* [lustre-devel] [PATCH] staging: lustre: Move assignments out of function calls. @ 2016-10-21 22:14 Elise Lennion 2016-10-21 22:32 ` Oleg Drokin 2016-10-21 22:34 ` Dilger, Andreas 0 siblings, 2 replies; 4+ messages in thread From: Elise Lennion @ 2016-10-21 22:14 UTC (permalink / raw) To: lustre-devel Assignments inside of function calls confuse the reader and should be avoided, so they were moved out before the call. Found with Coccinelle, semantic patch: @@ identifier f; expression e1, e2; assignment operator a; @@ + e1 a e2; f(..., - (e1 a e2) + e1 ,...); Signed-off-by: Elise Lennion <elise.lennion@gmail.com> --- drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c | 5 ++--- drivers/staging/lustre/lustre/llite/vvp_io.c | 3 ++- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c b/drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c index 12647af..ec7f2a8 100644 --- a/drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c +++ b/drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c @@ -764,9 +764,8 @@ static int ldlm_bl_thread_main(void *arg) if (!blwi) { atomic_dec(&blp->blp_busy_threads); - l_wait_event_exclusive(blp->blp_waitq, - (blwi = ldlm_bl_get_work(blp)), - &lwi); + blwi = ldlm_bl_get_work(blp); + l_wait_event_exclusive(blp->blp_waitq, blwi, &lwi); busy = atomic_inc_return(&blp->blp_busy_threads); } else { busy = atomic_read(&blp->blp_busy_threads); diff --git a/drivers/staging/lustre/lustre/llite/vvp_io.c b/drivers/staging/lustre/lustre/llite/vvp_io.c index 1a4b897..87fdab2 100644 --- a/drivers/staging/lustre/lustre/llite/vvp_io.c +++ b/drivers/staging/lustre/lustre/llite/vvp_io.c @@ -450,7 +450,8 @@ static void vvp_io_advance(const struct lu_env *env, struct vvp_io *vio = cl2vvp_io(env, ios); CLOBINVRNT(env, obj, vvp_object_invariant(obj)); - iov_iter_reexpand(vio->vui_iter, vio->vui_tot_count -= nob); + vio->vui_tot_count -= nob; + iov_iter_reexpand(vio->vui_iter, vio->vui_tot_count); } static void vvp_io_update_iov(const struct lu_env *env, -- 2.7.4 ^ permalink raw reply related [flat|nested] 4+ messages in thread
* [lustre-devel] [PATCH] staging: lustre: Move assignments out of function calls. 2016-10-21 22:14 [lustre-devel] [PATCH] staging: lustre: Move assignments out of function calls Elise Lennion @ 2016-10-21 22:32 ` Oleg Drokin 2016-10-21 22:34 ` Dilger, Andreas 1 sibling, 0 replies; 4+ messages in thread From: Oleg Drokin @ 2016-10-21 22:32 UTC (permalink / raw) To: lustre-devel On Oct 21, 2016, at 6:14 PM, Elise Lennion wrote: > Assignments inside of function calls confuse the reader and should be > avoided, so they were moved out before the call. > > Found with Coccinelle, semantic patch: > @@ > identifier f; > expression e1, e2; > assignment operator a; > @@ > > + e1 a e2; > f(..., > - (e1 a e2) > + e1 > ,...); > > Signed-off-by: Elise Lennion <elise.lennion@gmail.com> > --- > drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c | 5 ++--- > drivers/staging/lustre/lustre/llite/vvp_io.c | 3 ++- > 2 files changed, 4 insertions(+), 4 deletions(-) > > diff --git a/drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c b/drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c > index 12647af..ec7f2a8 100644 > --- a/drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c > +++ b/drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c > @@ -764,9 +764,8 @@ static int ldlm_bl_thread_main(void *arg) > > if (!blwi) { > atomic_dec(&blp->blp_busy_threads); > - l_wait_event_exclusive(blp->blp_waitq, > - (blwi = ldlm_bl_get_work(blp)), > - &lwi); > + blwi = ldlm_bl_get_work(blp); > + l_wait_event_exclusive(blp->blp_waitq, blwi, &lwi); This one is wrong. This is not a function call, this is a macro whee second argument determines when to terminate the waiting. ldlm_bl_get_work returns the next work item, so we need to save it somewhere for later processing. The way you coded it it would just hang in there as blwi becomes a constant. > busy = atomic_inc_return(&blp->blp_busy_threads); > } else { > busy = atomic_read(&blp->blp_busy_threads); > diff --git a/drivers/staging/lustre/lustre/llite/vvp_io.c b/drivers/staging/lustre/lustre/llite/vvp_io.c > index 1a4b897..87fdab2 100644 > --- a/drivers/staging/lustre/lustre/llite/vvp_io.c > +++ b/drivers/staging/lustre/lustre/llite/vvp_io.c > @@ -450,7 +450,8 @@ static void vvp_io_advance(const struct lu_env *env, > struct vvp_io *vio = cl2vvp_io(env, ios); > CLOBINVRNT(env, obj, vvp_object_invariant(obj)); > > - iov_iter_reexpand(vio->vui_iter, vio->vui_tot_count -= nob); > + vio->vui_tot_count -= nob; > + iov_iter_reexpand(vio->vui_iter, vio->vui_tot_count); This one is probably ok. > } > > static void vvp_io_update_iov(const struct lu_env *env, > -- > 2.7.4 ^ permalink raw reply [flat|nested] 4+ messages in thread
* [lustre-devel] [PATCH] staging: lustre: Move assignments out of function calls. 2016-10-21 22:14 [lustre-devel] [PATCH] staging: lustre: Move assignments out of function calls Elise Lennion 2016-10-21 22:32 ` Oleg Drokin @ 2016-10-21 22:34 ` Dilger, Andreas 2016-10-23 0:21 ` Elise Lennion 1 sibling, 1 reply; 4+ messages in thread From: Dilger, Andreas @ 2016-10-21 22:34 UTC (permalink / raw) To: lustre-devel On Oct 21, 2016, at 16:14, Elise Lennion <elise.lennion@gmail.com> wrote: > > Assignments inside of function calls confuse the reader and should be > avoided, so they were moved out before the call. Thanks for your patch. I agree that this kind of usage is bad and avoid it in my own code and try to remove it when I see it. However, in some cases it is needed to simplify the code flow (e.g. in conditionals that would otherwise need to duplicate logic). More comments below. > Found with Coccinelle, semantic patch: > @@ > identifier f; > expression e1, e2; > assignment operator a; > @@ > > + e1 a e2; > f(..., > - (e1 a e2) > + e1 > ,...); > > Signed-off-by: Elise Lennion <elise.lennion@gmail.com> > --- > drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c | 5 ++--- > drivers/staging/lustre/lustre/llite/vvp_io.c | 3 ++- > 2 files changed, 4 insertions(+), 4 deletions(-) > > diff --git a/drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c b/drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c > index 12647af..ec7f2a8 100644 > --- a/drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c > +++ b/drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c > @@ -764,9 +764,8 @@ static int ldlm_bl_thread_main(void *arg) > > if (!blwi) { > atomic_dec(&blp->blp_busy_threads); > - l_wait_event_exclusive(blp->blp_waitq, > - (blwi = ldlm_bl_get_work(blp)), > - &lwi); > + blwi = ldlm_bl_get_work(blp); > + l_wait_event_exclusive(blp->blp_waitq, blwi, &lwi); This first change is broken, since l_wait_event_exclusive() is actually a macro that is waiting until ldlm_bl_get_work() returns a work item. Moving the blwi assignment before the wait means that this will wait forever. This can be ignored for now, as there is a patch in the pipeline to clean up this code (http://review.whamcloud.com/5843) that James should be getting close to pushing, and it removes this inline assignment. > busy = atomic_inc_return(&blp->blp_busy_threads); > } else { > busy = atomic_read(&blp->blp_busy_threads); > diff --git a/drivers/staging/lustre/lustre/llite/vvp_io.c b/drivers/staging/lustre/lustre/llite/vvp_io.c > index 1a4b897..87fdab2 100644 > --- a/drivers/staging/lustre/lustre/llite/vvp_io.c > +++ b/drivers/staging/lustre/lustre/llite/vvp_io.c > @@ -450,7 +450,8 @@ static void vvp_io_advance(const struct lu_env *env, > struct vvp_io *vio = cl2vvp_io(env, ios); > CLOBINVRNT(env, obj, vvp_object_invariant(obj)); > > - iov_iter_reexpand(vio->vui_iter, vio->vui_tot_count -= nob); > + vio->vui_tot_count -= nob; > + iov_iter_reexpand(vio->vui_iter, vio->vui_tot_count); This one is fine. Please resubmit the patch with just this change. You can add "Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>" for that patch. Cheers, Andreas > } > > static void vvp_io_update_iov(const struct lu_env *env, > -- > 2.7.4 > ^ permalink raw reply [flat|nested] 4+ messages in thread
* [lustre-devel] [PATCH] staging: lustre: Move assignments out of function calls. 2016-10-21 22:34 ` Dilger, Andreas @ 2016-10-23 0:21 ` Elise Lennion 0 siblings, 0 replies; 4+ messages in thread From: Elise Lennion @ 2016-10-23 0:21 UTC (permalink / raw) To: lustre-devel On Fri, Oct 21, 2016 at 10:34:53PM +0000, Dilger, Andreas wrote: > On Oct 21, 2016, at 16:14, Elise Lennion <elise.lennion@gmail.com> wrote: > > > > Assignments inside of function calls confuse the reader and should be > > avoided, so they were moved out before the call. > > Thanks for your patch. I agree that this kind of usage is bad and avoid > it in my own code and try to remove it when I see it. However, in some > cases it is needed to simplify the code flow (e.g. in conditionals that > would otherwise need to duplicate logic). > > More comments below. > > > Found with Coccinelle, semantic patch: > > @@ > > identifier f; > > expression e1, e2; > > assignment operator a; > > @@ > > > > + e1 a e2; > > f(..., > > - (e1 a e2) > > + e1 > > ,...); > > > > Signed-off-by: Elise Lennion <elise.lennion@gmail.com> > > --- > > drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c | 5 ++--- > > drivers/staging/lustre/lustre/llite/vvp_io.c | 3 ++- > > 2 files changed, 4 insertions(+), 4 deletions(-) > > > > diff --git a/drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c b/drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c > > index 12647af..ec7f2a8 100644 > > --- a/drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c > > +++ b/drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c > > @@ -764,9 +764,8 @@ static int ldlm_bl_thread_main(void *arg) > > > > if (!blwi) { > > atomic_dec(&blp->blp_busy_threads); > > - l_wait_event_exclusive(blp->blp_waitq, > > - (blwi = ldlm_bl_get_work(blp)), > > - &lwi); > > + blwi = ldlm_bl_get_work(blp); > > + l_wait_event_exclusive(blp->blp_waitq, blwi, &lwi); > > This first change is broken, since l_wait_event_exclusive() is actually a > macro that is waiting until ldlm_bl_get_work() returns a work item. Moving > the blwi assignment before the wait means that this will wait forever. > > This can be ignored for now, as there is a patch in the pipeline to clean > up this code (http://review.whamcloud.com/5843) that James should be > getting close to pushing, and it removes this inline assignment. > > > busy = atomic_inc_return(&blp->blp_busy_threads); > > } else { > > busy = atomic_read(&blp->blp_busy_threads); > > diff --git a/drivers/staging/lustre/lustre/llite/vvp_io.c b/drivers/staging/lustre/lustre/llite/vvp_io.c > > index 1a4b897..87fdab2 100644 > > --- a/drivers/staging/lustre/lustre/llite/vvp_io.c > > +++ b/drivers/staging/lustre/lustre/llite/vvp_io.c > > @@ -450,7 +450,8 @@ static void vvp_io_advance(const struct lu_env *env, > > struct vvp_io *vio = cl2vvp_io(env, ios); > > CLOBINVRNT(env, obj, vvp_object_invariant(obj)); > > > > - iov_iter_reexpand(vio->vui_iter, vio->vui_tot_count -= nob); > > + vio->vui_tot_count -= nob; > > + iov_iter_reexpand(vio->vui_iter, vio->vui_tot_count); > > This one is fine. Please resubmit the patch with just this change. You > can add "Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>" for that > patch. > > Cheers, Andreas > > > } > > > > static void vvp_io_update_iov(const struct lu_env *env, > > -- > > 2.7.4 > > > Sorry for the confusion with the macro usage, I'll send a v2 without this undesired change. thank you all for the detailed answer, elise ^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2016-10-23 0:21 UTC | newest] Thread overview: 4+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2016-10-21 22:14 [lustre-devel] [PATCH] staging: lustre: Move assignments out of function calls Elise Lennion 2016-10-21 22:32 ` Oleg Drokin 2016-10-21 22:34 ` Dilger, Andreas 2016-10-23 0:21 ` Elise Lennion
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox