[Valgrind-users] custom allocators and redzones

[Nicholas M. L. <nm...@cs...>]

hi all,

haven't managed to find the answer to this in the valgrind headers, in the
manual, faq or by googling, so if theres something simple i'm missing please 
let me know.

i'm writing a custom memory allocator that allocates chunks of memory of the
same size.  It's capable of placing a redzone after each allocated chunk
(so that each chunk has either a redzone before, or the superblock header, and
a redzone afterward).  I'd like to integrate valgrind checking into my
allocator, so that it detects accesses to the allocators structures from
outside of the allocator (by using valgrind to mark the structures invalid when
not in allocator code), detects memory leaks (presumably by using
VALGRIND_MALLOCLIKE_BLOCK etc) and detecting writes into the redzones.
VALGRIND_MALLOCLIKE_BLOCK takes an argument that describes the size of the
red-zone associated with the block.  What i don't understand is where the
red-zone is expected to be relative to the addr argument.  And what does
MALLLOCLIKE_BLOCK do with the redzone info?  Is it going to mark a section of
memory non-accessable?  I've looked at the examples given and i'm still not
quite sure how it operates.

On a related note, what do the _MEMPOOL user requests do?  Are they just used
to associate a bit of memory with a given memory pool address?  Will memory
leaks be reported if i DESTROY a mempool that has outstanding allocations?

thanks in advance

Nick