* [PATCH] mm: alloc_tag: change the KASAN_TAG_WIDTH for HW_TAGS
@ 2025-04-26 18:08 Guilherme Giacomo Simoes
2025-04-28 0:25 ` Andrew Morton
0 siblings, 1 reply; 3+ messages in thread
From: Guilherme Giacomo Simoes @ 2025-04-26 18:08 UTC (permalink / raw)
To: akpm, pasha.tatashin, surenb
Cc: linux-mm, linux-kernel, Guilherme Giacomo Simoes, Andrey Konovalov
the KASAN_TAG_WIDTH is 8 bits for both (HW_TAGS and SW_TAGS), but for
HW_TAGS the KASAN_TAG_WIDTH can be 4 bits bits because due to the design
of the MTE the memory words for storing metadata only need 4 bits.
Change the preprocessor define KASAN_TAG_WIDTH for check if SW_TAGS is
define, so KASAN_TAG_WIDTH should be 8 bits, but if HW_TAGS is define,
so KASAN_TAG_WIDTH should be 4 bits.
Suggested-by: Andrey Konovalov <andreyknvl@gmail.com>
Signed-off-by: Guilherme Giacomo Simoes <trintaeoitogc@gmail.com>
---
include/linux/page-flags-layout.h | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/include/linux/page-flags-layout.h b/include/linux/page-flags-layout.h
index 4f5c9e979bb9..760006b1c480 100644
--- a/include/linux/page-flags-layout.h
+++ b/include/linux/page-flags-layout.h
@@ -72,8 +72,10 @@
#define NODE_NOT_IN_PAGE_FLAGS 1
#endif
-#if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
+#if defined(CONFIG_KASAN_SW_TAGS)
#define KASAN_TAG_WIDTH 8
+#elif defined(CONFIG_KASAN_HW_TAGS)
+#define KASAN_TAG_WIDTH 4
#else
#define KASAN_TAG_WIDTH 0
#endif
--
2.34.1
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] mm: alloc_tag: change the KASAN_TAG_WIDTH for HW_TAGS
2025-04-26 18:08 [PATCH] mm: alloc_tag: change the KASAN_TAG_WIDTH for HW_TAGS Guilherme Giacomo Simoes
@ 2025-04-28 0:25 ` Andrew Morton
2025-04-28 10:57 ` Guilherme Giacomo Simoes
0 siblings, 1 reply; 3+ messages in thread
From: Andrew Morton @ 2025-04-28 0:25 UTC (permalink / raw)
To: Guilherme Giacomo Simoes
Cc: pasha.tatashin, surenb, linux-mm, linux-kernel, Andrey Konovalov
On Sat, 26 Apr 2025 15:08:37 -0300 Guilherme Giacomo Simoes <trintaeoitogc@gmail.com> wrote:
> the KASAN_TAG_WIDTH is 8 bits for both (HW_TAGS and SW_TAGS), but for
> HW_TAGS the KASAN_TAG_WIDTH can be 4 bits bits because due to the design
> of the MTE the memory words for storing metadata only need 4 bits.
> Change the preprocessor define KASAN_TAG_WIDTH for check if SW_TAGS is
> define, so KASAN_TAG_WIDTH should be 8 bits, but if HW_TAGS is define,
> so KASAN_TAG_WIDTH should be 4 bits.
Why?
I assume the effect is to save a few flags bits under some
circumstances? But that's just me guessing - please fully describe the
motivation and effects within changelogs.
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] mm: alloc_tag: change the KASAN_TAG_WIDTH for HW_TAGS
2025-04-28 0:25 ` Andrew Morton
@ 2025-04-28 10:57 ` Guilherme Giacomo Simoes
0 siblings, 0 replies; 3+ messages in thread
From: Guilherme Giacomo Simoes @ 2025-04-28 10:57 UTC (permalink / raw)
To: akpm
Cc: andreyknvl, linux-kernel, linux-mm, pasha.tatashin, surenb,
trintaeoitogc
Andrew Morton <akpm@linux-foundation.org> wrote:
> > the KASAN_TAG_WIDTH is 8 bits for both (HW_TAGS and SW_TAGS), but for
> > HW_TAGS the KASAN_TAG_WIDTH can be 4 bits bits because due to the design
> > of the MTE the memory words for storing metadata only need 4 bits.
> > Change the preprocessor define KASAN_TAG_WIDTH for check if SW_TAGS is
> > define, so KASAN_TAG_WIDTH should be 8 bits, but if HW_TAGS is define,
> > so KASAN_TAG_WIDTH should be 4 bits.
>
> Why?
>
> I assume the effect is to save a few flags bits under some
> circumstances? But that's just me guessing - please fully describe the
> motivation and effects within changelogs.
Ok, I will change the commit message with the "why", and send a v2.
Thanks.
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2025-04-28 10:57 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2025-04-26 18:08 [PATCH] mm: alloc_tag: change the KASAN_TAG_WIDTH for HW_TAGS Guilherme Giacomo Simoes
2025-04-28 0:25 ` Andrew Morton
2025-04-28 10:57 ` Guilherme Giacomo Simoes
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox