Re: Reducing the chunk header sizes on all memory context types

Hi,

On 2022-08-29 11:43:14 -0400, Tom Lane wrote:
> Tomas Vondra <tomas.vondra@enterprisedb.com> writes:
> > I suspect it's a pre-existing bug in Slab allocator, because it does this:
> 
> > #define SlabBlockGetChunk(slab, block, idx) \
> >     ((MemoryChunk *) ((char *) (block) + sizeof(SlabBlock)    \
> >                     + (idx * slab->fullChunkSize)))
> 
> > and SlabBlock is only 20B, i.e. not a multiple of 8B. Which would mean
> > that even if we allocate block and size the chunks carefully (with all
> > the MAXALIGN things), we ultimately slice the block incorrectly.
> 
> Right, same conclusion I just came to.  But it's not a "pre-existing"
> bug, because sizeof(SlabBlock) *was* maxaligned until David added
> another field to it.
> 
> I think adding a padding field to SlabBlock would be a less messy
> solution than your patch.

That just seems to invite the same problem happening again later and it's
harder to ensure that the padding is correct across platforms.

Greetings,

Andres Freund