On Tue, Jan 28 2014, Grant Grundler wrote: > On Tue, Jan 28, 2014 at 2:03 PM, Jens Axboe <axboe@xxxxxxxxx> wrote: > ... > >> - o->rand_seed = le32_to_cpu(top->rand_seed); > >> + o->rand_seed = le64_to_cpu(top->rand_seed); > >> > >> I saw this when reviewing the code but didn't realize "top" was the > >> packed version that got changed. > > > > Yeah, and vice versa for top -> o. > > Oh! :( > > (And I assume you mean line 288 as shown below) > > If I add this: > > diff --git a/cconv.c b/cconv.c > index c4941ba..57996fb 100644 > --- a/cconv.c > +++ b/cconv.c > @@ -285,7 +285,7 @@ void convert_thread_options_to_net(struct thread_options_pac > top->do_disk_util = cpu_to_le32(o->do_disk_util); > top->override_sync = cpu_to_le32(o->override_sync); > top->rand_repeatable = cpu_to_le32(o->rand_repeatable); > - top->rand_seed = cpu_to_le32(o->rand_seed); > + top->rand_seed = cpu_to_le64(o->rand_seed); > top->use_os_rand = cpu_to_le32(o->use_os_rand); > top->log_avg_msec = cpu_to_le32(o->log_avg_msec); > top->norandommap = cpu_to_le32(o->norandommap); > > I get this warning: > > grundler <2090>make V=1 > gcc -o cconv.o -std=gnu99 -Wwrite-strings -Wall > -Wdeclaration-after-statement -O3 -g -ffast-math -D_GNU_SOURCE > -include config-host.h -DBITS_PER_LONG=64 > -DFIO_VERSION='"fio-2.1.4-26-g3a2a"' -D_LARGEFILE_SOURCE > -D_FILE_OFFSET_BITS=64 -D_FORTIFY_SOURCE=2 -DFIO_INC_DEBUG -c cconv.c > > CC cconv.o > cconv.c: In function ‘convert_thread_options_to_net’: > cconv.c:288:19: warning: initialization from incompatible pointer type > [enabled by default] > LINK fio > > grundler <2091>file cconv.o > cconv.o: ELF 64-bit LSB relocatable, x86-64, version 1 (SYSV), not stripped > > 286 top->override_sync = cpu_to_le32(o->override_sync); > 287 top->rand_repeatable = cpu_to_le32(o->rand_repeatable); > 288 top->rand_seed = cpu_to_le64(o->rand_seed); > 289 top->use_os_rand = cpu_to_le32(o->use_os_rand); > 290 top->log_avg_msec = cpu_to_le32(o->log_avg_msec); > > Now I'm just confused. It's because the parser is using non uintxx_t types, so you can't really please it here. There's __cpu_to_le64() to avoid this crap. But ideally, the parser just needs changing so that anything that is now unsigned int is uint32_t and unsigned long long needs to be uint64_t. If you look at your committed patch, I made those changes for you. -- Jens Axboe -- To unsubscribe from this list: send the line "unsubscribe fio" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
