On Wed, Sep 06, 2023 at 07:48:14PM +0200, Sabrina Dubroca wrote: > The networking code has multiple users of async crypto (tls, ipsec, > macsec), but networking developers often don't have access to crypto > accelerators, which limits the testing we can do on the async > completion codepaths. > > I've been using this patch to uncover some bugs [1] in the tls > implementation, forcing AESNI (which I do have on my laptop and test > machines) to go through cryptd. > > The intent with the cryptd_delay_ms knob was to try to trigger some > issues with async completions and closing tls sockets. > > Link: https://lore.kernel.org/netdev/cover.1694018970.git.sd@xxxxxxxxxxxxxxx/ [1] > Signed-off-by: Sabrina Dubroca <sd@xxxxxxxxxxxxxxx> Any reason why pcrypt can't be used for the same purpose? Thanks, -- Email: Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx> Home Page: http://gondor.apana.org.au/~herbert/ PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
