| Message ID | 20191009234951.2850-1-npiggin@gmail.com |
|---|---|
| Headers | show |
| Series | little endian skiboot | expand |
On Wed, Oct 9, 2019, at 4:49 PM, Nicholas Piggin wrote: > A few reasons for this: > - The rest of the software is moving (moved) to LE. BE is still > supported, but the fact is LE will be better tested from now on in > terms of toolchain. That alone is a pretty solid reason. I wonder how on top of this other firmware components are? Hostboot could probably save a *lot* of boot time and thrashing if it switched. > XICS, XICS-on-XIVE, KVM, VAS, and niche features have not been tested > yet, but at this point it's complete enough that it's easy to work on > and improve so I think it is ready for proper submission now. i.e. all the things there's no op-test test for? :) > Since v3: > - Several fixes to cvc code, including a stack trashing bug that would > sometimes result in skiboots that failed container verification, > depending on where the compiler put things. that looks like it would have been a fun one to find
Stewart Smith's on October 11, 2019 1:19 am: > On Wed, Oct 9, 2019, at 4:49 PM, Nicholas Piggin wrote: >> A few reasons for this: >> - The rest of the software is moving (moved) to LE. BE is still >> supported, but the fact is LE will be better tested from now on in >> terms of toolchain. > > That alone is a pretty solid reason. I wonder how on top of this other firmware components are? Hostboot could probably save a *lot* of boot time and thrashing if it switched. Good question, I don't know I'm still pretty hostbootphobic. The image size shrinks pretty significantly too. Less problem for skiboot because compressed doesn't change much, but if HB is memory constrained early on this might help. text data bss dec hex filename 1471553 273880 48424 1793857 1b5f41 skiboot.elf 1283853 199328 48424 1531605 175ed5 skiboot.elf.le > >> XICS, XICS-on-XIVE, KVM, VAS, and niche features have not been tested >> yet, but at this point it's complete enough that it's easy to work on >> and improve so I think it is ready for proper submission now. > > i.e. all the things there's no op-test test for? :) Yeah, well it was more just get just one real system to boot, then hope other people can pick up the rest of the pieces :) > >> Since v3: >> - Several fixes to cvc code, including a stack trashing bug that would >> sometimes result in skiboots that failed container verification, >> depending on where the compiler put things. > > that looks like it would have been a fun one to find Well at least it wasn't checkstopping, so not _quite_ so fun as the r13 clobber in this code you found with the virtual memory patches! Thanks, Nick