On Mon Feb 1, 2021 at 11:37 AM CST, David Laight wrote: > From: Christopher M. Riedl > > Sent: 01 February 2021 16:55 > ... > > > > > > + int i; > > > > > > \ > > > > > > + > > > > > > \ > > > > > > + unsafe_copy_from_user(buf, __f, ELF_NFPREG * sizeof(double), > > > > > > \ > > > > > > + label); > > > > > > \ > > > > > > + for (i = 0; i < ELF_NFPREG - 1; i++) > > > > > > \ > > > > > > + __t->thread.TS_FPR(i) = buf[i]; > > > > > > \ > > > > > > + __t->thread.fp_state.fpscr = buf[i]; > > > > > > \ > > > > > > +} while (0) > > > > > > On further reflection, since you immediately loop through the buffer > > > why not just use user_access_begin() and unsafe_get_user() in the loop. > > > > Christophe had suggested this a few revisions ago as well. When I tried > > this approach, the signal handling performance took a pretty big hit: > > https://lists.ozlabs.org/pipermail/linuxppc-dev/2020-October/219351.html > > > > I included some numbers on v3 as well but decided to drop the approach > > altogether for this one since it just didn't seem worth the hit. > > Was that using unsafe_get_user (which relies on user_access_begin() > having 'opened up' user accesses) or just get_user() that does > it for every access? > > The former should be ok, the latter will be horrid.
It was using unsafe_get_user() whereas unsafe_copy_from_user() will just call the optimized __copy_tofrom_user() a single time - assuming that user access is open. > > David > > - > Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, > MK1 1PT, UK > Registration No: 1397386 (Wales)
