linux-kernel - mm/slab.c: remove effectively dead code from kmem_cache

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Date:	Wed, 08 Feb 2012 22:16:23 -0600
From:	Daniel Santos <danielfsantos@....net>
To:	torvalds@...ux-foundation.org
CC:	linux-kernel@...r.kernel.org
Subject: mm/slab.c: remove effectively dead code from kmem_cache_create

I was examining slab.c when I noticed that there is code that will never
be executed, but that the compiler probably wouldn't determine as such. 
It turns out to be the case.  The below instructions (from a "disas /m 
kmem_cache_create" in gdb) will never be executed (or will have no
effect) since CONFIG_DEBUG_SLAB is not set and line 2301 (BUG_ON(flags &
~CREATE_MASK);) will oops us if we're using the flags in question.

2329            /*
2330             * Redzoning and user store require word alignment or
possibly larger.
2331             * Note this will be overridden by architecture or
caller mandated
2332             * alignment if either is greater than BYTES_PER_WORD.
2333             */
2334            if (flags & SLAB_STORE_USER)
2335                    ralign = BYTES_PER_WORD;
   0x00000000000038ae <+350>:   testq  $0x10000,0x20(%rsp)
   0x00000000000038b7 <+359>:   mov    $0x8,%eax
   0x00000000000038bc <+364>:   cmovne %rax,%r13

2336
2337            if (flags & SLAB_RED_ZONE) {
   0x00000000000038c0 <+368>:   testq  $0x400,0x20(%rsp)
   0x00000000000038c9 <+377>:   jne    0x3ba8 <kmem_cache_create+1112>

2338                    ralign = REDZONE_ALIGN;
   0x0000000000003bae <+1118>:  mov    $0x8,%r13d

2339                    /* If redzoning, ensure that the second redzone
is suitably
2340                     * aligned, by adjusting the object size
accordingly. */
2341                    size += REDZONE_ALIGN - 1;
   0x0000000000003ba8 <+1112>:  addq   $0x7,0x18(%rsp)

2342                    size &= ~(REDZONE_ALIGN - 1);
   0x0000000000003bb4 <+1124>:  andq   $0xfffffffffffffff8,0x18(%rsp)
   0x0000000000003bba <+1130>:  jmpq   0x38d7 <kmem_cache_create+391>
   0x0000000000003bbf <+1135>:  nop

2343            }
2344
2345            /* 2) arch mandated alignment */
2346            if (ralign < ARCH_SLAB_MINALIGN) {
   0x00000000000038d7 <+391>:   cmp    0x28(%rsp),%r13
   0x00000000000038e8 <+408>:   cmovb  0x28(%rsp),%r13

2347                    ralign = ARCH_SLAB_MINALIGN;
   0x00000000000038cf <+383>:   cmp    $0x7,%r13
   0x00000000000038d3 <+387>:   cmovbe %rax,%r13

2348            }
2349            /* 3) caller mandated alignment */
2350            if (ralign < align) {
2351                    ralign = align;
2352            }
2353            /* disable debug if necessary */
2354            if (ralign > __alignof__(unsigned long long))
2355                    flags &= ~(SLAB_RED_ZONE | SLAB_STORE_USER);
   0x00000000000038dc <+396>:   mov    0x20(%rsp),%rax
   0x00000000000038ee <+414>:   and    $0xfffffffffffefbff,%rax
   0x00000000000038f4 <+420>:   cmp    $0x9,%r13
   0x00000000000038f8 <+424>:   cmovb  0x20(%rsp),%rax
   0x0000000000003907 <+439>:   mov    %rax,0x20(%rsp)

2356            /*
2357             * 4) Store it.
2358             */
2359            align = ralign;

There's another little block that I can't illustrate since
CONFIG_PAGE_POISONING doesn't get enabled on my arch, but I've added it
into the patch as well.

Of note, in situations like this where I have a pre-process macro (i.e.,
DEBUG) that's defined to either zero or non-zero, my personal coding
style is to just use it directly in the the if() and let the optomizer
compile it out (as opposed to a #if/#endif block) but I was trying to
copy the coding style already in use.




View attachment "0001-compile-out-effectively-dead-code-from-kmem_cache_cr.patch" of type "text/x-patch" (2176 bytes)