> On July 23, 2012, 11:12 a.m., Marc Orr wrote: > > src/sim/syscall_emul.hh, line 601 > > <http://reviews.gem5.org/r/1318/diff/1/?file=28137#file28137line601> > > > > Why is this fatal being changed to a warn? I'm not necessarily against > > this, but I thought a lot of the effort and controversy generated by this > > patch was that unsupported ioctl's should cause a fatal. > > Steve Reinhardt wrote: > Good question. I think one thing that happened was that, in the process > of digging into this a little more, I found that, for historical reasons, > ENOTTY really is the right error code to return for an unrecognized ioctl, > even in situations where there's no relationship to ttys at all. It turns > out Vince had already pointed this out yet somehow it didn't sink in for me. > So based on that, I felt better about simply warning and returning ENOTTY. > However, the white list is still very important in avoiding warnings when > ENOTTY is actually the right error code in the sense of "we recognize this > ioctl and ENOTTY is what we really want to return" as opposed to "we're > returning ENOTTY because we don't recognize this ioctl". > > I'm not 100% committed to this though, and I'm willing to change this > back to fatal if people feel that's still appropriate.
I'm sort of indifferent, I'd just like to see the patch committed and it's taken about 100x longer than it should for the size of the change (imho). - Ali ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: http://reviews.gem5.org/r/1318/#review3133 ----------------------------------------------------------- On July 22, 2012, 5:29 p.m., Steve Reinhardt wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > http://reviews.gem5.org/r/1318/ > ----------------------------------------------------------- > > (Updated July 22, 2012, 5:29 p.m.) > > > Review request for Default. > > > Description > ------- > > [Note: this is an updated version of Marc's patch #1187. I realized I hadn't > pushed that, but when I went to test it, it didn't compile for ARM. I ended > up doing some more restructuring in the process of fixing that problem.] > > syscall emulation: Clean up ioctl handling, and implement for x86. > > Enable different whitelists for different OS/arch combinations, > since some use the generic Linux definitions only, and others > use definitions inherited from earlier Unix flavors on those > architectures. > > Also update x86 function pointers so ioctl is no longer > unimplemented on that platform. > > This patch is a revised version of Vince Weaver's earlier patch. > > > Diffs > ----- > > src/arch/alpha/linux/linux.hh UNKNOWN > src/arch/alpha/tru64/tru64.hh UNKNOWN > src/arch/arm/linux/linux.hh UNKNOWN > src/arch/mips/linux/linux.hh UNKNOWN > src/arch/power/linux/linux.hh UNKNOWN > src/arch/sparc/linux/linux.hh UNKNOWN > src/arch/x86/linux/syscalls.cc UNKNOWN > src/kern/linux/linux.hh UNKNOWN > src/sim/syscall_emul.hh UNKNOWN > > Diff: http://reviews.gem5.org/r/1318/diff/ > > > Testing > ------- > > passes 'util/regress quick' > > > Thanks, > > Steve Reinhardt > > _______________________________________________ gem5-dev mailing list [email protected] http://m5sim.org/mailman/listinfo/gem5-dev
