| Message ID | 20210408092011.52763-1-david@redhat.com |
|---|---|
| Headers | show |
| Series | drivers: don't select DMA_CMA or CMA | expand |
On Thu, Apr 08, 2021 at 11:20:11AM +0200, David Hildenbrand wrote: > Random drivers should not override a user configuration of core knobs > (e.g., CONFIG_DMA_CMA=n). Use "imply" instead, to still respect > dependencies and manual overrides. > > "This is similar to "select" as it enforces a lower limit on another > symbol except that the "implied" symbol's value may still be set to n > from a direct dependency or with a visible prompt." > > Implying DRM_CMA should be sufficient, as that depends on CMA. ^ DMA_CMA > > Note: If this is a real dependency, we should use "depends on DMA_CMA" > instead - but I assume the driver can work without CMA just fine -- > esp. when we wouldn't have HAVE_DMA_CONTIGUOUS right now. > > Signed-off-by: David Hildenbrand <david@redhat.com> > --- > drivers/gpu/drm/aspeed/Kconfig | 3 +-- > drivers/gpu/drm/etnaviv/Kconfig | 3 +-- > 2 files changed, 2 insertions(+), 4 deletions(-) > > diff --git a/drivers/gpu/drm/aspeed/Kconfig b/drivers/gpu/drm/aspeed/Kconfig > index 5e95bcea43e9..a055f763d230 100644 > --- a/drivers/gpu/drm/aspeed/Kconfig > +++ b/drivers/gpu/drm/aspeed/Kconfig > @@ -6,9 +6,8 @@ config DRM_ASPEED_GFX > depends on MMU > select DRM_KMS_HELPER > select DRM_KMS_CMA_HELPER > - select DMA_CMA if HAVE_DMA_CONTIGUOUS > - select CMA if HAVE_DMA_CONTIGUOUS > select MFD_SYSCON > + imply DRM_CMA Ditto > help > Chose this option if you have an ASPEED AST2500 SOC Display > Controller (aka GFX). > diff --git a/drivers/gpu/drm/etnaviv/Kconfig b/drivers/gpu/drm/etnaviv/Kconfig > index faa7fc68b009..5f5576b7221a 100644 > --- a/drivers/gpu/drm/etnaviv/Kconfig > +++ b/drivers/gpu/drm/etnaviv/Kconfig > @@ -9,9 +9,8 @@ config DRM_ETNAVIV > select THERMAL if DRM_ETNAVIV_THERMAL > select TMPFS > select WANT_DEV_COREDUMP > - select CMA if HAVE_DMA_CONTIGUOUS > - select DMA_CMA if HAVE_DMA_CONTIGUOUS > select DRM_SCHED > + imply DMA_CMA > help > DRM driver for Vivante GPUs. > > -- > 2.30.2 >
On Thu, Apr 8, 2021 at 11:22 AM David Hildenbrand <david@redhat.com> wrote: > > Random drivers should not override a user configuration of core knobs > (e.g., CONFIG_DMA_CMA=n). Use "imply" instead, to still respect > dependencies and manual overrides. > > "This is similar to "select" as it enforces a lower limit on another > symbol except that the "implied" symbol's value may still be set to n > from a direct dependency or with a visible prompt." > > Implying DRM_CMA should be sufficient, as that depends on CMA. > > Note: If this is a real dependency, we should use "depends on DMA_CMA" > instead - but I assume the driver can work without CMA just fine -- > esp. when we wouldn't have HAVE_DMA_CONTIGUOUS right now. 'imply' is almost never the right solution, and it tends to cause more problems than it solves. In particular, it does not prevent a configuration with 'DRM_CMA=m' and 'DRMA_ASPEED_GFX=y', or any build failures from such a configuration. If you want this kind of soft dependency, you need 'depends on DRM_CMA || !DRM_CMA'. Arnd
On 08.04.21 12:27, David Hildenbrand wrote: > On 08.04.21 12:20, Arnd Bergmann wrote: >> On Thu, Apr 8, 2021 at 11:22 AM David Hildenbrand <david@redhat.com> wrote: >>> >>> Random drivers should not override a user configuration of core knobs >>> (e.g., CONFIG_DMA_CMA=n). Use "imply" instead, to still respect >>> dependencies and manual overrides. >>> >>> "This is similar to "select" as it enforces a lower limit on another >>> symbol except that the "implied" symbol's value may still be set to n >>> from a direct dependency or with a visible prompt." >>> >>> Implying DRM_CMA should be sufficient, as that depends on CMA. >>> >>> Note: If this is a real dependency, we should use "depends on DMA_CMA" >>> instead - but I assume the driver can work without CMA just fine -- >>> esp. when we wouldn't have HAVE_DMA_CONTIGUOUS right now. >> >> 'imply' is almost never the right solution, and it tends to cause more >> problems than it solves. > > I thought that was the case with "select" :) > >> >> In particular, it does not prevent a configuration with 'DRM_CMA=m' > > I assume you meant "DRM_CMA=n" ? DRM_CMA cannot be built as a module. > >> and 'DRMA_ASPEED_GFX=y', or any build failures from such >> a configuration. > > I don't follow. "DRM_CMA=n" and 'DRMA_ASPEED_GFX=y' is supposed to work > just fine (e.g., without HAVE_DMA_CONTIGUOUS) or what am I missing? > >> >> If you want this kind of soft dependency, you need >> 'depends on DRM_CMA || !DRM_CMA'. > > Seriously? I think the point of imply is "please enable if possible and > not prevented by someone else". Your example looks more like a NOP - no? > Or will it have the same effect? I just tried (remove CONFIG_DMA_CMA from .config followed by make) and the default will be set to "N" (when querying the user). So it indeed looks like a NOP - unless I am missing something.
>>> In particular, it does not prevent a configuration with 'DRM_CMA=m' >> >> I assume you meant "DRM_CMA=n" ? DRM_CMA cannot be built as a module. > > Ok, at least that makes it easier. > >>> and 'DRMA_ASPEED_GFX=y', or any build failures from such >>> a configuration. >> >> I don't follow. "DRM_CMA=n" and 'DRMA_ASPEED_GFX=y' is supposed to work >> just fine (e.g., without HAVE_DMA_CONTIGUOUS) or what am I missing? > > I thought you were trying to solve the problem where DRMA_ASPEED_GFX > can optionally link against CMA but would fail to build when the CMA code > is in a loadable module. Yes. I was trying to say: it works with this patch just fine. The issue you described does not seem to apply (DRM_CMA=m). > >> Your example looks more like a NOP - no? >> Or will it have the same effect? > > The example I gave is only meaningful if both are tristate, which is > not the case here as you explain. Okay, thanks. > > It is a somewhat awkward way to say "prevent this symbol from > being =y if the dependency is =m". What would be the right thing to do in the case here then to achieve the "if DRMA_ASPEED_GFX is enabled, also enable DMA_CMA id possible"? One approach could be to have for DMA_CMA default y if DRMA_ASPEED_GFX but it feels like the wrong way to tackle this. Thanks!
On Thu, Apr 8, 2021 at 1:00 PM David Hildenbrand <david@redhat.com> wrote: > > > > It is a somewhat awkward way to say "prevent this symbol from > > being =y if the dependency is =m". > > What would be the right thing to do in the case here then to achieve the > "if DRMA_ASPEED_GFX is enabled, also enable DMA_CMA id possible"? > > One approach could be to have for DMA_CMA > > default y if DRMA_ASPEED_GFX > > but it feels like the wrong way to tackle this. I'm still not sure what you are trying to achieve. Is the idea only to provide a useful default for DMA_CMA depending on which drivers are enabled? This is something you could do using a hidden helper symbol like config DRMA_ASPEED_GFX bool "Aspeed display driver" select DRM_WANT_CMA config DRM_WANT_CMA bool help Select this from any driver that benefits from CMA being enabled config DMA_CMA bool "Use CMA helpers for DRM" default DRM_WANT_CMA Arnd
On Thu, Apr 8, 2021 at 2:00 PM David Hildenbrand <david@redhat.com> wrote: > > On 08.04.21 13:44, Arnd Bergmann wrote: > > On Thu, Apr 8, 2021 at 1:00 PM David Hildenbrand <david@redhat.com> wrote: > >>> > >>> It is a somewhat awkward way to say "prevent this symbol from > >>> being =y if the dependency is =m". > >> > >> What would be the right thing to do in the case here then to achieve the > >> "if DRMA_ASPEED_GFX is enabled, also enable DMA_CMA id possible"? > >> > >> One approach could be to have for DMA_CMA > >> > >> default y if DRMA_ASPEED_GFX > >> > >> but it feels like the wrong way to tackle this. > > > > I'm still not sure what you are trying to achieve. Is the idea only to provide > > a useful default for DMA_CMA depending on which drivers are enabled? > > "Random drivers should not override a user configuration of core knobs > (e.g., CONFIG_DMA_CMA=n)." > > Let's assume I'm a distribution and want to set CONFIG_CMA=n or want to > set CONFIG_DMA_CMA=n with CONFIG_CMA=y; there is no way to do that with > e.g., DRMA_ASPEED_GFX=y because it will always override my (user!) > setting -- even though it doesn't really always need it. Using "select" > is the problem here. I agree on the part of removing the 'select' if we don't need it. The part I couldn't figure out was what the 'imply' is supposed to help with. Most other users that added imply tried (and failed) to fix a build problem. > > This is something you could do using a hidden helper symbol like > > > > config DRMA_ASPEED_GFX > > bool "Aspeed display driver" > > select DRM_WANT_CMA > > > > config DRM_WANT_CMA > > bool > > help > > Select this from any driver that benefits from CMA being enabled > > > > config DMA_CMA > > bool "Use CMA helpers for DRM" > > default DRM_WANT_CMA > > > > Arnd > > > > That's precisely what I had first, with an additional "WANT_CMA" -- but > looking at the number of such existing options (I was able to spot 1 !) > I wondered if there is a better approach to achieve the same; "imply" > sounded like a good candidate. I can probably find a couple more, but regardless of how many others exist, this would be a much clearer way of doing it than 'imply' since it has none of the ambiguity and misuse problems. I think the reason we don't see more is that generally speaking, those defaults are widely ignored anyway. You almost always start out with a defconfig file that contains everything you know you need, and then you add bits to that. Having the default in any form only helps to make that defconfig file one line shorter, while requiring other users to add another line to turn it off when they do not want it. Arnd
On Thu, Apr 8, 2021 at 2:50 PM Linus Walleij <linus.walleij@linaro.org> wrote: > > On Thu, Apr 8, 2021 at 2:01 PM David Hildenbrand <david@redhat.com> wrote: > > > > This is something you could do using a hidden helper symbol like > > > > > > config DRMA_ASPEED_GFX > > > bool "Aspeed display driver" > > > select DRM_WANT_CMA > > > > > > config DRM_WANT_CMA > > > bool > > > help > > > Select this from any driver that benefits from CMA being enabled > > > > > > config DMA_CMA > > > bool "Use CMA helpers for DRM" > > > default DRM_WANT_CMA > > > > > > Arnd > > > > > > > That's precisely what I had first, with an additional "WANT_CMA" -- but > > looking at the number of such existing options (I was able to spot 1 !) > > If you do this it probably makes sense to fix a few other drivers > Kconfig in the process. It's not just a problem with your driver. > "my" drivers: > > drivers/gpu/drm/mcde/Kconfig > drivers/gpu/drm/pl111/Kconfig > drivers/gpu/drm/tve200/Kconfig > > certainly needs this as well, and pretty much anything that is > selecting DRM_KMS_CMA_HELPER or > DRM_GEM_CMA_HELPER "wants" DMA_CMA. Are there any that don't select either of the helpers and still want CMA? If not, it would be easy to just add default DRM_KMS_CMA_HELPER || DRM_GEM_CMA_HELPER and skipt the extra symbol. Arnd
On Thu, Apr 08, 2021 at 01:38:17PM +0200, Daniel Vetter wrote: > If you want to change this, we need automatic conflict resolution like apt > and other package managers have, with suggestions how to fix the config if > you want to enable a driver, but some of its requirements are missing. The > current approach of hiding driver symbols complete if any of their > dependencies are off is really not great. +1 to this.. Though Kconfig is basically already unusuable unless you have hours to carefully craft the kconfig you need to get out. I'm not sure trying to optimize this by abusing the existing language rules is such a good idea. I gave a very half hearted go at a simple heuristic solution solve for kconfig a while ago. It is good enough to sometimes automate a kconfig task, but it is not so nice. I use it to do things like "turn on all RDMA drivers" which is quite a hard to do by hand. It looks liked heursitics need a lot of fine tuning as the conditionals are complex enough that it is hard to guess which branch is going to yield a success. Jason
On Thu, Apr 8, 2021 at 6:44 PM David Hildenbrand <david@redhat.com> wrote: > > drivers/gpu/drm/mcde/Kconfig > > drivers/gpu/drm/pl111/Kconfig > > drivers/gpu/drm/tve200/Kconfig > > I was assuming these are "real" dependencies. Will it also work without > DMA_CMA? It will mostly work but that is only because the reservations are mostly contiguous anyway because they are done early and are small. The hardware requires contiguous buffers in all three cases. I'm not sure I always got it right. > > certainly needs this as well, and pretty much anything that is > > selecting DRM_KMS_CMA_HELPER or > > DRM_GEM_CMA_HELPER "wants" DMA_CMA. > > "wants" as in "desires to use but can life without" or "wants" as in > "really needs it". ? I don't know the exact semantics of using DRM_KMS_CMA* without actually using DMA_CMA. I suspect small allocations will be contiguous and big allocations will start to fragment? but it's just my guess. I guess "really need it"? Yours, Linus Walleij
On 08.04.21 15:19, Arnd Bergmann wrote: > On Thu, Apr 8, 2021 at 2:50 PM Linus Walleij <linus.walleij@linaro.org> wrote: >> >> On Thu, Apr 8, 2021 at 2:01 PM David Hildenbrand <david@redhat.com> wrote: >> >>>> This is something you could do using a hidden helper symbol like >>>> >>>> config DRMA_ASPEED_GFX >>>> bool "Aspeed display driver" >>>> select DRM_WANT_CMA >>>> >>>> config DRM_WANT_CMA >>>> bool >>>> help >>>> Select this from any driver that benefits from CMA being enabled >>>> >>>> config DMA_CMA >>>> bool "Use CMA helpers for DRM" >>>> default DRM_WANT_CMA >>>> >>>> Arnd >>>> >>> >>> That's precisely what I had first, with an additional "WANT_CMA" -- but >>> looking at the number of such existing options (I was able to spot 1 !) >> >> If you do this it probably makes sense to fix a few other drivers >> Kconfig in the process. It's not just a problem with your driver. >> "my" drivers: >> >> drivers/gpu/drm/mcde/Kconfig >> drivers/gpu/drm/pl111/Kconfig >> drivers/gpu/drm/tve200/Kconfig >> >> certainly needs this as well, and pretty much anything that is >> selecting DRM_KMS_CMA_HELPER or >> DRM_GEM_CMA_HELPER "wants" DMA_CMA. > > Are there any that don't select either of the helpers and > still want CMA? If not, it would be easy to just add > > default DRM_KMS_CMA_HELPER || DRM_GEM_CMA_HELPER > > and skipt the extra symbol. That sounds like a reasonable thing to do. I'll look into that. -- Thanks, David / dhildenb
On 08.04.21 22:29, Arnd Bergmann wrote: > On Thu, Apr 8, 2021 at 6:45 PM David Hildenbrand <david@redhat.com> wrote: >> On 08.04.21 14:49, Linus Walleij wrote: >>> On Thu, Apr 8, 2021 at 2:01 PM David Hildenbrand <david@redhat.com> wrote: >>> >>>>> This is something you could do using a hidden helper symbol like >>>>> >>>>> config DRMA_ASPEED_GFX >>>>> bool "Aspeed display driver" >>>>> select DRM_WANT_CMA >>>>> >>>>> config DRM_WANT_CMA >>>>> bool >>>>> help >>>>> Select this from any driver that benefits from CMA being enabled >>>>> >>>>> config DMA_CMA >>>>> bool "Use CMA helpers for DRM" >>>>> default DRM_WANT_CMA >>>>> >>>>> Arnd >>>>> >>>> >>>> That's precisely what I had first, with an additional "WANT_CMA" -- but >>>> looking at the number of such existing options (I was able to spot 1 !) >>> >>> If you do this it probably makes sense to fix a few other drivers >>> Kconfig in the process. It's not just a problem with your driver. >>> "my" drivers: >>> >> >> :) I actually wanted to convert them to "depends on DMA_CMA" but ran >> into recursive dependencies ... >> >>> drivers/gpu/drm/mcde/Kconfig >>> drivers/gpu/drm/pl111/Kconfig >>> drivers/gpu/drm/tve200/Kconfig > > Right, this is the main problem caused by using 'select' to > force-enable symbols that other drivers depend on. > > Usually, the answer is to be consistent about the use of 'select' > and 'depends on', using the former only to enable symbols that > are hidden, while using 'depends on' for anything that is an > actual build time dependency. > >> I was assuming these are "real" dependencies. Will it also work without >> DMA_CMA? > > I think in this case, it is fairly likely to work without DMA_CMA when the > probe function gets called during a fresh boot, but fairly likely to fail if > it gets called after the system has run for long enough to fragment the > free memory. > > The point of DMA_CMA is to make it work reliably. Right, and even at runtime there is no guarantee that DMA_CMA will do anything -- especially if we don't reserve a CMA region (e.g., "cma=X"). So this really sounds like a "desires DMA_CMA" and achieving that via an additional symbol or via "default y if ..." for DMA_CMA sounds reasonable. Thanks! -- Thanks, David / dhildenb