| Message ID | 1476441745-10765-1-git-send-email-mark.rutland@arm.com |
|---|---|
| State | New |
| Headers | show |
On Fri, Oct 14, 2016 at 11:42:25AM +0100, Mark Rutland wrote: > When CONFIG_THREAD_INFO_IN_TASK is selected, the current_thread_info() > macro relies on current having been defined prior to its use. However, > not all users of current_thread_info() include <asm/current.h>, and thus > current is not guaranteed to be defined. > > When CONFIG_THREAD_INFO_IN_TASK is not selected, it's possible that > get_current() / current are based upon current_thread_info(), and > <asm/current.h> includes <asm/thread_info.h>. Thus always including > <asm/current.h> would result in circular dependences on some platforms. > > To ensure both cases work, this patch includes <asm/current.h>, but only > when CONFIG_THREAD_INFO_IN_TASK is selected. > > Signed-off-by: Mark Rutland <mark.rutland@arm.com> > --- > include/linux/thread_info.h | 1 + > 1 file changed, 1 insertion(+) > > As discussed, it would be great if this could go in along with the patch to > make thread_info arch-specific again, to make merging the arch-specific parts > easier (for arm64 and s390). Urrgh; I've just recalled that this patch alone is insufficient. The h8300 arch code has an unfixed bug [1], and relies on some implicit definition ordering that will be broken by this patch. I have a workaround/cleanup for core code that I'll send with an updated version of my arm64 series shortly. Thanks, Mark. [1] http://lkml.kernel.org/r/<1476714934-11635-1-git-send-email-mark.rutland@arm.com > Mark. > > diff --git a/include/linux/thread_info.h b/include/linux/thread_info.h > index 45f004e..521f84b 100644 > --- a/include/linux/thread_info.h > +++ b/include/linux/thread_info.h > @@ -25,6 +25,7 @@ struct thread_info { > #endif > > #ifdef CONFIG_THREAD_INFO_IN_TASK > +#include <asm/current.h> > #define current_thread_info() ((struct thread_info *)current) > #endif > > -- > 1.9.1 >
On Mon, Oct 17, 2016 at 03:48:13PM +0100, Mark Rutland wrote: > On Fri, Oct 14, 2016 at 11:42:25AM +0100, Mark Rutland wrote: > > When CONFIG_THREAD_INFO_IN_TASK is selected, the current_thread_info() > > macro relies on current having been defined prior to its use. However, > > not all users of current_thread_info() include <asm/current.h>, and thus > > current is not guaranteed to be defined. > > > > When CONFIG_THREAD_INFO_IN_TASK is not selected, it's possible that > > get_current() / current are based upon current_thread_info(), and > > <asm/current.h> includes <asm/thread_info.h>. Thus always including > > <asm/current.h> would result in circular dependences on some platforms. > > > > To ensure both cases work, this patch includes <asm/current.h>, but only > > when CONFIG_THREAD_INFO_IN_TASK is selected. > > > > Signed-off-by: Mark Rutland <mark.rutland@arm.com> > > --- > > include/linux/thread_info.h | 1 + > > 1 file changed, 1 insertion(+) > > > > As discussed, it would be great if this could go in along with the patch to > > make thread_info arch-specific again, to make merging the arch-specific parts > > easier (for arm64 and s390). > > Urrgh; I've just recalled that this patch alone is insufficient. The > h8300 arch code has an unfixed bug [1], and relies on some implicit > definition ordering that will be broken by this patch. > > I have a workaround/cleanup for core code that I'll send with an updated > version of my arm64 series shortly. Looks like I spoke too soon. I have another circular include issue with raw_smp_processor_id() and task_struct::cpu to solve -- it looks like s390 doesn't suffer from that per my reading of your headers. In the mean time, I've pushed out a branch [2] with the common patches, atop of v4.9-rc1. Thanks, Mark. [1] http://lkml.kernel.org/r/<1476714934-11635-1-git-send-email-mark.rutland@arm.com [2] https://git.kernel.org/cgit/linux/kernel/git/mark/linux.git/log/?h=core/ti-stack-split
On Mon, Oct 17, 2016 at 06:33:15PM +0100, Mark Rutland wrote: > On Mon, Oct 17, 2016 at 03:48:13PM +0100, Mark Rutland wrote: > > On Fri, Oct 14, 2016 at 11:42:25AM +0100, Mark Rutland wrote: > > > When CONFIG_THREAD_INFO_IN_TASK is selected, the current_thread_info() > > > macro relies on current having been defined prior to its use. However, > > > not all users of current_thread_info() include <asm/current.h>, and thus > > > current is not guaranteed to be defined. > > > > > > When CONFIG_THREAD_INFO_IN_TASK is not selected, it's possible that > > > get_current() / current are based upon current_thread_info(), and > > > <asm/current.h> includes <asm/thread_info.h>. Thus always including > > > <asm/current.h> would result in circular dependences on some platforms. > > > > > > To ensure both cases work, this patch includes <asm/current.h>, but only > > > when CONFIG_THREAD_INFO_IN_TASK is selected. > > > > > > Signed-off-by: Mark Rutland <mark.rutland@arm.com> > > > --- > > > include/linux/thread_info.h | 1 + > > > 1 file changed, 1 insertion(+) > > > > > > As discussed, it would be great if this could go in along with the patch to > > > make thread_info arch-specific again, to make merging the arch-specific parts > > > easier (for arm64 and s390). > > > > Urrgh; I've just recalled that this patch alone is insufficient. The > > h8300 arch code has an unfixed bug [1], and relies on some implicit > > definition ordering that will be broken by this patch. > > > > I have a workaround/cleanup for core code that I'll send with an updated > > version of my arm64 series shortly. > > Looks like I spoke too soon. I have another circular include issue with > raw_smp_processor_id() and task_struct::cpu to solve -- it looks like > s390 doesn't suffer from that per my reading of your headers. That's correct. > In the mean time, I've pushed out a branch [2] with the common patches, > atop of v4.9-rc1. I just verified that your branch works fine for s390 (with and without the THREAD_INFO_IN_TASK conversion). Thanks, Heiko
diff --git a/include/linux/thread_info.h b/include/linux/thread_info.h index 45f004e..521f84b 100644 --- a/include/linux/thread_info.h +++ b/include/linux/thread_info.h @@ -25,6 +25,7 @@ struct thread_info { #endif #ifdef CONFIG_THREAD_INFO_IN_TASK +#include <asm/current.h> #define current_thread_info() ((struct thread_info *)current) #endif
When CONFIG_THREAD_INFO_IN_TASK is selected, the current_thread_info() macro relies on current having been defined prior to its use. However, not all users of current_thread_info() include <asm/current.h>, and thus current is not guaranteed to be defined. When CONFIG_THREAD_INFO_IN_TASK is not selected, it's possible that get_current() / current are based upon current_thread_info(), and <asm/current.h> includes <asm/thread_info.h>. Thus always including <asm/current.h> would result in circular dependences on some platforms. To ensure both cases work, this patch includes <asm/current.h>, but only when CONFIG_THREAD_INFO_IN_TASK is selected. Signed-off-by: Mark Rutland <mark.rutland@arm.com> --- include/linux/thread_info.h | 1 + 1 file changed, 1 insertion(+) As discussed, it would be great if this could go in along with the patch to make thread_info arch-specific again, to make merging the arch-specific parts easier (for arm64 and s390). Mark. -- 1.9.1