On Sat, Mar 01, 2025 at 03:03:48PM +0800, Herbert Xu wrote: > On Sat, Mar 01, 2025 at 02:36:50PM +0800, Herbert Xu wrote: > > > > I thought this was a lot more complicated and you had some weird > > arbtirary pointer from an unknown source. But if it's just highmem > > I can get rid of the memcpy for you right now. > > So it appears that your highmem usage is coming from zsmalloc. In > that case you don't want virtual addresses at all, you want an SG > list. > > In fact you've already gone through a totally unnecessary copy > in _zs_map_object. Had it simply given us a 2-entry SG list, > the Crypto API can process the data directly with no copies at > all. > > The whole point of SG lists is to deal with memory fragmentation. > When your object is too big to fit in a single page, you need an > SG list to describe it. Forcing virtual addresses just leads to > an unnecessary copy. I have seen the other thread with Sergey, I believe the conclusion is that zsmalloc will be updated to use SG lists, at which point zswap can just pass this as-is to the crypto API, and we don't need any copies in either zsmalloc or zswap. Is this correct? Will this patch series be dropped? > > Chers, > -- > Email: Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx> > Home Page: http://gondor.apana.org.au/~herbert/ > PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
