NFSD: Rework encoding and decoding of nfsd4_deviceid
Subsystems (3)
PATCH 1/1 NFSD: Rework encoding and decoding of nfsd4_deviceid
Compilers may optimize the layout of C structures, so we should not rely on sizeof and memcpy to encode and decode XDR structures. The byte order of the fields should also be taken into account. This patch adds the correct functions to handle the nfsd4_deviceid structure and removes the pad field, which is currently unused.
---
Tested on the block layout setup, checked with smatch.
fs/nfsd/blocklayoutxdr.c | 7 ++----- fs/nfsd/flexfilelayoutxdr.c | 3 +-- fs/nfsd/nfs4layouts.c | 1 - fs/nfsd/nfs4xdr.c | 14 +------------- fs/nfsd/xdr4.h | 31 ++++++++++++++++++++++++++++++- 5 files changed, 34 insertions(+), 22 deletions(-)
fs/nfsd/blocklayoutxdr.c +2 -5
@@ -29,8 +29,7 @@ nfsd4_block_encode_layoutget(struct xdr_stream *xdr, *p++ =cpu_to_be32 (len); *p++ =cpu_to_be32 (1 );/* we always return a single extent */ - p =xdr_encode_opaque_fixed (p, &b->vol_id, -sizeof (struct nfsd4_deviceid )); + p =nfsd4_encode_deviceid (p, &b->vol_id); p =xdr_encode_hyper (p, b->foff); p =xdr_encode_hyper (p, b->len); p =xdr_encode_hyper (p, b->soff);
@@ -156,9 +155,7 @@ nfsd4_block_decode_layoutupdate(__be32 *p, u32 len, struct iomap **iomapp,for (i =0 ; i < nr_iomaps; i++) {struct pnfs_block_extent bex; -memcpy (&bex.vol_id, p,sizeof (struct nfsd4_deviceid )); - p +=XDR_QUADLEN (sizeof (struct nfsd4_deviceid )); - + p =nfsd4_decode_deviceid (p, &bex.vol_id); p =xdr_decode_hyper (p, &bex.foff);if (bex.foff & (block_size -1 )) { goto fail;
fs/nfsd/flexfilelayoutxdr.c +1 -2
@@ -54,8 +54,7 @@ nfsd4_ff_encode_layoutget(struct xdr_stream *xdr, *p++ =cpu_to_be32 (1 );/* single mirror */ *p++ =cpu_to_be32 (1 );/* single data server */ - p =xdr_encode_opaque_fixed (p, &fl->deviceid, -sizeof (struct nfsd4_deviceid )); + p =nfsd4_encode_deviceid (p, &fl->deviceid); *p++ =cpu_to_be32 (1 );/* efficiency */
fs/nfsd/nfs4layouts.c +0 -1
@@ -120,7 +120,6 @@ nfsd4_set_deviceid(struct nfsd4_deviceid *id, const struct svc_fh *fhp, id->fsid_idx = fhp->fh_export->ex_devid_map->idx; id->generation = device_generation; - id->pad =0 ;return 0 ; }
fs/nfsd/nfs4xdr.c +1 -13
@@ -587,18 +587,6 @@ nfsd4_decode_state_owner4(struct nfsd4_compoundargs *argp, }#ifdef CONFIG_NFSD_PNFS -static __be32 -nfsd4_decode_deviceid4 (struct nfsd4_compoundargs *argp, -struct nfsd4_deviceid *devid) -{ -__be32 *p; - - p =xdr_inline_decode (argp->xdr,NFS4_DEVICEID4_SIZE ); -if (!p) -return nfserr_bad_xdr; -memcpy (devid, p,sizeof (*devid)); -return nfs_ok; -} static __be32 nfsd4_decode_layoutupdate4(struct nfsd4_compoundargs *argp,
@@ -1783,7 +1771,7 @@ nfsd4_decode_getdeviceinfo(struct nfsd4_compoundargs *argp,__be32 status;memset (gdev,0 ,sizeof (*gdev)); - status =nfsd4_decode_deviceid4 (argp, &gdev->gd_devid); + status =nfsd4_stream_decode_deviceid (argp->xdr, &gdev->gd_devid);if (status)return status;if (xdr_stream_decode_u32 (argp->xdr, &gdev->gd_layout_type) <0 )
fs/nfsd/xdr4.h +30 -1
@@ -595,9 +595,38 @@ struct nfsd4_reclaim_complete {struct nfsd4_deviceid {u64 fsid_idx;u32 generation; -u32 pad; }; +static inline __be32 * +nfsd4_encode_deviceid (__be32 *p,const struct nfsd4_deviceid *devid) +{ + p =xdr_encode_hyper (p, devid->fsid_idx); + *p++ =cpu_to_be32 (devid->generation); + *p++ =cpu_to_be32 (0 );/* pad field is currently unused */ +return p; +} + +static inline __be32 * +nfsd4_decode_deviceid (__be32 *p,struct nfsd4_deviceid *devid) +{ + p =xdr_decode_hyper (p, &devid->fsid_idx); + devid->generation =be32_to_cpup (p++); + p++;/* pad field is currently unused */ +return p; +} + +static inline __be32 +nfsd4_stream_decode_deviceid (struct xdr_stream *xdr, +struct nfsd4_deviceid *devid) +{ +__be32 *p =xdr_inline_decode (xdr,NFS4_DEVICEID4_SIZE ); + +if (unlikely (!p)) +return nfserr_bad_xdr; +nfsd4_decode_deviceid (p, devid); +return nfs_ok; +} +struct nfsd4_layout_seg {u32 iomode;u64 offset;
Signature · 1 line
2.43.0
Quoted 7 lines · Sergey Bashirov
Compilers may optimize the layout of C structures, so we should not rely on sizeof and memcpy to encode and decode XDR structures. The byte order of the fields should also be taken into account. This patch adds the correct functions to handle the
nfsd4_deviceidstructure and removes the pad field, which is currently unused.Signed-off-by Sergey Bashirov <[email protected]>
Very thorough. At first I didn't think we would need to update the other layout types, but yes, I now agree that's prudent to do.
A few nits below.
Quoted 98 lines
---
Tested on the block layout setup, checked with smatch.fs/nfsd/blocklayoutxdr.c | 7 ++----- fs/nfsd/flexfilelayoutxdr.c | 3 +-- fs/nfsd/nfs4layouts.c | 1 - fs/nfsd/nfs4xdr.c | 14 +------------- fs/nfsd/xdr4.h | 31 ++++++++++++++++++++++++++++++- 5 files changed, 34 insertions(+), 22 deletions(-)fs/nfsd/blocklayoutxdr.c +2
-5@@ -29,8 +29,7 @@ nfsd4_block_encode_layoutget(struct xdr_stream *xdr, *p++ =cpu_to_be32 (len); *p++ =cpu_to_be32 (1 );/* we always return a single extent */ - p =xdr_encode_opaque_fixed (p, &b->vol_id, -sizeof (struct nfsd4_deviceid )); + p =nfsd4_encode_deviceid (p, &b->vol_id); p =xdr_encode_hyper (p, b->foff); p =xdr_encode_hyper (p, b->len); p =xdr_encode_hyper (p, b->soff);@@ -156,9 +155,7 @@ nfsd4_block_decode_layoutupdate(__be32 *p, u32 len, struct iomap **iomapp,for (i =0 ; i < nr_iomaps; i++) {struct pnfs_block_extent bex; -memcpy (&bex.vol_id, p,sizeof (struct nfsd4_deviceid )); - p +=XDR_QUADLEN (sizeof (struct nfsd4_deviceid )); - + p =nfsd4_decode_deviceid (p, &bex.vol_id); p =xdr_decode_hyper (p, &bex.foff);if (bex.foff & (block_size -1 )) { goto fail;fs/nfsd/flexfilelayoutxdr.c +1
-2@@ -54,8 +54,7 @@ nfsd4_ff_encode_layoutget(struct xdr_stream *xdr, *p++ =cpu_to_be32 (1 );/* single mirror */ *p++ =cpu_to_be32 (1 );/* single data server */ - p =xdr_encode_opaque_fixed (p, &fl->deviceid, -sizeof (struct nfsd4_deviceid )); + p =nfsd4_encode_deviceid (p, &fl->deviceid); *p++ =cpu_to_be32 (1 );/* efficiency */ fs/nfsd/nfs4layouts.c +0
-1@@ -120,7 +120,6 @@ nfsd4_set_deviceid(struct nfsd4_deviceid *id, const struct svc_fh *fhp, id->fsid_idx = fhp->fh_export->ex_devid_map->idx; id->generation = device_generation; - id->pad =0 ;return 0 ; }fs/nfsd/nfs4xdr.c +1
-13@@ -587,18 +587,6 @@ nfsd4_decode_state_owner4(struct nfsd4_compoundargs *argp, }#ifdef CONFIG_NFSD_PNFS -static __be32 -nfsd4_decode_deviceid4 (struct nfsd4_compoundargs *argp, -struct nfsd4_deviceid *devid) -{ -__be32 *p; - - p =xdr_inline_decode (argp->xdr,NFS4_DEVICEID4_SIZE ); -if (!p) -return nfserr_bad_xdr; -memcpy (devid, p,sizeof (*devid)); -return nfs_ok; -} static __be32 nfsd4_decode_layoutupdate4(struct nfsd4_compoundargs *argp,@@ -1783,7 +1771,7 @@ nfsd4_decode_getdeviceinfo(struct nfsd4_compoundargs *argp,__be32 status;memset (gdev,0 ,sizeof (*gdev)); - status =nfsd4_decode_deviceid4 (argp, &gdev->gd_devid); + status =nfsd4_stream_decode_deviceid (argp->xdr, &gdev->gd_devid);if (status)return status;if (xdr_stream_decode_u32 (argp->xdr, &gdev->gd_layout_type) <0 )fs/nfsd/xdr4.h +0
-0
fs/nfsd/xdr4.h struct nfsd4_reclaim_complete { · line 601
Quoted 8 lines
fs/nfsd/xdr4.h +2
-1@@ -595,9 +595,38 @@ struct nfsd4_reclaim_complete {struct nfsd4_deviceid {u64 fsid_idx;u32 generation; -u32 pad; }; +static inline __be32 * +nfsd4_encode_deviceid (__be32 *p,const struct nfsd4_deviceid *devid)
Let's call this one svcxdr_encode_deviceid4()
(Note the "4" on the end, which matches the name of the XDR type as it is specified in RFC 8881).
fs/nfsd/xdr4.h struct nfsd4_reclaim_complete { · line 607
Quoted 6 lines
fs/nfsd/xdr4.h +6
-0+{ + p =xdr_encode_hyper (p, devid->fsid_idx); + *p++ =cpu_to_be32 (devid->generation); + *p++ =cpu_to_be32 (0 );/* pad field is currently unused */ +return p; +}
The byte swap operations are unnecessary here. The deviceid4 blob on the wire is not used by clients in any way other than as a cookie. So the content of the blob can remain in the same endian-ness as the server.
What we do want is to document that fact. The usual way to do that is with "(__force xxx)". Maybe we want to do something clever with a union type, but here's something quick, dirty, and untested:
{
__be64 *q = (__be64)p;
*q = (__force __be64)devid->fsid_idx;
*p += 2;
*p++ (__force __be32)devid->generation;
*p++ = xdr_zero;
return p;
}For the new "is unused" comment, IMHO it's not adding much value. But if you want to keep it, let's make it active voice so it's clear this is an peculiarity of our implementation, rather than a protocol requirement. Something like:
/* NFSD does not use the remaining octets */
fs/nfsd/xdr4.h struct nfsd4_reclaim_complete { · line 610
Quoted 3 lines
fs/nfsd/xdr4.h +3
-0+ +static inline __be32 * +nfsd4_decode_deviceid (__be32 *p,struct nfsd4_deviceid *devid)
This one should be named svcxdr_decode_deviceid4()
fs/nfsd/xdr4.h struct nfsd4_reclaim_complete { · line 616
Quoted 6 lines
fs/nfsd/xdr4.h +6
-0+{ + p =xdr_decode_hyper (p, &devid->fsid_idx); + devid->generation =be32_to_cpup (p++); + p++;/* pad field is currently unused */ +return p; +}
Likewise, replace the byte-swaps with forced type casts here, and fix up the new comment.
fs/nfsd/xdr4.h struct nfsd4_reclaim_complete { · line 620
Quoted 4 lines
fs/nfsd/xdr4.h +4
-0+ +static inline __be32 +nfsd4_stream_decode_deviceid (struct xdr_stream *xdr, +struct nfsd4_deviceid *devid)
And now this one can be called nfsd4_decode_deviceid4(), which matches the naming scheme of the helpers at the top of fs/nfsd/xdr4.h.
Quoted 12 lines
fs/nfsd/xdr4.h +9
-0+{ +__be32 *p =xdr_inline_decode (xdr,NFS4_DEVICEID4_SIZE ); + +if (unlikely (!p)) +return nfserr_bad_xdr; +nfsd4_decode_deviceid (p, devid); +return nfs_ok; +} +struct nfsd4_layout_seg {u32 iomode;u64 offset;
Signature · 1 line
Chuck Lever