Re: [PATCH] powerpc: fix personality handling in ppc64_personality()

Jiri Kosina [off-list ref] writes:

 	if (personality(current->personality) == PER_LINUX32
-	    && personality == PER_LINUX)
-		personality = PER_LINUX32;
+	    && personality(personality) == PER_LINUX)
+		personality &= ~PER_LINUX | PER_LINUX32;

That doesn't work.  ~PER_LINUX is -1, so this is a no-op.

 	ret = sys_personality(personality);
-	if (ret == PER_LINUX32)
-		ret = PER_LINUX;
+	if (personality(ret) == PER_LINUX32)
+		ret &= ~PER_LINUX32 | PER_LINUX;

That only "works" because PER_LINUX is 0.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Directly comparing current->personality against PER_LINUX32 doesn't work
in cases when any of the personality flags stored in the top three bytes
are used.

Directly forcefully setting personality to PER_LINUX32 or PER_LINUX
discards any flags stored in the top three bytes

Use personality() macro to compare only PER_MASK bytes and make sure that
we are setting only the bits that should be set, instead of
overwriting the whole value.

Signed-off-by: Jiri Kosina <redacted>
---

changed since v1: fix the bit ops to reflect the fact that PER_LINUX is 
actually 0

 arch/powerpc/kernel/syscalls.c |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/arch/powerpc/kernel/syscalls.c b/arch/powerpc/kernel/syscalls.c
index f2496f2..dc1558e 100644
--- a/arch/powerpc/kernel/syscalls.c
+++ b/arch/powerpc/kernel/syscalls.c

@@ -107,11 +107,11 @@ long ppc64_personality(unsigned long personality)
 	long ret;
 
 	if (personality(current->personality) == PER_LINUX32
-	    && personality == PER_LINUX)
-		personality = PER_LINUX32;
+	    && personality(personality) == PER_LINUX)
+		personality |= PER_LINUX32;
 	ret = sys_personality(personality);
-	if (ret == PER_LINUX32)
-		ret = PER_LINUX;
+	if (personality(ret) == PER_LINUX32)
+		ret &= ~PER_LINUX32;
 	return ret;
 }
 #endif

-- 
1.7.3.1

-- 
Jiri Kosina
SUSE Labs

On Thu, 2012-08-02 at 09:10 +0200, Jiri Kosina wrote:

Directly comparing current->personality against PER_LINUX32 doesn't work
in cases when any of the personality flags stored in the top three bytes
are used.

Directly forcefully setting personality to PER_LINUX32 or PER_LINUX
discards any flags stored in the top three bytes

Use personality() macro to compare only PER_MASK bytes and make sure that
we are setting only the bits that should be set, instead of
overwriting the whole value.

Signed-off-by: Jiri Kosina <redacted>
---

changed since v1: fix the bit ops to reflect the fact that PER_LINUX is 
actually 0

Had already merged v1 (oops.. didn't spot the issue with PER_LINUX being
0). Can you send an incremental fixup ?

Cheers,
Ben.

On Wed, 5 Sep 2012, Benjamin Herrenschmidt wrote:

quoted

Directly comparing current->personality against PER_LINUX32 doesn't work
in cases when any of the personality flags stored in the top three bytes
are used.

Directly forcefully setting personality to PER_LINUX32 or PER_LINUX
discards any flags stored in the top three bytes

Use personality() macro to compare only PER_MASK bytes and make sure that
we are setting only the bits that should be set, instead of
overwriting the whole value.

Signed-off-by: Jiri Kosina <redacted>
---

changed since v1: fix the bit ops to reflect the fact that PER_LINUX is 
actually 0

Had already merged v1 (oops.. didn't spot the issue with PER_LINUX being
0). Can you send an incremental fixup ?

Hi Benjamin,

actually commit 7256a5d2da56 seems to contain the correct PER_LINUX 
handling, so seems like you picked the right one :)

Thanks,

-- 
Jiri Kosina
SUSE Labs

On Wed, 2012-09-05 at 10:56 +0200, Jiri Kosina wrote:

Hi Benjamin,

actually commit 7256a5d2da56 seems to contain the correct PER_LINUX 
handling, so seems like you picked the right one :)

Odd, they looked different around the use of PER_MASK when I looked but
I was tired & jet lagged, so I might have just had a brain fail...

Cheers,
Ben.

On Thu, 6 Sep 2012, Benjamin Herrenschmidt wrote:

quoted

actually commit 7256a5d2da56 seems to contain the correct PER_LINUX 
handling, so seems like you picked the right one :)

Odd, they looked different around the use of PER_MASK when I looked but

The original patch had

	personality &= ~PER_LINUX | PER_LINUX32;

Which is bogus, exactly because ~PER_LINUX is -1.

I then used

	personality = (personality & ~PER_MASK) | PER_LINUX32;

which is correct and perhaps a little bit more descriptive, and that is 
what you have merged, so all is fine.

I was tired & jet lagged, so I might have just had a brain fail...

Probably just missed that the first patch used PER_LINUX and the second 
one PER_MASK, or whatever.

Anyway, thanks.

-- 
Jiri Kosina
SUSE Labs