On Mon, Feb 22, 2021 at 5:53 PM Randy Dunlap <rdunlap@xxxxxxxxxxxxx> wrote: > On 2/22/21 8:46 AM, Jason Gunthorpe wrote: > > > There is nearly 1000 places that use a 'select CRYPTO_*' in the > > kernel. > > > > I see only 60 'select CRYPTO' statements. I think most of these are correct, as you typically have a single 'select CRYPTO' in combination with a couple of 'select CRYPTO_*' ones for the specific algorithms. I just added a lot of 'select CRC32' statements to deal with all the build failures in drivers that require this but fail to have an extra select statement. The way I got the list was to start with 'make allmodconfig' and then recursively disable everything that had an explicit select, until I was left with all the modules that need it without selecting it. The same method could be used for CONFIG_CRYPTO, but it's a few hours of work. > Correct. :( > We (community) tend to fix bug reports, not do global audits > to see what needs to be fixed (with a few exceptions). > > > I'll gladly wait to see what Arnd says. For the specific case of CRC32, it might actually a good idea to change the code to call into the CRC32 code directly instead of the CRYPTO_CRC32 abstraction. Would that work for RDMA_RXE? Arnd