| Message ID | 201308051051060248738@gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Mon, 5 Aug 2013, majianpeng wrote: > As the 'ceph help' print, it will print pgs.But now it can't. > So we add this.There are two type name of pg, temp and stable. > Because the command 'ceph pg dump' print temp name,so we also print the > temp name of pg. Unfortunately we can't modify the ioctl structure like this without breaking the kernel/userspace ABI (a big no-no). The usual way around that is to make a new ioctl (like CEPH_IOC_GET_DATALOC2) and a new structure (struct ceph_ioctl_dataloc2) with the additional fields. What is this going to be used for? You can get the full info from the montior with the 'ceph osd map <pool name> <object name> --format=json'. Does it need to come from the kernel this way? Assuming we do want to go down this path, I would change a few things: > > Signed-off-by: Jianpeng Ma <majianpeng@gmail.com> > --- > fs/ceph/ioctl.c | 10 +++++++++- > fs/ceph/ioctl.h | 1 + > 2 files changed, 10 insertions(+), 1 deletion(-) > > diff --git a/fs/ceph/ioctl.c b/fs/ceph/ioctl.c > index e0b4ef3..1d0f24c 100644 > --- a/fs/ceph/ioctl.c > +++ b/fs/ceph/ioctl.c > @@ -209,9 +209,17 @@ static long ceph_ioctl_get_dataloc(struct file *file, void __user *arg) > snprintf(dl.object_name, sizeof(dl.object_name), "%llx.%08llx", > ceph_ino(inode), dl.object_no); > > + r = ceph_calc_ceph_temp_pg(&pgid, dl.object_name, osdc->osdmap, > + ceph_file_layout_pg_pool(ci->i_layout)); > + if (r < 0) { > + down_read(&osdc->map_sem); > + return r; > + } > + snprintf(dl.pg_name, sizeof(dl.pg_name), "%llx.%08x", > + pgid.pool, pgid.seed); The pg name formatting is "%lld.%x", if we want this to match with the pg names used throughout the rest of the code. But it sounds like you're actually after the raw hash seed, though, in which case I would drop the string and instead do > @@ -58,6 +58,7 @@ struct ceph_ioctl_dataloc { > __u64 object_no; /* out: object # */ > __u64 object_size; /* out: object size */ > char object_name[64]; /* out: object name */ > + char pg_name[32]; /* out: pg name*/ __u64 pg_raw_seed; /* out: placement seed */ __u64 pg_num; /* out: pg number */ __u64 pg_pool; /* out: pool id */ so that you get the raw value (which e.g. goes in the request) as well as which pg that maps to and the pool id. sage > __u64 block_offset; /* out: offset in block */ > __u64 block_size; /* out: block length */ > __s64 osd; /* out: osd # */ > -- > 1.8.3.rc1.44.gb387c77 > > > Thanks! > Jianpeng MaN????y????b?????v?????{.n??????z??ay????????j???f????????????????:+v??????????zZ+??????"?!? -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
>On Mon, 5 Aug 2013, majianpeng wrote: >> As the 'ceph help' print, it will print pgs.But now it can't. >> So we add this.There are two type name of pg, temp and stable. >> Because the command 'ceph pg dump' print temp name,so we also print the >> temp name of pg. > >Unfortunately we can't modify the ioctl structure like this without >breaking the kernel/userspace ABI (a big no-no). The usual way around >that is to make a new ioctl (like CEPH_IOC_GET_DATALOC2) and a new >structure (struct ceph_ioctl_dataloc2) with the additional fields. > >What is this going to be used for? You can get the full info from the >montior with the 'ceph osd map <pool name> <object name> --format=json'. >Does it need to come from the kernel this way? > I only want to print the pgid when print other infos using cephfs map or sho_location. I'll dig into the implementation of 'ceph osd map <pool name> <object name> --format=json'. Thanks! Jianpeng Ma
diff --git a/fs/ceph/ioctl.c b/fs/ceph/ioctl.c index e0b4ef3..1d0f24c 100644 --- a/fs/ceph/ioctl.c +++ b/fs/ceph/ioctl.c @@ -209,9 +209,17 @@ static long ceph_ioctl_get_dataloc(struct file *file, void __user *arg) snprintf(dl.object_name, sizeof(dl.object_name), "%llx.%08llx", ceph_ino(inode), dl.object_no); + r = ceph_calc_ceph_temp_pg(&pgid, dl.object_name, osdc->osdmap, + ceph_file_layout_pg_pool(ci->i_layout)); + if (r < 0) { + down_read(&osdc->map_sem); + return r; + } + snprintf(dl.pg_name, sizeof(dl.pg_name), "%llx.%08x", + pgid.pool, pgid.seed); + ceph_calc_ceph_pg(&pgid, dl.object_name, osdc->osdmap, ceph_file_layout_pg_pool(ci->i_layout)); - dl.osd = ceph_calc_pg_primary(osdc->osdmap, pgid); if (dl.osd >= 0) { struct ceph_entity_addr *a = diff --git a/fs/ceph/ioctl.h b/fs/ceph/ioctl.h index c77028a..618eef2 100644 --- a/fs/ceph/ioctl.h +++ b/fs/ceph/ioctl.h @@ -58,6 +58,7 @@ struct ceph_ioctl_dataloc { __u64 object_no; /* out: object # */ __u64 object_size; /* out: object size */ char object_name[64]; /* out: object name */ + char pg_name[32]; /* out: pg name*/ __u64 block_offset; /* out: offset in block */ __u64 block_size; /* out: block length */ __s64 osd; /* out: osd # */
As the 'ceph help' print, it will print pgs.But now it can't. So we add this.There are two type name of pg, temp and stable. Because the command 'ceph pg dump' print temp name,so we also print the temp name of pg. Signed-off-by: Jianpeng Ma <majianpeng@gmail.com> --- fs/ceph/ioctl.c | 10 +++++++++- fs/ceph/ioctl.h | 1 + 2 files changed, 10 insertions(+), 1 deletion(-) -- 1.8.3.rc1.44.gb387c77 Thanks! Jianpeng Ma