Fix VDSO gettimeofday() when called with NULL struct timeval.
Will Schmidt [Fri, 29 Jun 2007 20:49:50 +0000 (15:49 -0500)]
The vdso64 portion of patch 74609f4536f2b8fd6a48381bbbe3cd37da20a527 for
fixing problems with NULL gettimeofday input mistakenly checks for a
null tz field twice, when it should be checking for null tz once, and
null tv once; by way of a r10/r11 typo.

Any application calling gettimeofday(&tv,NULL) will "fail".

This corrects that typo, and makes my G5 happy.

Tested on G5.

Signed-off-by: Will Schmidt <will_schmidt@vnet.ibm.com>
Cc: Tony Breeds <tony@bakeyournoodle.com>
Forwarded-by: Ben Herrenschmidt <benh@kernel.crashing.org>

[ Ben says: "I checked the 32 bits part of the change is correct.  You
  can probably blame me for originally writing the 2 versions with
  inversed usage of r10 and r11, thus confusing Tony :-)"

  Ben duly blamed.   - Linus ]

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arch/powerpc/kernel/vdso64/gettimeofday.S

index 2d7a510..c6401f9 100644 (file)
@@ -31,7 +31,7 @@ V_FUNCTION_BEGIN(__kernel_gettimeofday)
        mr      r11,r3                  /* r11 holds tv */
        mr      r10,r4                  /* r10 holds tz */
        bl      V_LOCAL_FUNC(__get_datapage)    /* get data page */
-       cmpldi  r10,0                   /* check if tv is NULL */
+       cmpldi  r11,0                   /* check if tv is NULL */
        beq     2f
        bl      V_LOCAL_FUNC(__do_get_xsec)     /* get xsec from tb & kernel */
        lis     r7,15                   /* r7 = 1000000 = USEC_PER_SEC */