Message ID | 20240905122020.872466-1-ubizjak@gmail.com (mailing list archive) |
---|---|
Headers | show |
Series | random: Include <linux/percpu.h> and resolve circular include dependency | expand |
On Thu, Sep 05, 2024 at 02:17:08PM +0200, Uros Bizjak wrote: > There were several attempts to resolve circular include dependency > after the addition of percpu.h: 1c9df907da83 ("random: fix circular > include dependency on arm64 after addition of percpu.h"), c0842fbc1b18 > ("random32: move the pseudo-random 32-bit definitions to prandom.h") and > finally d9f29deb7fe8 ("prandom: Remove unused include") that completely > removes inclusion of <linux/percpu.h>. > > Due to legacy reasons, <linux/random.h> includes <linux/prandom.h>, but > with the commit entry remark: > > --quote-- > A further cleanup step would be to remove this from <linux/random.h> > entirely, and make people who use the prandom infrastructure include > just the new header file. That's a bit of a churn patch, but grepping > for "prandom_" and "next_pseudo_random32" "struct rnd_state" should > catch most users. > > But it turns out that that nice cleanup step is fairly painful, because > a _lot_ of code currently seems to depend on the implicit include of > <linux/random.h>, which can currently come in a lot of ways, including > such fairly core headfers as <linux/net.h>. > > So the "nice cleanup" part may or may never happen. > --/quote-- > > __percpu tag is currently defined in include/linux/compiler_types.h, > so there is no direct need for the inclusion of <linux/percpu.h>. > However, in [1] we would like to repurpose __percpu tag as a named > address space qualifier, where __percpu macro uses defines from > <linux/percpu.h>. > > This patch series is the "nice cleanup" part, and allows us to finally > include <linux/percpu.h> in prandom.h. > > The whole series was tested by compiling the kernel for x86_64 allconfig > and some popular architectures, namely arm64 defconfig, powerpc defconfig > and loongarch defconfig. Obvious question(s) is(are): 1) have you seen the Ingo's gigantic patch series towards resolving issues with the headers? 2) if not, please look at the preliminary work and take something from there, I believe there are many useful changes already waiting for a couple of years to be applied. Because I haven't found any references nor mentions of that in the cover letter here and explanation why it was not taking into consideration. > [1] https://lore.kernel.org/lkml/20240812115945.484051-4-ubizjak@gmail.com/
On Thu, Sep 5, 2024 at 2:41 PM Andy Shevchenko <andriy.shevchenko@linux.intel.com> wrote: > > On Thu, Sep 05, 2024 at 02:17:08PM +0200, Uros Bizjak wrote: > > There were several attempts to resolve circular include dependency > > after the addition of percpu.h: 1c9df907da83 ("random: fix circular > > include dependency on arm64 after addition of percpu.h"), c0842fbc1b18 > > ("random32: move the pseudo-random 32-bit definitions to prandom.h") and > > finally d9f29deb7fe8 ("prandom: Remove unused include") that completely > > removes inclusion of <linux/percpu.h>. > > > > Due to legacy reasons, <linux/random.h> includes <linux/prandom.h>, but > > with the commit entry remark: > > > > --quote-- > > A further cleanup step would be to remove this from <linux/random.h> > > entirely, and make people who use the prandom infrastructure include > > just the new header file. That's a bit of a churn patch, but grepping > > for "prandom_" and "next_pseudo_random32" "struct rnd_state" should > > catch most users. > > > > But it turns out that that nice cleanup step is fairly painful, because > > a _lot_ of code currently seems to depend on the implicit include of > > <linux/random.h>, which can currently come in a lot of ways, including > > such fairly core headfers as <linux/net.h>. > > > > So the "nice cleanup" part may or may never happen. > > --/quote-- > > > > __percpu tag is currently defined in include/linux/compiler_types.h, > > so there is no direct need for the inclusion of <linux/percpu.h>. > > However, in [1] we would like to repurpose __percpu tag as a named > > address space qualifier, where __percpu macro uses defines from > > <linux/percpu.h>. > > > > This patch series is the "nice cleanup" part, and allows us to finally > > include <linux/percpu.h> in prandom.h. > > > > The whole series was tested by compiling the kernel for x86_64 allconfig > > and some popular architectures, namely arm64 defconfig, powerpc defconfig > > and loongarch defconfig. > > Obvious question(s) is(are): > 1) have you seen the Ingo's gigantic patch series towards resolving issues with > the headers? > 2) if not, please look at the preliminary work and take something from there, I > believe there are many useful changes already waiting for a couple of years to > be applied. > > Because I haven't found any references nor mentions of that in the cover letter > here and explanation why it was not taking into consideration. I am aware of the series, but the patch takes only a small bite of it and specifically resolves the inclusion of <linux/prandom.h> from linux/random.h. The series fixes the missing inclusion of <linux/prandom.h> in files that use pseudo-random function and it was not meant to be a general header cleanup. The end goal is to allow us to include <linux/percpu.h> in linux/prandom.h - which uses __percpu tag without the correct include. Thus, the patch series is only tangentially connected to Ingo's patch series. Best regards, Uros.
On Thu, Sep 05, 2024 at 03:03:24PM +0200, Uros Bizjak wrote: > On Thu, Sep 5, 2024 at 2:41 PM Andy Shevchenko > <andriy.shevchenko@linux.intel.com> wrote: > > > > On Thu, Sep 05, 2024 at 02:17:08PM +0200, Uros Bizjak wrote: > > > There were several attempts to resolve circular include dependency > > > after the addition of percpu.h: 1c9df907da83 ("random: fix circular > > > include dependency on arm64 after addition of percpu.h"), c0842fbc1b18 > > > ("random32: move the pseudo-random 32-bit definitions to prandom.h") and > > > finally d9f29deb7fe8 ("prandom: Remove unused include") that completely > > > removes inclusion of <linux/percpu.h>. > > > > > > Due to legacy reasons, <linux/random.h> includes <linux/prandom.h>, but > > > with the commit entry remark: > > > > > > --quote-- > > > A further cleanup step would be to remove this from <linux/random.h> > > > entirely, and make people who use the prandom infrastructure include > > > just the new header file. That's a bit of a churn patch, but grepping > > > for "prandom_" and "next_pseudo_random32" "struct rnd_state" should > > > catch most users. > > > > > > But it turns out that that nice cleanup step is fairly painful, because > > > a _lot_ of code currently seems to depend on the implicit include of > > > <linux/random.h>, which can currently come in a lot of ways, including > > > such fairly core headfers as <linux/net.h>. > > > > > > So the "nice cleanup" part may or may never happen. > > > --/quote-- > > > > > > __percpu tag is currently defined in include/linux/compiler_types.h, > > > so there is no direct need for the inclusion of <linux/percpu.h>. > > > However, in [1] we would like to repurpose __percpu tag as a named > > > address space qualifier, where __percpu macro uses defines from > > > <linux/percpu.h>. > > > > > > This patch series is the "nice cleanup" part, and allows us to finally > > > include <linux/percpu.h> in prandom.h. > > > > > > The whole series was tested by compiling the kernel for x86_64 allconfig > > > and some popular architectures, namely arm64 defconfig, powerpc defconfig > > > and loongarch defconfig. > > > > Obvious question(s) is(are): > > 1) have you seen the Ingo's gigantic patch series towards resolving issues with > > the headers? > > 2) if not, please look at the preliminary work and take something from there, I > > believe there are many useful changes already waiting for a couple of years to > > be applied. > > > > Because I haven't found any references nor mentions of that in the cover letter > > here and explanation why it was not taking into consideration. > > I am aware of the series, but the patch takes only a small bite of it > and specifically resolves the inclusion of <linux/prandom.h> from > linux/random.h. The series fixes the missing inclusion of > <linux/prandom.h> in files that use pseudo-random function and it was > not meant to be a general header cleanup. The end goal is to allow us > to include <linux/percpu.h> in linux/prandom.h - which uses __percpu > tag without the correct include. Thank you for elaboration, it's all clear now. > Thus, the patch series is only tangentially connected to Ingo's patch series.