| Message ID | alpine.LRH.2.02.1805300818140.24812@file01.intranet.prod.int.rdu2.redhat.com (mailing list archive) |
|---|---|
| State | Accepted, archived |
| Delegated to: | Mike Snitzer |
| Headers | show |
On Wed, 30 May 2018 08:19:22 -0400 (EDT) Mikulas Patocka <mpatocka@redhat.com> wrote: > The function __builtin_expect returns long type (see the gcc > documentation), and so do macros likely and unlikely. Unfortunatelly, when > CONFIG_PROFILE_ANNOTATED_BRANCHES is selected, the macros likely and > unlikely expand to __branch_check__ and __branch_check__ truncates the Nice catch. > long type to int. This unintended truncation may cause bugs in various > kernel code (we found a bug in dm-writecache because of it), so it's I'm curious to what that bug was. > better to fix __branch_check__ to return long. > > Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> > Cc: stable@vger.kernel.org Anyway, I can pull this in my tree and test it. -- Steve > > --- > include/linux/compiler.h | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > Index: linux-2.6/include/linux/compiler.h > =================================================================== > --- linux-2.6.orig/include/linux/compiler.h 2018-02-26 20:34:17.000000000 +0100 > +++ linux-2.6/include/linux/compiler.h 2018-05-30 14:11:53.000000000 +0200 > @@ -21,7 +21,7 @@ void ftrace_likely_update(struct ftrace_ > #define unlikely_notrace(x) __builtin_expect(!!(x), 0) > > #define __branch_check__(x, expect, is_constant) ({ \ > - int ______r; \ > + long ______r; \ > static struct ftrace_likely_data \ > __attribute__((__aligned__(4))) \ > __attribute__((section("_ftrace_annotated_branch"))) \ -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel
On Mon, 4 Jun 2018, Steven Rostedt wrote: > On Wed, 30 May 2018 08:19:22 -0400 (EDT) > Mikulas Patocka <mpatocka@redhat.com> wrote: > > > The function __builtin_expect returns long type (see the gcc > > documentation), and so do macros likely and unlikely. Unfortunatelly, when > > CONFIG_PROFILE_ANNOTATED_BRANCHES is selected, the macros likely and > > unlikely expand to __branch_check__ and __branch_check__ truncates the > > Nice catch. > > > long type to int. This unintended truncation may cause bugs in various > > kernel code (we found a bug in dm-writecache because of it), so it's > > I'm curious to what that bug was. printk("%ld", writecache_has_error(wc)) ... and writecache_has_error was defined as #define writecache_has_error(wc) (unlikely(READ_ONCE((wc)->error))) Mikulas -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel
Index: linux-2.6/include/linux/compiler.h =================================================================== --- linux-2.6.orig/include/linux/compiler.h 2018-02-26 20:34:17.000000000 +0100 +++ linux-2.6/include/linux/compiler.h 2018-05-30 14:11:53.000000000 +0200 @@ -21,7 +21,7 @@ void ftrace_likely_update(struct ftrace_ #define unlikely_notrace(x) __builtin_expect(!!(x), 0) #define __branch_check__(x, expect, is_constant) ({ \ - int ______r; \ + long ______r; \ static struct ftrace_likely_data \ __attribute__((__aligned__(4))) \ __attribute__((section("_ftrace_annotated_branch"))) \
The function __builtin_expect returns long type (see the gcc documentation), and so do macros likely and unlikely. Unfortunatelly, when CONFIG_PROFILE_ANNOTATED_BRANCHES is selected, the macros likely and unlikely expand to __branch_check__ and __branch_check__ truncates the long type to int. This unintended truncation may cause bugs in various kernel code (we found a bug in dm-writecache because of it), so it's better to fix __branch_check__ to return long. Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Cc: stable@vger.kernel.org --- include/linux/compiler.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel