Re: [BUG] x86info fails on 2.6.34 ?

[Eric Dumazet <eric.dumazet@gmail.com>]

Le lundi 05 avril 2010 à 22:24 +0200, Eric Dumazet a écrit :
> Hi Dave
> 
> I just noticed x86info was not working anymore on latest kernel
> 
> x86info v1.24.  Dave Jones 2001-2009
> Feedback to <davej@redhat.com>.
> 
> type seek: Value too large for defined data type
> 
> 
> Offending syscall is an lseek(fd, -1, SEEK_CUR) on /dev/mem file
> 
> 
> 
> open("/dev/mem", O_RDONLY)              = 3
> ...
> lseek(3, 1023092, SEEK_SET)             = 1023092
> read(3, "PCMP\4\3\4\266HP      PROLIANT    \0\0\0\0\0"..., 44) = 44
> read(3, "\0"..., 1)                     = 1
> lseek(3, -1, SEEK_CUR)                  = -1 EOVERFLOW (Value too large
> for defined data type)
> 
> It seems -1 is taken as an unsigned quantity ?
> 

Andrew, commit dcefafb6ac90ece8d68a6c203105f3d313e52da4 seems to be the
problem. 

It breaks rewinds (negative offsets to lseek (... SEEK_CUR))


commit dcefafb6ac90ece8d68a6c203105f3d313e52da4
Author: Wu Fengguang <fengguang.wu@intel.com>
Date:   Wed Mar 10 15:21:51 2010 -0800

    /dev/mem: dont allow seek to last page
    
    So as to return a uniform error -EOVERFLOW instead of a random one:
    
    # kmem-seek 0xfffffffffffffff0
    seek /dev/kmem: Device or resource busy
    # kmem-seek 0xfffffffffffffff1
    seek /dev/kmem: Block device required
    
    Suggested by OGAWA Hirofumi.
    
    Cc: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
    Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
    Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
    Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>