percpu: add __percpu sparse annotations
in [Kernel]
|
Prev: [PATCH] scripts/kallsyms build warning fix
Next: serial: bfin_5xx: kgdboc should accept gdb break only when it is active
From: Al Viro on 25 Jan 2010 18:20 On Tue, Jan 26, 2010 at 12:22:07AM +0900, Tejun Heo wrote: > This patchset adds __percpu sparse annotations to all percpu users > covered by x86_64 allmodconfig. __percpu annotation teaches sparse > that percpu variables live in a separate address space and can't be > accessed directly without going through percpu accessors. This allows > detection of most percpu access mistakes involving both static and > dyanmic percpu variables. > > This patchset contains the following eight patches. > > 0001-percpu-add-__percpu-sparse-annotations-to-core-kerne.patch > 0002-percpu-add-__percpu-sparse-annotations-to-fs.patch > 0003-percpu-add-__percpu-sparse-annotations-to-net.patch > 0004-percpu-add-__percpu-sparse-annotations-to-net-driver.patch > 0005-percpu-add-__percpu-sparse-annotations-to-x86.patch > 0006-percpu-add-__percpu-sparse-annotations-to-trace.patch > 0007-percpu-add-__percpu-sparse-annotations-to-hw_breakpo.patch > 0008-percpu-add-__percpu-sparse-annotations-to-what-s-lef.patch > > As these annotations are for sparse, none of the above patches affects > normal kernel build and most of the conversions are straight-forward > and trivial. There are a few places where the conversion isn't > completely straight-forward (but still fairly trivial). Those are > mentioned in each patch description. > > I can route the patch through percpu and conflict resolution, if > necessary, wouldn't be difficult at all for these changes. If anyone > wants to route one of these patches through a different tree, please > let me know. All that's necessary would be adding dummy __percpu > definition to the patch. > > If nobody objects, I'll push these into percpu tree in three or four > days. Um. Where *is* the definition of __percpu? Presumably, that'd be something like __attribute__((noderef,address_space(4)) under ifdef __CHECKER__ and empty otherwise? If so, I'm fine with that patchset, provided that it does grow that #define and becomes self-contained... -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
From: Tejun Heo on 25 Jan 2010 19:30 Hello, On 01/26/2010 08:14 AM, Al Viro wrote: > Um. Where *is* the definition of __percpu? Presumably, that'd be > something like __attribute__((noderef,address_space(4)) under ifdef > __CHECKER__ and empty otherwise? If so, I'm fine with that patchset, > provided that it does grow that #define and becomes self-contained... Ah... sorry, I forgot about the base tree. The definition has been in the percpu tree for some time now. The patchset is on top of the following branch. git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu.git for-next and the series is available as git tree in the following branch. git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu.git percpu-sparse-review The definition is in include/linux/compiler.h. # define __percpu __attribute__((noderef, address_space(3))) which gets replaced by dummy def if !__CHECKER__. Thanks. -- tejun -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
From: Tejun Heo on 2 Feb 2010 00:50
On 01/26/2010 12:22 AM, Tejun Heo wrote: > If nobody objects, I'll push these into percpu tree in three or four > days. Patchset pushed out w/ the array workarounds dropped. Thanks. -- tejun -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/ |