On Thu, 7 Mar 2019 14:25:00 -0500 "Jason J. Herne" <[email protected]> wrote:
> On 3/4/19 1:25 PM, Cornelia Huck wrote: > > On Fri, 1 Mar 2019 13:59:30 -0500 > > "Jason J. Herne" <[email protected]> wrote: > >> + consume_io_int(); > >> + > >> + /* collect status */ > >> + rc = tsch(schid, irb); > >> + if (rc) { > >> + print_int("tsch failed with rc=", rc); > >> + } > > > > Hm. The whole code flow relies on the fact that not only no more than > > one cpu is enabled for I/O interrupts, but also only one subchannel. > > Otherwise, you could get an interrupt for another subchannel, which > > would be the only way you'd get cc 1 on the tsch for this subchannel > > here (no status pending). Maybe peek at the interruption information > > stored into the lowcore first? > > > > Won't be a problem with the code as it is now, though, AFAICS. > > > Agreed, voting to leave as is. Perhaps a comment to explain that we rely on > only one > "Active" i/o device? Yes, sounds good.
