Skip to contentSkip to conversation

minor mmap bug ?

DISCUSSION John Levon 1 participant 1 message Oct 4, 2001

lore
Share & tools
lore

John Levon

author
#0 lore raw

2.4.10

In do_mmap_pgoff() :

    422         addr = get_unmapped_area(file, addr, len, pgoff, flags);
    423         if (addr & ~PAGE_MASK)
    424                 return addr;

in get_unmapped_area() :

    621         if (flags & MAP_FIXED) {
    622                 if (addr > TASK_SIZE - len)
    623                         return -EINVAL;
    624                 if (addr & ~PAGE_MASK)
    625                         return -EINVAL;
    626                 return addr;
    627         }

So it seems that if we hit the first case (addr > TASK_SIZE - len), we don't return with EINVAL immediately. This code is a little weird anyway.

Is this a bug ?

Also, what's the reason for :

    405         if ((len = PAGE_ALIGN(len)) == 0)
    406                 return addr;

Also, shouldn't calc_vm_flags() mention MAP_LOCKED and the others specifically ? The current "oh, it happens to be the same value" is distinctly non-obvious to me.

Also, why do several (all ?) of the do_mmap2()s do this :

     51         flags &= ~(MAP_EXECUTABLE | MAP_DENYWRITE);

Is it something to do with mm->def_flags ?

thanks
john

Signature · 4 lines
" It is quite humbling to realize that the storage occupied by the longest line
from a typical Usenet posting is sufficient to provide a state space so vast
that all the computation power in the world can not conquer it."
	- Dave Wallace