| Message ID | 154701504098.26726.7180667249484886220.stgit@noble (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | assorted osc cleanups. | expand |
On Jan 8, 2019, at 23:24, NeilBrown <neilb@suse.com> wrote: > > assert_spin_locked() is preferred to > spin_is_locked() for affirming that a > spinlock is locked. > > Signed-off-by: NeilBrown <neilb@suse.com> We used to have an LASSERT_SPIN_LOCKED() macro, but it was removed a few years ago. It is nice to get better checking in the kernel. One question inline below: > --- > drivers/staging/lustre/lustre/osc/osc_cache.c | 29 +++++++++----------- > .../staging/lustre/lustre/osc/osc_cl_internal.h | 15 +--------- > 2 files changed, 15 insertions(+), 29 deletions(-) > > diff --git a/drivers/staging/lustre/lustre/osc/osc_cache.c b/drivers/staging/lustre/lustre/osc/osc_cache.c > index fbf16547003d..1ce9f673f1bf 100644 > --- a/drivers/staging/lustre/lustre/osc/osc_cache.c > +++ b/drivers/staging/lustre/lustre/osc/osc_cache.c > @@ -181,10 +181,7 @@ static int osc_extent_sanity_check0(struct osc_extent *ext, > size_t page_count; > int rc = 0; > > - if (!osc_object_is_locked(obj)) { > - rc = 9; > - goto out; > - } > + assert_osc_object_is_locked(obj); Is this actually a fatal error? It looks like if the object is not locked then the checking isn't consistent and could just be skipped? There is a macro that lends credence to this: #define sanity_check_nolock(ext) \ osc_extent_sanity_check0(ext, __func__, __LINE__) #define sanity_check(ext) ({ \ int __res; \ osc_object_lock((ext)->oe_obj); \ __res = sanity_check_nolock(ext); \ osc_object_unlock((ext)->oe_obj); \ __res; \ }) However, reading deeper into the code sanity_check_nolock() looks like is only ever called when the object is already locked, so indeed it does seem like a valid change that should be described in the commit message like: __osc_extent_sanity_check() is only ever called with obj already locked, so change the check into an assertion. It might also be an improvement to rename sanity_check{,_nolock}() to osc_extent_sanity_check() and osc_extent_sanity_check_nolock(), and use __osc_extent_sanity_check() instead of ...0() (which is not standard)? I was going to suggest making sanity_check() an inline function, but that would break the __func__ and __LINE__ expansion and isn't onerous. Reviewed-by: Andreas Dilger <adilger@whamcloud.com> > > if (ext->oe_state >= OES_STATE_MAX) { > rc = 10; > @@ -324,7 +321,7 @@ static int osc_extent_is_overlapped(struct osc_object *obj, > { > struct osc_extent *tmp; > > - LASSERT(osc_object_is_locked(obj)); > + assert_osc_object_is_locked(obj); > > if (!extent_debug) > return 0; > @@ -341,7 +338,7 @@ static int osc_extent_is_overlapped(struct osc_object *obj, > > static void osc_extent_state_set(struct osc_extent *ext, int state) > { > - LASSERT(osc_object_is_locked(ext->oe_obj)); > + assert_osc_object_is_locked(ext->oe_obj); > LASSERT(state >= OES_INV && state < OES_STATE_MAX); > > /* Never try to sanity check a state changing extent :-) */ > @@ -414,7 +411,7 @@ static void osc_extent_put(const struct lu_env *env, struct osc_extent *ext) > static void osc_extent_put_trust(struct osc_extent *ext) > { > LASSERT(atomic_read(&ext->oe_refc) > 1); > - LASSERT(osc_object_is_locked(ext->oe_obj)); > + assert_osc_object_is_locked(ext->oe_obj); > atomic_dec(&ext->oe_refc); > } > > @@ -428,7 +425,7 @@ static struct osc_extent *osc_extent_search(struct osc_object *obj, > struct rb_node *n = obj->oo_root.rb_node; > struct osc_extent *tmp, *p = NULL; > > - LASSERT(osc_object_is_locked(obj)); > + assert_osc_object_is_locked(obj); > while (n) { > tmp = rb_extent(n); > if (index < tmp->oe_start) { > @@ -467,7 +464,7 @@ static void osc_extent_insert(struct osc_object *obj, struct osc_extent *ext) > > LASSERT(RB_EMPTY_NODE(&ext->oe_node)); > LASSERT(ext->oe_obj == obj); > - LASSERT(osc_object_is_locked(obj)); > + assert_osc_object_is_locked(obj); > while (*n) { > tmp = rb_extent(*n); > parent = *n; > @@ -489,7 +486,7 @@ static void osc_extent_erase(struct osc_extent *ext) > { > struct osc_object *obj = ext->oe_obj; > > - LASSERT(osc_object_is_locked(obj)); > + assert_osc_object_is_locked(obj); > if (!RB_EMPTY_NODE(&ext->oe_node)) { > rb_erase(&ext->oe_node, &obj->oo_root); > RB_CLEAR_NODE(&ext->oe_node); > @@ -502,7 +499,7 @@ static struct osc_extent *osc_extent_hold(struct osc_extent *ext) > { > struct osc_object *obj = ext->oe_obj; > > - LASSERT(osc_object_is_locked(obj)); > + assert_osc_object_is_locked(obj); > LASSERT(ext->oe_state == OES_ACTIVE || ext->oe_state == OES_CACHE); > if (ext->oe_state == OES_CACHE) { > osc_extent_state_set(ext, OES_ACTIVE); > @@ -515,7 +512,7 @@ static struct osc_extent *osc_extent_hold(struct osc_extent *ext) > > static void __osc_extent_remove(struct osc_extent *ext) > { > - LASSERT(osc_object_is_locked(ext->oe_obj)); > + assert_osc_object_is_locked(ext->oe_obj); > LASSERT(list_empty(&ext->oe_pages)); > osc_extent_erase(ext); > list_del_init(&ext->oe_link); > @@ -546,7 +543,7 @@ static int osc_extent_merge(const struct lu_env *env, struct osc_extent *cur, > int ppc_bits; > > LASSERT(cur->oe_state == OES_CACHE); > - LASSERT(osc_object_is_locked(obj)); > + assert_osc_object_is_locked(obj); > if (!victim) > return -EINVAL; > > @@ -2079,7 +2076,7 @@ static unsigned int get_write_extents(struct osc_object *obj, > .erd_max_extents = 256, > }; > > - LASSERT(osc_object_is_locked(obj)); > + assert_osc_object_is_locked(obj); > while (!list_empty(&obj->oo_hp_exts)) { > ext = list_entry(obj->oo_hp_exts.next, struct osc_extent, > oe_link); > @@ -2146,7 +2143,7 @@ osc_send_write_rpc(const struct lu_env *env, struct client_obd *cli, > int srvlock = 0; > int rc = 0; > > - LASSERT(osc_object_is_locked(osc)); > + assert_osc_object_is_locked(osc); > > page_count = get_write_extents(osc, &rpclist); > LASSERT(equi(page_count == 0, list_empty(&rpclist))); > @@ -2224,7 +2221,7 @@ osc_send_read_rpc(const struct lu_env *env, struct client_obd *cli, > }; > int rc = 0; > > - LASSERT(osc_object_is_locked(osc)); > + assert_osc_object_is_locked(osc); > list_for_each_entry_safe(ext, next, &osc->oo_reading_exts, oe_link) { > EASSERT(ext->oe_state == OES_LOCK_DONE, ext); > if (!try_to_add_extent_for_io(cli, ext, &data)) > diff --git a/drivers/staging/lustre/lustre/osc/osc_cl_internal.h b/drivers/staging/lustre/lustre/osc/osc_cl_internal.h > index b78deef3963a..aa1b753fc88d 100644 > --- a/drivers/staging/lustre/lustre/osc/osc_cl_internal.h > +++ b/drivers/staging/lustre/lustre/osc/osc_cl_internal.h > @@ -183,19 +183,8 @@ static inline void osc_object_unlock(struct osc_object *obj) > spin_unlock(&obj->oo_lock); > } > > -static inline int osc_object_is_locked(struct osc_object *obj) > -{ > -#if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK) > - return spin_is_locked(&obj->oo_lock); > -#else > - /* > - * It is not perfect to return true all the time. > - * But since this function is only used for assertion > - * and checking, it seems OK. > - */ > - return 1; > -#endif > -} > +#define assert_osc_object_is_locked(obj) \ > + assert_spin_locked(&obj->oo_lock) > > /* > * Lock "micro-states" for osc layer. > > Cheers, Andreas --- Andreas Dilger CTO Whamcloud
On Thu, Jan 10 2019, Andreas Dilger wrote: > On Jan 8, 2019, at 23:24, NeilBrown <neilb@suse.com> wrote: >> >> assert_spin_locked() is preferred to >> spin_is_locked() for affirming that a >> spinlock is locked. >> >> Signed-off-by: NeilBrown <neilb@suse.com> > > We used to have an LASSERT_SPIN_LOCKED() macro, but it was removed a few > years ago. It is nice to get better checking in the kernel. > > One question inline below: > >> --- >> drivers/staging/lustre/lustre/osc/osc_cache.c | 29 +++++++++----------- >> .../staging/lustre/lustre/osc/osc_cl_internal.h | 15 +--------- >> 2 files changed, 15 insertions(+), 29 deletions(-) >> >> diff --git a/drivers/staging/lustre/lustre/osc/osc_cache.c b/drivers/staging/lustre/lustre/osc/osc_cache.c >> index fbf16547003d..1ce9f673f1bf 100644 >> --- a/drivers/staging/lustre/lustre/osc/osc_cache.c >> +++ b/drivers/staging/lustre/lustre/osc/osc_cache.c >> @@ -181,10 +181,7 @@ static int osc_extent_sanity_check0(struct osc_extent *ext, >> size_t page_count; >> int rc = 0; >> >> - if (!osc_object_is_locked(obj)) { >> - rc = 9; >> - goto out; >> - } >> + assert_osc_object_is_locked(obj); > > Is this actually a fatal error? It looks like if the object is not > locked then the checking isn't consistent and could just be skipped? The assert is fatal (-> BUG_ON()). > > There is a macro that lends credence to this: > > #define sanity_check_nolock(ext) \ > osc_extent_sanity_check0(ext, __func__, __LINE__) > > #define sanity_check(ext) ({ \ > int __res; \ > osc_object_lock((ext)->oe_obj); \ > __res = sanity_check_nolock(ext); \ > osc_object_unlock((ext)->oe_obj); \ > __res; \ > }) > > However, reading deeper into the code sanity_check_nolock() looks > like is only ever called when the object is already locked, so > indeed it does seem like a valid change that should be described > in the commit message like: Agreed. > > __osc_extent_sanity_check() is only ever called with obj > already locked, so change the check into an assertion. Good suggestion - I've added this text. > > It might also be an improvement to rename sanity_check{,_nolock}() > to osc_extent_sanity_check() and osc_extent_sanity_check_nolock(), > and use __osc_extent_sanity_check() instead of ...0() (which is not > standard)? I was going to suggest making sanity_check() an inline > function, but that would break the __func__ and __LINE__ expansion > and isn't onerous. That renaming makes sense - I'll add it as another patch. > > Reviewed-by: Andreas Dilger <adilger@whamcloud.com> Thanks, NeilBrown
diff --git a/drivers/staging/lustre/lustre/osc/osc_cache.c b/drivers/staging/lustre/lustre/osc/osc_cache.c index fbf16547003d..1ce9f673f1bf 100644 --- a/drivers/staging/lustre/lustre/osc/osc_cache.c +++ b/drivers/staging/lustre/lustre/osc/osc_cache.c @@ -181,10 +181,7 @@ static int osc_extent_sanity_check0(struct osc_extent *ext, size_t page_count; int rc = 0; - if (!osc_object_is_locked(obj)) { - rc = 9; - goto out; - } + assert_osc_object_is_locked(obj); if (ext->oe_state >= OES_STATE_MAX) { rc = 10; @@ -324,7 +321,7 @@ static int osc_extent_is_overlapped(struct osc_object *obj, { struct osc_extent *tmp; - LASSERT(osc_object_is_locked(obj)); + assert_osc_object_is_locked(obj); if (!extent_debug) return 0; @@ -341,7 +338,7 @@ static int osc_extent_is_overlapped(struct osc_object *obj, static void osc_extent_state_set(struct osc_extent *ext, int state) { - LASSERT(osc_object_is_locked(ext->oe_obj)); + assert_osc_object_is_locked(ext->oe_obj); LASSERT(state >= OES_INV && state < OES_STATE_MAX); /* Never try to sanity check a state changing extent :-) */ @@ -414,7 +411,7 @@ static void osc_extent_put(const struct lu_env *env, struct osc_extent *ext) static void osc_extent_put_trust(struct osc_extent *ext) { LASSERT(atomic_read(&ext->oe_refc) > 1); - LASSERT(osc_object_is_locked(ext->oe_obj)); + assert_osc_object_is_locked(ext->oe_obj); atomic_dec(&ext->oe_refc); } @@ -428,7 +425,7 @@ static struct osc_extent *osc_extent_search(struct osc_object *obj, struct rb_node *n = obj->oo_root.rb_node; struct osc_extent *tmp, *p = NULL; - LASSERT(osc_object_is_locked(obj)); + assert_osc_object_is_locked(obj); while (n) { tmp = rb_extent(n); if (index < tmp->oe_start) { @@ -467,7 +464,7 @@ static void osc_extent_insert(struct osc_object *obj, struct osc_extent *ext) LASSERT(RB_EMPTY_NODE(&ext->oe_node)); LASSERT(ext->oe_obj == obj); - LASSERT(osc_object_is_locked(obj)); + assert_osc_object_is_locked(obj); while (*n) { tmp = rb_extent(*n); parent = *n; @@ -489,7 +486,7 @@ static void osc_extent_erase(struct osc_extent *ext) { struct osc_object *obj = ext->oe_obj; - LASSERT(osc_object_is_locked(obj)); + assert_osc_object_is_locked(obj); if (!RB_EMPTY_NODE(&ext->oe_node)) { rb_erase(&ext->oe_node, &obj->oo_root); RB_CLEAR_NODE(&ext->oe_node); @@ -502,7 +499,7 @@ static struct osc_extent *osc_extent_hold(struct osc_extent *ext) { struct osc_object *obj = ext->oe_obj; - LASSERT(osc_object_is_locked(obj)); + assert_osc_object_is_locked(obj); LASSERT(ext->oe_state == OES_ACTIVE || ext->oe_state == OES_CACHE); if (ext->oe_state == OES_CACHE) { osc_extent_state_set(ext, OES_ACTIVE); @@ -515,7 +512,7 @@ static struct osc_extent *osc_extent_hold(struct osc_extent *ext) static void __osc_extent_remove(struct osc_extent *ext) { - LASSERT(osc_object_is_locked(ext->oe_obj)); + assert_osc_object_is_locked(ext->oe_obj); LASSERT(list_empty(&ext->oe_pages)); osc_extent_erase(ext); list_del_init(&ext->oe_link); @@ -546,7 +543,7 @@ static int osc_extent_merge(const struct lu_env *env, struct osc_extent *cur, int ppc_bits; LASSERT(cur->oe_state == OES_CACHE); - LASSERT(osc_object_is_locked(obj)); + assert_osc_object_is_locked(obj); if (!victim) return -EINVAL; @@ -2079,7 +2076,7 @@ static unsigned int get_write_extents(struct osc_object *obj, .erd_max_extents = 256, }; - LASSERT(osc_object_is_locked(obj)); + assert_osc_object_is_locked(obj); while (!list_empty(&obj->oo_hp_exts)) { ext = list_entry(obj->oo_hp_exts.next, struct osc_extent, oe_link); @@ -2146,7 +2143,7 @@ osc_send_write_rpc(const struct lu_env *env, struct client_obd *cli, int srvlock = 0; int rc = 0; - LASSERT(osc_object_is_locked(osc)); + assert_osc_object_is_locked(osc); page_count = get_write_extents(osc, &rpclist); LASSERT(equi(page_count == 0, list_empty(&rpclist))); @@ -2224,7 +2221,7 @@ osc_send_read_rpc(const struct lu_env *env, struct client_obd *cli, }; int rc = 0; - LASSERT(osc_object_is_locked(osc)); + assert_osc_object_is_locked(osc); list_for_each_entry_safe(ext, next, &osc->oo_reading_exts, oe_link) { EASSERT(ext->oe_state == OES_LOCK_DONE, ext); if (!try_to_add_extent_for_io(cli, ext, &data)) diff --git a/drivers/staging/lustre/lustre/osc/osc_cl_internal.h b/drivers/staging/lustre/lustre/osc/osc_cl_internal.h index b78deef3963a..aa1b753fc88d 100644 --- a/drivers/staging/lustre/lustre/osc/osc_cl_internal.h +++ b/drivers/staging/lustre/lustre/osc/osc_cl_internal.h @@ -183,19 +183,8 @@ static inline void osc_object_unlock(struct osc_object *obj) spin_unlock(&obj->oo_lock); } -static inline int osc_object_is_locked(struct osc_object *obj) -{ -#if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK) - return spin_is_locked(&obj->oo_lock); -#else - /* - * It is not perfect to return true all the time. - * But since this function is only used for assertion - * and checking, it seems OK. - */ - return 1; -#endif -} +#define assert_osc_object_is_locked(obj) \ + assert_spin_locked(&obj->oo_lock) /* * Lock "micro-states" for osc layer.
assert_spin_locked() is preferred to spin_is_locked() for affirming that a spinlock is locked. Signed-off-by: NeilBrown <neilb@suse.com> --- drivers/staging/lustre/lustre/osc/osc_cache.c | 29 +++++++++----------- .../staging/lustre/lustre/osc/osc_cl_internal.h | 15 +--------- 2 files changed, 15 insertions(+), 29 deletions(-)