan 2022, Helge Deller <deller@gmx.de> wrote: > On 1/19/22 16:42, Daniel Vetter wrote: >> If otoh you get the maintainership first to be able to cram in reverts >> without discussions, then it's very backwards. > I'm working on the Linux kernel since at least 23 years and am a maintainer of parts of it. > I know that and would never push something which is controversal without discussions. I think the entire MAINTAINERS change was controversial and rushed to Linus without discussion over a weekend. BR, Jani.
On Wed, Jan 19, 2022 at 06:13:39PM +0200, Jani Nikula wrote: > On Wed, 19 Jan 2022, Helge Deller <deller@gmx.de> wrote: > > On 1/19/22 16:42, Daniel Vetter wrote: > >> If otoh you get the maintainership first to be able to cram in reverts > >> without discussions, then it's very backwards. > > I'm working on the Linux kernel since at least 23 years and am a maintainer of parts of it. > > I know that and would never push something which is controversal without discussions. > > I think the entire MAINTAINERS change was controversial and rushed to > Linus without discussion over a weekend. Yeah just looking at the size of the thread is pretty clear indiciation that this went wrong real good. And I said that you need to clean this up in my very first reply. None of this should have been a surprise, but somehow it all happened. -Daniel
Hi Thomas, On 1/19/22 16:37, Thomas Zimmermann wrote: > Am 19.01.22 um 16:05 schrieb Sven Schnelle: >> Daniel Vetter <daniel@ffwll.ch> writes: >> >>> On Wed, Jan 19, 2022 at 3:01 PM Linus Torvalds >>> <torvalds@linux-foundation.org> wrote: >>> Irrespective of this code being buggy or not buggy I think the bigger >>> pictures, and really the reason I want to see as much code ditched >>> from the fbdev/fbcon stack as we possible can, are very clear: >>> >>> - it's full of bugs >>> - there's no test coverage/CI to speak of >>> - it's very arcane code which is damn hard to understand and fix issues within >>> - the locking is busted (largely thanks to console_lock, and the >>> effort to make that reasonable from -rt folks has been slowly creeping >>> forward for years). >>> >>> Iow this subsystem is firmly stuck in the 90s, and I think it's best >>> to just leave it there. There's also not been anyone actually capable >>> and willing to put in the work to change this (pretty much all actual >>> changes/fixes have been done by drm folks anyway, like me having a >>> small pet project to make the fbdev vs fbcon locking slightly less >>> busted). >> >> Saying it's stuck in the 90ies, and actively trying to prevent >> Helge from taking over maintainership at the same time looks odd. > > The issues are in the design itself. It's impossible to model today's > hardware and constraints with fbdev. It's impossible to change > configuration in a reliable way (i.e., what DRM calls atomic). Fbdev > mmaps plain video ram to userspace, which is one of the reasons why > DRM's fbdev support is hard to improve. That's fully understood, but I think you are mixing up things here... The fbdev userspace api is most likely not the best way forward. I'm sure that drm can and will provide better solutions for userspace. And userspace will surely pick up those new interfaces. DRM folks will drive it in the right direction, I'm sure! But in addition fbdev/fbcon is the kernel framework for nearly all existing graphic cards which are not (yet) supported by DRM. They need fbdev/fbcon to show their text console and maybe a simple X server. If you break fbdev for those cards, they are completely stuck. Hopefully those drivers will be ported to DRM, but that's currently not easily possible (or they would be so slow that they are unuseable). So, I don't think you should try to improve DRM's /dev/fb0 support further, but instead work forward for a new interface which perfectly suits DRM. That's Ok, and my goal is not to prevent that. >> I think Helge should at least get a chance to fix the issues. If the >> state is still the same in a year or so it should be discussed again. > > You cannot fix that in 10yrs. > >> >>> The other side is that being a maintainer is about collaboration, and >>> this entire fbdev maintainership takeover has been a demonstration of >>> anything but that. MAINTAINERS entry was a bit confusing since defacto >>> drm has been maintaining it for years. >> >> It was marked as 'Orphaned'. Anyone is free to send a Patch/PR to take >> over maintainership. If you have strong opinions about that code (And you >> obviously have reading your mail, set it to 'maintained' and care about >> it. Everything else is just wrong in my opinion. > > No, it's not wrong. Helge takes fbdev over the weekend, without noteworthy experience, and ignores advice from the people that have kept it alive over the past years. This isn't going to work in the long term. > > Best regards > Thomas > >> >> /Sven >
On Wed, Jan 19, 2022 at 12:08:39PM +0100, Helge Deller wrote: > This reverts commit 39aead8373b3c20bb5965c024dfb51a94e526151. > > Revert this patch. This patch started to introduce the regression that > all hardware acceleration of more than 35 existing fbdev drivers were > bypassed and thus fbcon console output for those was dramatically slowed > down by factor of 10 and more. > > Reverting this commit has no impact on DRM, since none of the DRM drivers are > tagged with the acceleration flags FBINFO_HWACCEL_COPYAREA, > FBINFO_HWACCEL_FILLRECT or others. > > Signed-off-by: Helge Deller <deller@gmx.de> > Cc: stable@vger.kernel.org # v5.16 So if this really has to come back then I think the pragmatic approach is to do it behind a CONFIG_FBCON_ACCEL, default n, and with a huge warning that enabling that shouldn't be done for any distro which only enables firmware and drm fbdev drivers. Plus adjusting the todo to limit it to drm drivers. Maybe also #ifdef out the code that's then dead from fbcon. Also in that case I guess it's ok to cc: stable, and really if you cc: stable it needs to go down to 5.11, not 5.16. And if we do that, I think that should go in through a -next cycle, or at least quite some soaking before it's cherry-picked over. Enough to give syzbot a chance to discover any path we've missed at least. -Daniel > --- > Documentation/gpu/todo.rst | 21 --------------- > drivers/video/fbdev/core/fbcon.c | 45 ++++++++++++++++++++++++++------ > 2 files changed, 37 insertions(+), 29 deletions(-) > > diff --git a/Documentation/gpu/todo.rst b/Documentation/gpu/todo.rst > index 29506815d24a..a1212b5b3026 100644 > --- a/Documentation/gpu/todo.rst > +++ b/Documentation/gpu/todo.rst > @@ -300,27 +300,6 @@ Contact: Daniel Vetter, Noralf Tronnes > > Level: Advanced > > -Garbage collect fbdev scrolling acceleration > --------------------------------------------- > - > -Scroll acceleration is disabled in fbcon by hard-wiring p->scrollmode = > -SCROLL_REDRAW. There's a ton of code this will allow us to remove: > - > -- lots of code in fbcon.c > - > -- a bunch of the hooks in fbcon_ops, maybe the remaining hooks could be called > - directly instead of the function table (with a switch on p->rotate) > - > -- fb_copyarea is unused after this, and can be deleted from all drivers > - > -Note that not all acceleration code can be deleted, since clearing and cursor > -support is still accelerated, which might be good candidates for further > -deletion projects. > - > -Contact: Daniel Vetter > - > -Level: Intermediate > - > idr_init_base() > --------------- > > diff --git a/drivers/video/fbdev/core/fbcon.c b/drivers/video/fbdev/core/fbcon.c > index 22bb3892f6bd..b813985f1403 100644 > --- a/drivers/video/fbdev/core/fbcon.c > +++ b/drivers/video/fbdev/core/fbcon.c > @@ -1025,7 +1025,7 @@ static void fbcon_init(struct vc_data *vc, int init) > struct vc_data *svc = *default_mode; > struct fbcon_display *t, *p = &fb_display[vc->vc_num]; > int logo = 1, new_rows, new_cols, rows, cols; > - int ret; > + int cap, ret; > > if (WARN_ON(info_idx == -1)) > return; > @@ -1034,6 +1034,7 @@ static void fbcon_init(struct vc_data *vc, int init) > con2fb_map[vc->vc_num] = info_idx; > > info = registered_fb[con2fb_map[vc->vc_num]]; > + cap = info->flags; > > if (logo_shown < 0 && console_loglevel <= CONSOLE_LOGLEVEL_QUIET) > logo_shown = FBCON_LOGO_DONTSHOW; > @@ -1135,13 +1136,11 @@ static void fbcon_init(struct vc_data *vc, int init) > > ops->graphics = 0; > > - /* > - * No more hw acceleration for fbcon. > - * > - * FIXME: Garbage collect all the now dead code after sufficient time > - * has passed. > - */ > - p->scrollmode = SCROLL_REDRAW; > + if ((cap & FBINFO_HWACCEL_COPYAREA) && > + !(cap & FBINFO_HWACCEL_DISABLED)) > + p->scrollmode = SCROLL_MOVE; > + else /* default to something safe */ > + p->scrollmode = SCROLL_REDRAW; > > /* > * ++guenther: console.c:vc_allocate() relies on initializing > @@ -1953,15 +1952,45 @@ static void updatescrollmode(struct fbcon_display *p, > { > struct fbcon_ops *ops = info->fbcon_par; > int fh = vc->vc_font.height; > + int cap = info->flags; > + u16 t = 0; > + int ypan = FBCON_SWAP(ops->rotate, info->fix.ypanstep, > + info->fix.xpanstep); > + int ywrap = FBCON_SWAP(ops->rotate, info->fix.ywrapstep, t); > int yres = FBCON_SWAP(ops->rotate, info->var.yres, info->var.xres); > int vyres = FBCON_SWAP(ops->rotate, info->var.yres_virtual, > info->var.xres_virtual); > + int good_pan = (cap & FBINFO_HWACCEL_YPAN) && > + divides(ypan, vc->vc_font.height) && vyres > yres; > + int good_wrap = (cap & FBINFO_HWACCEL_YWRAP) && > + divides(ywrap, vc->vc_font.height) && > + divides(vc->vc_font.height, vyres) && > + divides(vc->vc_font.height, yres); > + int reading_fast = cap & FBINFO_READS_FAST; > + int fast_copyarea = (cap & FBINFO_HWACCEL_COPYAREA) && > + !(cap & FBINFO_HWACCEL_DISABLED); > + int fast_imageblit = (cap & FBINFO_HWACCEL_IMAGEBLIT) && > + !(cap & FBINFO_HWACCEL_DISABLED); > > p->vrows = vyres/fh; > if (yres > (fh * (vc->vc_rows + 1))) > p->vrows -= (yres - (fh * vc->vc_rows)) / fh; > if ((yres % fh) && (vyres % fh < yres % fh)) > p->vrows--; > + > + if (good_wrap || good_pan) { > + if (reading_fast || fast_copyarea) > + p->scrollmode = good_wrap ? > + SCROLL_WRAP_MOVE : SCROLL_PAN_MOVE; > + else > + p->scrollmode = good_wrap ? SCROLL_REDRAW : > + SCROLL_PAN_REDRAW; > + } else { > + if (reading_fast || (fast_copyarea && !fast_imageblit)) > + p->scrollmode = SCROLL_MOVE; > + else > + p->scrollmode = SCROLL_REDRAW; > + } > } > > #define PITCH(w) (((w) + 7) >> 3) > -- > 2.31.1 >
Hello Daniel, On 1/20/22 15:30, Daniel Vetter wrote: > On Wed, Jan 19, 2022 at 12:08:39PM +0100, Helge Deller wrote: >> This reverts commit 39aead8373b3c20bb5965c024dfb51a94e526151. >> >> Revert this patch. This patch started to introduce the regression that >> all hardware acceleration of more than 35 existing fbdev drivers were >> bypassed and thus fbcon console output for those was dramatically slowed >> down by factor of 10 and more. >> >> Reverting this commit has no impact on DRM, since none of the DRM drivers are >> tagged with the acceleration flags FBINFO_HWACCEL_COPYAREA, >> FBINFO_HWACCEL_FILLRECT or others. >> >> Signed-off-by: Helge Deller <deller@gmx.de> >> Cc: stable@vger.kernel.org # v5.16 > > So if this really has to come back then I think the pragmatic approach is > to do it behind a CONFIG_FBCON_ACCEL, default n, and with a huge warning > that enabling that shouldn't be done for any distro which only enables > firmware and drm fbdev drivers. Thanks for coming back on this, but quite frankly I don't understand that request. How should that warning look like, something along: "BE WARNED: The framebuffer text console on your non-DRM supported graphic card will then run faster and smoother if you enable this option." That doesn't make sense. People and distros would want to enable that. And if a distro *just* has firmware and drm fbdev drivers enabled, none of the non-DRM graphic cards would be loaded anyway and this code wouldn't be executed anyway. I think what you want is that DRM drivers are preferred over standard fbdev drivers, esp. if there is a driver for both available. But that's completely independed of fbdev-drivers console hardware acceleration. > Plus adjusting the todo to limit it to drm drivers. Maybe also #ifdef out > the code that's then dead from fbcon. Sorry, I don't understand that either. I assume you mean to put code of fbcon which is only used by fbdev-drivers into and #ifdef CONFIG_FB .. #endif (CONFIG_FB may be wrong in this example). That's probably possible, but I don't see a big win. If there is no fbdev driver compiled-in or as module, none of this fbdev-drivers will be loaded and that code path wouldn't be executed anyway. In that case you will win a few bytes of code, but probably not much. > Also in that case I guess it's ok to cc: stable, and really if you cc: > stable it needs to go down to 5.11, not 5.16. Yes, I missed that in my patch request. Will fix. > And if we do that, I think that should go in through a -next cycle, or at > least quite some soaking before it's cherry-picked over. Enough to give > syzbot a chance to discover any path we've missed at least. Sure. We don't need to hurry. Thanks! Helge > -Daniel > >> --- >> Documentation/gpu/todo.rst | 21 --------------- >> drivers/video/fbdev/core/fbcon.c | 45 ++++++++++++++++++++++++++------ >> 2 files changed, 37 insertions(+), 29 deletions(-) >> >> diff --git a/Documentation/gpu/todo.rst b/Documentation/gpu/todo.rst >> index 29506815d24a..a1212b5b3026 100644 >> --- a/Documentation/gpu/todo.rst >> +++ b/Documentation/gpu/todo.rst >> @@ -300,27 +300,6 @@ Contact: Daniel Vetter, Noralf Tronnes >> >> Level: Advanced >> >> -Garbage collect fbdev scrolling acceleration >> --------------------------------------------- >> - >> -Scroll acceleration is disabled in fbcon by hard-wiring p->scrollmode = >> -SCROLL_REDRAW. There's a ton of code this will allow us to remove: >> - >> -- lots of code in fbcon.c >> - >> -- a bunch of the hooks in fbcon_ops, maybe the remaining hooks could be called >> - directly instead of the function table (with a switch on p->rotate) >> - >> -- fb_copyarea is unused after this, and can be deleted from all drivers >> - >> -Note that not all acceleration code can be deleted, since clearing and cursor >> -support is still accelerated, which might be good candidates for further >> -deletion projects. >> - >> -Contact: Daniel Vetter >> - >> -Level: Intermediate >> - >> idr_init_base() >> --------------- >> >> diff --git a/drivers/video/fbdev/core/fbcon.c b/drivers/video/fbdev/core/fbcon.c >> index 22bb3892f6bd..b813985f1403 100644 >> --- a/drivers/video/fbdev/core/fbcon.c >> +++ b/drivers/video/fbdev/core/fbcon.c >> @@ -1025,7 +1025,7 @@ static void fbcon_init(struct vc_data *vc, int init) >> struct vc_data *svc = *default_mode; >> struct fbcon_display *t, *p = &fb_display[vc->vc_num]; >> int logo = 1, new_rows, new_cols, rows, cols; >> - int ret; >> + int cap, ret; >> >> if (WARN_ON(info_idx == -1)) >> return; >> @@ -1034,6 +1034,7 @@ static void fbcon_init(struct vc_data *vc, int init) >> con2fb_map[vc->vc_num] = info_idx; >> >> info = registered_fb[con2fb_map[vc->vc_num]]; >> + cap = info->flags; >> >> if (logo_shown < 0 && console_loglevel <= CONSOLE_LOGLEVEL_QUIET) >> logo_shown = FBCON_LOGO_DONTSHOW; >> @@ -1135,13 +1136,11 @@ static void fbcon_init(struct vc_data *vc, int init) >> >> ops->graphics = 0; >> >> - /* >> - * No more hw acceleration for fbcon. >> - * >> - * FIXME: Garbage collect all the now dead code after sufficient time >> - * has passed. >> - */ >> - p->scrollmode = SCROLL_REDRAW; >> + if ((cap & FBINFO_HWACCEL_COPYAREA) && >> + !(cap & FBINFO_HWACCEL_DISABLED)) >> + p->scrollmode = SCROLL_MOVE; >> + else /* default to something safe */ >> + p->scrollmode = SCROLL_REDRAW; >> >> /* >> * ++guenther: console.c:vc_allocate() relies on initializing >> @@ -1953,15 +1952,45 @@ static void updatescrollmode(struct fbcon_display *p, >> { >> struct fbcon_ops *ops = info->fbcon_par; >> int fh = vc->vc_font.height; >> + int cap = info->flags; >> + u16 t = 0; >> + int ypan = FBCON_SWAP(ops->rotate, info->fix.ypanstep, >> + info->fix.xpanstep); >> + int ywrap = FBCON_SWAP(ops->rotate, info->fix.ywrapstep, t); >> int yres = FBCON_SWAP(ops->rotate, info->var.yres, info->var.xres); >> int vyres = FBCON_SWAP(ops->rotate, info->var.yres_virtual, >> info->var.xres_virtual); >> + int good_pan = (cap & FBINFO_HWACCEL_YPAN) && >> + divides(ypan, vc->vc_font.height) && vyres > yres; >> + int good_wrap = (cap & FBINFO_HWACCEL_YWRAP) && >> + divides(ywrap, vc->vc_font.height) && >> + divides(vc->vc_font.height, vyres) && >> + divides(vc->vc_font.height, yres); >> + int reading_fast = cap & FBINFO_READS_FAST; >> + int fast_copyarea = (cap & FBINFO_HWACCEL_COPYAREA) && >> + !(cap & FBINFO_HWACCEL_DISABLED); >> + int fast_imageblit = (cap & FBINFO_HWACCEL_IMAGEBLIT) && >> + !(cap & FBINFO_HWACCEL_DISABLED); >> >> p->vrows = vyres/fh; >> if (yres > (fh * (vc->vc_rows + 1))) >> p->vrows -= (yres - (fh * vc->vc_rows)) / fh; >> if ((yres % fh) && (vyres % fh < yres % fh)) >> p->vrows--; >> + >> + if (good_wrap || good_pan) { >> + if (reading_fast || fast_copyarea) >> + p->scrollmode = good_wrap ? >> + SCROLL_WRAP_MOVE : SCROLL_PAN_MOVE; >> + else >> + p->scrollmode = good_wrap ? SCROLL_REDRAW : >> + SCROLL_PAN_REDRAW; >> + } else { >> + if (reading_fast || (fast_copyarea && !fast_imageblit)) >> + p->scrollmode = SCROLL_MOVE; >> + else >> + p->scrollmode = SCROLL_REDRAW; >> + } >> } >> >> #define PITCH(w) (((w) + 7) >> 3) >> -- >> 2.31.1 >> >
Hi, > > So if this really has to come back then I think the pragmatic approach is > > to do it behind a CONFIG_FBCON_ACCEL, default n, and with a huge warning > > that enabling that shouldn't be done for any distro which only enables > > firmware and drm fbdev drivers. > > Thanks for coming back on this, but quite frankly I don't understand > that request. How should that warning look like, something along: > "BE WARNED: The framebuffer text console on your non-DRM supported > graphic card will then run faster and smoother if you enable this option." > That doesn't make sense. People and distros would want to enable that. Nope. Most distros want disable fbdev drivers rather sooner than later. The fbdev drivers enabled in the fedora kernel today: CONFIG_FB_VGA16=m CONFIG_FB_VESA=y CONFIG_FB_EFI=y CONFIG_FB_SSD1307=m CONFIG_FB_VESA + CONFIG_FB_EFI will go away soon, with simpledrm taking over their role. > And if a distro *just* has firmware and drm fbdev drivers enabled, > none of the non-DRM graphic cards would be loaded anyway and this code > wouldn't be executed anyway. Yes, exactly. That's why there is no point in compiling that code. take care, Gerd
On 1/21/22 08:20, Gerd Hoffmann wrote: >>> So if this really has to come back then I think the pragmatic approach is >>> to do it behind a CONFIG_FBCON_ACCEL, default n, and with a huge warning >>> that enabling that shouldn't be done for any distro which only enables >>> firmware and drm fbdev drivers. >> >> Thanks for coming back on this, but quite frankly I don't understand >> that request. How should that warning look like, something along: >> "BE WARNED: The framebuffer text console on your non-DRM supported >> graphic card will then run faster and smoother if you enable this option." >> That doesn't make sense. People and distros would want to enable that. > > Nope. Most distros want disable fbdev drivers rather sooner than later. > The fbdev drivers enabled in the fedora kernel today: > > CONFIG_FB_VGA16=m > CONFIG_FB_VESA=y > CONFIG_FB_EFI=y > CONFIG_FB_SSD1307=m > > CONFIG_FB_VESA + CONFIG_FB_EFI will go away soon, with simpledrm taking > over their role. That's Ok. Nevertheless, some distros and platforms will still need fbdev drivers for various reasons. >> And if a distro *just* has firmware and drm fbdev drivers enabled, >> none of the non-DRM graphic cards would be loaded anyway and this code >> wouldn't be executed anyway. > > Yes, exactly. That's why there is no point in compiling that code. As long as you have a graphic card which is not supported by DRM you still need it. Here is my proposed way forward: a) I will resend the patches which reverts the remove-fbcon-hardware-scolling patches to the mailing lists. I'll adjust the stable tags and update the commit messages. b) Then after some days I'll include it in the fbdev for-next git branch. That way it's included in the various build & test chains. c) If everything is working well, I'll push that change during the next merge window for kernel 5.18. If problems arise we will need to discuss. While the patches are in the fbdev git tree we should decide how to exclude code which is not needed for DRM. What about this proposal: a) adding a Kconfig option like: CONFIG_FB_DRIVERS - enable if you need the fbdev drivers. For DRM-only this should be disabled. b) Add to every native fbdev driver a "depends on FB_DRIVERS" in the Kconfig files. c) That way we can use "#if defined(CONFIG_FB_DRIVERS).." to exclude code in fbcon which isn't needed by DRM. Thoughts? Helge
[snip] > > What about this proposal: > a) adding a Kconfig option like: > CONFIG_FB_DRIVERS - enable if you need the fbdev drivers. For DRM-only this should be disabled. > b) Add to every native fbdev driver a "depends on FB_DRIVERS" in the Kconfig files. > c) That way we can use "#if defined(CONFIG_FB_DRIVERS).." to exclude code in fbcon which > isn't needed by DRM. > I proposed something similar in: https://lore.kernel.org/lkml/20210827100027.1577561-1-javierm@redhat.com/t/ Best regards,
Hi Am 24.01.22 um 12:10 schrieb Helge Deller: [...] > > Here is my proposed way forward: > a) I will resend the patches which reverts the remove-fbcon-hardware-scolling patches > to the mailing lists. I'll adjust the stable tags and update the commit messages. > b) Then after some days I'll include it in the fbdev for-next git branch. That way it's > included in the various build & test chains. > c) If everything is working well, I'll push that change during the next merge window > for kernel 5.18. If problems arise we will need to discuss. > > While the patches are in the fbdev git tree we should decide how to exclude code > which is not needed for DRM. > > What about this proposal: > a) adding a Kconfig option like: > CONFIG_FB_DRIVERS - enable if you need the fbdev drivers. For DRM-only this should be disabled. > b) Add to every native fbdev driver a "depends on FB_DRIVERS" in the Kconfig files. > c) That way we can use "#if defined(CONFIG_FB_DRIVERS).." to exclude code in fbcon which > isn't needed by DRM. > > Thoughts? I can't say I approve keeping fbdev alive, but... With fbdev emulation, every DRM driver is an fbdev driver too. So CONFIG_FB_DRIVER is somewhat misleading. Better add an option like CONFIG_FBCON_HW_SCROLLING and have it selected by the fbdev drivers that absolutely need HW acceleration. That option would then protect the rsp code. Best regards Thomas > > Helge
On 1/24/22 12:33, Thomas Zimmermann wrote: [snip] >> Thoughts? > > I can't say I approve keeping fbdev alive, but... > > With fbdev emulation, every DRM driver is an fbdev driver too. So > CONFIG_FB_DRIVER is somewhat misleading. Better add an option like > CONFIG_FBCON_HW_SCROLLING and have it selected by the fbdev drivers that > absolutely need HW acceleration. That option would then protect the rsp > code. > Agreed that this option would be better and allow distros to disable the code that was reverted. Best regards,
On 1/24/22 12:50, Javier Martinez Canillas wrote: > On 1/24/22 12:33, Thomas Zimmermann wrote: > > [snip] > >>> Thoughts? >> >> I can't say I approve keeping fbdev alive, but... >> >> With fbdev emulation, every DRM driver is an fbdev driver too. So >> CONFIG_FB_DRIVER is somewhat misleading. Better add an option like >> CONFIG_FBCON_HW_SCROLLING and have it selected by the fbdev drivers that >> absolutely need HW acceleration. That option would then protect the rsp >> code. I'm not a fan of something like CONFIG_FBCON_HW_SCROLLING, but I'm not against it either. For me it sounds that this is not the real direction you want to go, which is to prevent that any other drivers take the framebuffer before you take it with simpledrm or similiar. CONFIG_FBCON_HW_SCROLLING IMHO just disables the (from your POV) neglectable accleration part. With an option like CONFIG_FB_DRIVER (maybe better: CONFIG_FB_LEGACY_DRIVERS) it's an easy option for distros to disable all of the legacy drivers from being built & shipped. Instead of CONFIG_FBCON_HW_SCROLLING we could also choose CONFIG_FBCON_LEGACY_ACCELERATION, because it includes fillrect() as well... > Agreed that this option would be better and allow distros > to disable the code that was reverted. Yes, but IMHO it doesn't hurt either to leave it in. It doesn't break anything at least. Anyway... Helge
Hi Am 24.01.22 um 16:29 schrieb Helge Deller: > On 1/24/22 12:50, Javier Martinez Canillas wrote: >> On 1/24/22 12:33, Thomas Zimmermann wrote: >> >> [snip] >> >>>> Thoughts? >>> >>> I can't say I approve keeping fbdev alive, but... >>> >>> With fbdev emulation, every DRM driver is an fbdev driver too. So >>> CONFIG_FB_DRIVER is somewhat misleading. Better add an option like >>> CONFIG_FBCON_HW_SCROLLING and have it selected by the fbdev drivers that >>> absolutely need HW acceleration. That option would then protect the rsp >>> code. > > I'm not a fan of something like CONFIG_FBCON_HW_SCROLLING, but I'm not > against it either. > For me it sounds that this is not the real direction you want to go, > which is to prevent that any other drivers take the framebuffer before > you take it with simpledrm or similiar. > CONFIG_FBCON_HW_SCROLLING IMHO just disables the (from your POV) neglectable accleration part. > With an option like CONFIG_FB_DRIVER (maybe better: CONFIG_FB_LEGACY_DRIVERS) > it's an easy option for distros to disable all of the legacy drivers > from being built & shipped. These drivers have been disabled by most distros a long time ago. Those that still remain are the generic, soon to be replaced, ones; and drivers for niche architectures where no DRM-based replacement exists. If I run DRM with fbdev emulation, HW scrolling is unused, possibly buggy, and I'd want to not built it if possible. I guess that's what most distros would want as well. That's the use case for FBCON_HW_SCROLLING. Best regards Thomas
Hi Helge, On Mon, Jan 24, 2022 at 4:30 PM Helge Deller <deller@gmx.de> wrote: > On 1/24/22 12:50, Javier Martinez Canillas wrote: > > On 1/24/22 12:33, Thomas Zimmermann wrote: > > > > [snip] > > > >>> Thoughts? > >> > >> I can't say I approve keeping fbdev alive, but... > >> > >> With fbdev emulation, every DRM driver is an fbdev driver too. So > >> CONFIG_FB_DRIVER is somewhat misleading. Better add an option like > >> CONFIG_FBCON_HW_SCROLLING and have it selected by the fbdev drivers that > >> absolutely need HW acceleration. That option would then protect the rsp > >> code. > > I'm not a fan of something like CONFIG_FBCON_HW_SCROLLING, but I'm not > against it either. > For me it sounds that this is not the real direction you want to go, > which is to prevent that any other drivers take the framebuffer before > you take it with simpledrm or similiar. > CONFIG_FBCON_HW_SCROLLING IMHO just disables the (from your POV) neglectable accleration part. > With an option like CONFIG_FB_DRIVER (maybe better: CONFIG_FB_LEGACY_DRIVERS) > it's an easy option for distros to disable all of the legacy drivers > from being built & shipped. > > Instead of CONFIG_FBCON_HW_SCROLLING we could also choose > CONFIG_FBCON_LEGACY_ACCELERATION, because it includes fillrect() as well... As this is about resurrecting features indicated by the various FBINFO_HWACCEL_* flags, what about CONFIG_FB_HWACCEL? Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds
On Mon, Jan 24, 2022 at 12:33 PM Thomas Zimmermann <tzimmermann@suse.de> wrote:
> With fbdev emulation, every DRM driver is an fbdev driver too. So
Some are even without?
drivers/gpu/drm/vmwgfx/vmwgfx_fb.c: ret = register_framebuffer(info);
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
On Mon, Jan 24, 2022 at 04:29:34PM +0100, Helge Deller wrote: > On 1/24/22 12:50, Javier Martinez Canillas wrote: > > On 1/24/22 12:33, Thomas Zimmermann wrote: > > > > [snip] > > > >>> Thoughts? > >> > >> I can't say I approve keeping fbdev alive, but... > >> > >> With fbdev emulation, every DRM driver is an fbdev driver too. So > >> CONFIG_FB_DRIVER is somewhat misleading. Better add an option like > >> CONFIG_FBCON_HW_SCROLLING and have it selected by the fbdev drivers that > >> absolutely need HW acceleration. That option would then protect the rsp > >> code. > > I'm not a fan of something like CONFIG_FBCON_HW_SCROLLING, but I'm not > against it either. > For me it sounds that this is not the real direction you want to go, > which is to prevent that any other drivers take the framebuffer before > you take it with simpledrm or similiar. > CONFIG_FBCON_HW_SCROLLING IMHO just disables the (from your POV) neglectable accleration part. > With an option like CONFIG_FB_DRIVER (maybe better: CONFIG_FB_LEGACY_DRIVERS) > it's an easy option for distros to disable all of the legacy drivers > from being built & shipped. > > Instead of CONFIG_FBCON_HW_SCROLLING we could also choose > CONFIG_FBCON_LEGACY_ACCELERATION, because it includes fillrect() as well... +1 on that name, since on the lwn discussions I've also seen some noise about resurrecting scrollback. And I guess we could do that too and then just add it all behind that same option. -Daniel > > Agreed that this option would be better and allow distros > > to disable the code that was reverted. > > Yes, but IMHO it doesn't hurt either to leave it in. > It doesn't break anything at least. > Anyway... > > Helge
Hi Am 24.01.22 um 16:50 schrieb Geert Uytterhoeven: > On Mon, Jan 24, 2022 at 12:33 PM Thomas Zimmermann <tzimmermann@suse.de> wrote: >> With fbdev emulation, every DRM driver is an fbdev driver too. So > > Some are even without? > > drivers/gpu/drm/vmwgfx/vmwgfx_fb.c: ret = register_framebuffer(info); Well, I counted this as 'emulation' as well. There's fully driver-agnostic fbdev support in DRM, but some drivers still run their own. At some point, we want all drivers to use DRM's generic solution. Best regards Thomas > > Gr{oetje,eeting}s, > > Geert > > -- > Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org > > In personal conversations with technical people, I call myself a hacker. But > when I'm talking to journalists I just say "programmer" or something like that. > -- Linus Torvalds
diff --git a/Documentation/gpu/todo.rst b/Documentation/gpu/todo.rst index 29506815d24a..a1212b5b3026 100644 --- a/Documentation/gpu/todo.rst +++ b/Documentation/gpu/todo.rst @@ -300,27 +300,6 @@ Contact: Daniel Vetter, Noralf Tronnes Level: Advanced -Garbage collect fbdev scrolling acceleration --------------------------------------------- - -Scroll acceleration is disabled in fbcon by hard-wiring p->scrollmode = -SCROLL_REDRAW. There's a ton of code this will allow us to remove: - -- lots of code in fbcon.c - -- a bunch of the hooks in fbcon_ops, maybe the remaining hooks could be called - directly instead of the function table (with a switch on p->rotate) - -- fb_copyarea is unused after this, and can be deleted from all drivers - -Note that not all acceleration code can be deleted, since clearing and cursor -support is still accelerated, which might be good candidates for further -deletion projects. - -Contact: Daniel Vetter - -Level: Intermediate - idr_init_base() --------------- diff --git a/drivers/video/fbdev/core/fbcon.c b/drivers/video/fbdev/core/fbcon.c index 22bb3892f6bd..b813985f1403 100644 --- a/drivers/video/fbdev/core/fbcon.c +++ b/drivers/video/fbdev/core/fbcon.c @@ -1025,7 +1025,7 @@ static void fbcon_init(struct vc_data *vc, int init) struct vc_data *svc = *default_mode; struct fbcon_display *t, *p = &fb_display[vc->vc_num]; int logo = 1, new_rows, new_cols, rows, cols; - int ret; + int cap, ret; if (WARN_ON(info_idx == -1)) return; @@ -1034,6 +1034,7 @@ static void fbcon_init(struct vc_data *vc, int init) con2fb_map[vc->vc_num] = info_idx; info = registered_fb[con2fb_map[vc->vc_num]]; + cap = info->flags; if (logo_shown < 0 && console_loglevel <= CONSOLE_LOGLEVEL_QUIET) logo_shown = FBCON_LOGO_DONTSHOW; @@ -1135,13 +1136,11 @@ static void fbcon_init(struct vc_data *vc, int init) ops->graphics = 0; - /* - * No more hw acceleration for fbcon. - * - * FIXME: Garbage collect all the now dead code after sufficient time - * has passed. - */ - p->scrollmode = SCROLL_REDRAW; + if ((cap & FBINFO_HWACCEL_COPYAREA) && + !(cap & FBINFO_HWACCEL_DISABLED)) + p->scrollmode = SCROLL_MOVE; + else /* default to something safe */ + p->scrollmode = SCROLL_REDRAW; /* * ++guenther: console.c:vc_allocate() relies on initializing @@ -1953,15 +1952,45 @@ static void updatescrollmode(struct fbcon_display *p, { struct fbcon_ops *ops = info->fbcon_par; int fh = vc->vc_font.height; + int cap = info->flags; + u16 t = 0; + int ypan = FBCON_SWAP(ops->rotate, info->fix.ypanstep, + info->fix.xpanstep); + int ywrap = FBCON_SWAP(ops->rotate, info->fix.ywrapstep, t); int yres = FBCON_SWAP(ops->rotate, info->var.yres, info->var.xres); int vyres = FBCON_SWAP(ops->rotate, info->var.yres_virtual, info->var.xres_virtual); + int good_pan = (cap & FBINFO_HWACCEL_YPAN) && + divides(ypan, vc->vc_font.height) && vyres > yres; + int good_wrap = (cap & FBINFO_HWACCEL_YWRAP) && + divides(ywrap, vc->vc_font.height) && + divides(vc->vc_font.height, vyres) && + divides(vc->vc_font.height, yres); + int reading_fast = cap & FBINFO_READS_FAST; + int fast_copyarea = (cap & FBINFO_HWACCEL_COPYAREA) && + !(cap & FBINFO_HWACCEL_DISABLED); + int fast_imageblit = (cap & FBINFO_HWACCEL_IMAGEBLIT) && + !(cap & FBINFO_HWACCEL_DISABLED); p->vrows = vyres/fh; if (yres > (fh * (vc->vc_rows + 1))) p->vrows -= (yres - (fh * vc->vc_rows)) / fh; if ((yres % fh) && (vyres % fh < yres % fh)) p->vrows--; + + if (good_wrap || good_pan) { + if (reading_fast || fast_copyarea) + p->scrollmode = good_wrap ? + SCROLL_WRAP_MOVE : SCROLL_PAN_MOVE; + else + p->scrollmode = good_wrap ? SCROLL_REDRAW : + SCROLL_PAN_REDRAW; + } else { + if (reading_fast || (fast_copyarea && !fast_imageblit)) + p->scrollmode = SCROLL_MOVE; + else + p->scrollmode = SCROLL_REDRAW; + } } #define PITCH(w) (((w) + 7) >> 3)
This reverts commit 39aead8373b3c20bb5965c024dfb51a94e526151. Revert this patch. This patch started to introduce the regression that all hardware acceleration of more than 35 existing fbdev drivers were bypassed and thus fbcon console output for those was dramatically slowed down by factor of 10 and more. Reverting this commit has no impact on DRM, since none of the DRM drivers are tagged with the acceleration flags FBINFO_HWACCEL_COPYAREA, FBINFO_HWACCEL_FILLRECT or others. Signed-off-by: Helge Deller <deller@gmx.de> Cc: stable@vger.kernel.org # v5.16 --- Documentation/gpu/todo.rst | 21 --------------- drivers/video/fbdev/core/fbcon.c | 45 ++++++++++++++++++++++++++------ 2 files changed, 37 insertions(+), 29 deletions(-) -- 2.31.1