Hi, > Date: 2026-10-04 17:19:55-0700 > From: Rosen Penev <[email protected]> > > ceph_alloc_options() allocates the monitor address array separately, > always sized for CEPH_MAX_MON entries. Make it a flexible array member > at the end of struct ceph_options and allocate both together with > kzalloc_flex(). The array was already an 8 KiB slab object, so this > only drops the separate allocation and its error path. > > num_mon is the number of addresses in use, not the capacity, so the > array cannot be annotated with __counted_by(). > > ceph_compare_options() memcmp()s the simple fields up to the first > member that needs a custom comparison. With mon_addr gone from there, > use num_mon as that boundary; the compared fields are unchanged. > Document that num_mon has to stay first past that point. > > Assisted-by: LLM > Signed-off-by: Rosen Penev <[email protected]> > --- > include/linux/ceph/libceph.h | 7 ++++--- > net/ceph/ceph_common.c | 11 ++--------- > 2 files changed, 6 insertions(+), 12 deletions(-) > > diff --git a/include/linux/ceph/libceph.h b/include/linux/ceph/libceph.h > index f92fdd853f1f..e585d6d11303 100644 > --- a/include/linux/ceph/libceph.h > +++ b/include/linux/ceph/libceph.h > @@ -57,15 +57,16 @@ struct ceph_options { > /* > * any type that can't be simply compared or doesn't need > * to be compared should go beyond this point, > - * ceph_compare_options() should be updated accordingly > + * ceph_compare_options() should be updated accordingly. > + * num_mon must stay the first member past this point, as > + * ceph_compare_options() uses it as the end of the memcmp(). > */ > > - struct ceph_entity_addr *mon_addr; /* should be the first > - pointer type of args */ > int num_mon; > char *name; > struct ceph_crypto_key *key; > struct rb_root crush_locs; > + struct ceph_entity_addr mon_addr[]; > }; > > /* > diff --git a/net/ceph/ceph_common.c b/net/ceph/ceph_common.c > index a797c7360e3c..95dc18257056 100644 > --- a/net/ceph/ceph_common.c > +++ b/net/ceph/ceph_common.c > @@ -133,7 +133,7 @@ int ceph_compare_options(struct ceph_options *new_opt, > { > struct ceph_options *opt1 = new_opt; > struct ceph_options *opt2 = client->options; > - int ofs = offsetof(struct ceph_options, mon_addr); > + int ofs = offsetof(struct ceph_options, num_mon); > int i; > int ret; > > @@ -309,17 +309,11 @@ struct ceph_options *ceph_alloc_options(void) > { > struct ceph_options *opt; > > - opt = kzalloc_obj(*opt); > + opt = kzalloc_flex(*opt, mon_addr, CEPH_MAX_MON);
Oh, there's indeed a kzalloc_flex(). Why is it not used in the other
patches? (I'm not all that happy with the current API of this _flex()
variant, but it's better than nothing.)
Have a lovely day!
Alex
> if (!opt)
> return NULL;
>
> opt->crush_locs = RB_ROOT;
> - opt->mon_addr = kzalloc_objs(*opt->mon_addr, CEPH_MAX_MON);
> - if (!opt->mon_addr) {
> - kfree(opt);
> - return NULL;
> - }
> -
> opt->flags = CEPH_OPT_DEFAULT;
> opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
> opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT;
> @@ -344,7 +338,6 @@ void ceph_destroy_options(struct ceph_options *opt)
> ceph_crypto_key_destroy(opt->key);
> kfree(opt->key);
> }
> - kfree(opt->mon_addr);
> kfree(opt);
> }
> EXPORT_SYMBOL(ceph_destroy_options);
> --
> 2.56.0
>
>
--
<https://www.alejandro-colomar.es>
signature.asc
Description: PGP signature
