NFSv4.2: Various READ_PLUS fixes
Subsystems (3)
Anna Schumaker
maintainerv6 cover NFSv4.2: Various READ_PLUS fixes
From: Anna Schumaker <[email protected]>
These patches are a handul of fixes I've done recently to the READ_PLUS code. This includes fixing some smatch warnings, fixing the XDR reply length calculation, improving scratch buffer handling, and having xdr_inline_decode() kmap pages if we detect that they're HIGHMEM so we don't oops during READ_PLUS xdr decoding.
I also (optimistically) have a patch at the end to enable CONFIG_READ_PLUS by default. My hope right now is that we can enable READ_PLUS for Linux 6.6, and remove it entirely in 6.7 if all goes well.
Thoughts? Patch #4 probably needs the most review, and I'm open to other approaches if something else makes more sense!
Thanks,
Anna
Anna Schumaker (5):
NFSv4.2: Fix READ_PLUS smatch warnings NFSv4.2: Fix READ_PLUS size calculations NFSv4.2: Rework scratch handling for READ_PLUS (again)
SUNRPC: kmap() the xdr pages during decode
NFS: Enable the READ_PLUS operation by default
fs/nfs/Kconfig | 6 ++---- fs/nfs/internal.h | 1 + fs/nfs/nfs42.h | 1 + fs/nfs/nfs42xdr.c | 17 +++++++++++------ fs/nfs/nfs4proc.c | 13 +------------ fs/nfs/read.c | 10 ++++++++++ include/linux/sunrpc/xdr.h | 3 +++ net/sunrpc/clnt.c | 1 + net/sunrpc/svc.c | 2 ++ net/sunrpc/xdr.c | 27 ++++++++++++++++++++++++++- 10 files changed, 58 insertions(+), 23 deletions(-)
Signature · 1 line
2.41.0
Anna Schumaker
maintainerPATCH v6 1/5 NFSv4.2: Fix READ_PLUS smatch warnings
From: Anna Schumaker <[email protected]>
Smatch reports:
fs/nfs/nfs42xdr.c:1131 decode_read_plus() warn: missing error code? 'status'
Which Dan suggests to fix by doing a hardcoded "return 0" from the "if (segments == 0)" check.
Additionally, smatch reports that the "status = -EIO" assignment is not used. This patch addresses both these issues.
--- fs/nfs/nfs42xdr.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)
fs/nfs/nfs42xdr.c +1 -2
@@ -1056,13 +1056,12 @@ static int decode_read_plus(struct xdr_stream *xdr, struct nfs_pgio_res *res) res->eof =be32_to_cpup (p++); segments =be32_to_cpup (p++);if (segments ==0 ) -return status; +return 0 ; segs =kmalloc_array (segments,sizeof (*segs),GFP_KERNEL );if (!segs)return -ENOMEM ; - status = -EIO ;for (i =0 ; i < segments; i++) { status =decode_read_plus_segment (xdr, &segs[i]);if (status <0 )
Signature · 1 line
2.41.0
Anna Schumaker
maintainerPATCH v6 2/5 NFSv4.2: Fix READ_PLUS size calculations
In reply to Anna SchumakerFrom: Anna Schumaker <[email protected]>
I bump the decode_read_plus_maxsz to account for hole segments, but I need to subtract out this increase when calling rpc_prepare_reply_pages() so the common case of single data segment replies can be directly placed into the xdr pages without needing to be shifted around.
--- fs/nfs/nfs42xdr.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-)
fs/nfs/nfs42xdr.c +9 -3
@@ -54,10 +54,16 @@ (1 /* data_content4 */ + \ 2 /* data_info4.di_offset */ + \ 1 /* data_info4.di_length */) +#define NFS42_READ_PLUS_HOLE_SEGMENT_SIZE \ + (1 /* data_content4 */ + \ + 2 /* data_info4.di_offset */ + \ + 2 /* data_info4.di_length */) +#define READ_PLUS_SEGMENT_SIZE_DIFF (NFS42_READ_PLUS_HOLE_SEGMENT_SIZE - \ + NFS42_READ_PLUS_DATA_SEGMENT_SIZE)#define decode_read_plus_maxsz (op_decode_hdr_maxsz + \ 1 /* rpr_eof */ + \ 1 /* rpr_contents count */ + \ - NFS42_READ_PLUS_DATA_SEGMENT_SIZE) + NFS42_READ_PLUS_HOLE_SEGMENT_SIZE) #define encode_seek_maxsz (op_encode_hdr_maxsz + \ encode_stateid_maxsz + \ 2 /* offset */ + \
@@ -617,8 +623,8 @@ static void nfs4_xdr_enc_read_plus(struct rpc_rqst *req,encode_putfh (xdr, args->fh, &hdr);encode_read_plus (xdr, args, &hdr); -rpc_prepare_reply_pages (req, args->pages, args->pgbase, - args->count, hdr.replen); +rpc_prepare_reply_pages (req, args->pages, args->pgbase, args->count, + hdr.replen -READ_PLUS_SEGMENT_SIZE_DIFF );encode_nops (&hdr); }
Signature · 1 line
2.41.0
Anna Schumaker
maintainerPATCH v6 3/5 NFSv4.2: Rework scratch handling for READ_PLUS (again)
In reply to Anna SchumakerFrom: Anna Schumaker <[email protected]>
I found that the read code might send multiple requests using the same nfs_pgio_header, but nfs4_proc_read_setup() is only called once. This is how we ended up occasionally double-freeing the scratch buffer, but also means we set a NULL pointer but non-zero length to the xdr scratch buffer. This results in an oops the first time decoding needs to copy something to scratch, which frequently happens when decoding READ_PLUS hole segments.
I fix this by moving scratch handling into the pageio read code. I provide a function to allocate scratch space for decoding read replies, and free the scratch buffer when the nfs_pgio_header is freed.
--- fs/nfs/internal.h | 1 + fs/nfs/nfs42.h | 1 + fs/nfs/nfs42xdr.c | 2 +- fs/nfs/nfs4proc.c | 13 +------------ fs/nfs/read.c | 10 ++++++++++ 5 files changed, 14 insertions(+), 13 deletions(-)
fs/nfs/internal.h +1 -0
@@ -493,6 +493,7 @@ extern const struct nfs_pgio_completion_ops nfs_async_read_completion_ops;extern void nfs_pageio_init_read (struct nfs_pageio_descriptor *pgio,struct inode *inode,bool force_mds,const struct nfs_pgio_completion_ops *compl_ops); +extern bool nfs_read_alloc_scratch (struct nfs_pgio_header *hdr,size_t size);extern int nfs_read_add_folio (struct nfs_pageio_descriptor *pgio,struct nfs_open_context *ctx,struct folio *folio);
fs/nfs/nfs42.h +1 -0
@@ -13,6 +13,7 @@* more? Need to consider not to pre-alloc too much for a compound. */ #define PNFS_LAYOUTSTATS_MAXDEV (4) +#define READ_PLUS_SCRATCH_SIZE (16)/* nfs4.2proc.c */ #ifdef CONFIG_NFS_V4_2
fs/nfs/nfs42xdr.c +1 -1
@@ -1433,7 +1433,7 @@ static int nfs4_xdr_dec_read_plus(struct rpc_rqst *rqstp,struct compound_hdr hdr;int status; -xdr_set_scratch_buffer (xdr, res->scratch,sizeof (res->scratch)); +xdr_set_scratch_buffer (xdr, res->scratch,READ_PLUS_SCRATCH_SIZE ); status =decode_compound_hdr (xdr, &hdr);if (status)
fs/nfs/nfs4proc.c +1 -12
@@ -5438,18 +5438,8 @@ static bool nfs4_read_plus_not_supported(struct rpc_task *task,return false; } -static inline void nfs4_read_plus_scratch_free (struct nfs_pgio_header *hdr) -{ -if (hdr->res.scratch) { -kfree (hdr->res.scratch); - hdr->res.scratch =NULL ; - } -} -static int nfs4_read_done (struct rpc_task *task,struct nfs_pgio_header *hdr) { -nfs4_read_plus_scratch_free (hdr); -if (!nfs4_sequence_done (task, &hdr->res.seq_res))return -EAGAIN ;if (nfs4_read_stateid_changed (task, &hdr->args))
@@ -5469,8 +5459,7 @@ static bool nfs42_read_plus_support(struct nfs_pgio_header *hdr,/* Note: We don't use READ_PLUS with pNFS yet */ if (nfs_server_capable (hdr->inode,NFS_CAP_READ_PLUS ) && !hdr->ds_clp) { msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS ]; - hdr->res.scratch =kmalloc (32 ,GFP_KERNEL ); -return hdr->res.scratch !=NULL ; +return nfs_read_alloc_scratch (hdr,READ_PLUS_SCRATCH_SIZE ); }return false; }
fs/nfs/read.c +10 -0
@@ -47,6 +47,8 @@ static struct nfs_pgio_header *nfs_readhdr_alloc(void)static void nfs_readhdr_free (struct nfs_pgio_header *rhdr) { +if (rhdr->res.scratch !=NULL ) +kfree (rhdr->res.scratch);kmem_cache_free (nfs_rdata_cachep, rhdr); }
@@ -108,6 +110,14 @@ void nfs_pageio_reset_read_mds(struct nfs_pageio_descriptor *pgio) }EXPORT_SYMBOL_GPL (nfs_pageio_reset_read_mds); +bool nfs_read_alloc_scratch (struct nfs_pgio_header *hdr,size_t size) +{ +WARN_ON (hdr->res.scratch !=NULL ); + hdr->res.scratch =kmalloc (size,GFP_KERNEL ); +return hdr->res.scratch !=NULL ; +} +EXPORT_SYMBOL_GPL (nfs_read_alloc_scratch); +static void nfs_readpage_release (struct nfs_page *req,int error) {struct folio *folio =nfs_page_to_folio (req);
Signature · 1 line
2.41.0
Anna Schumaker
maintainerPATCH v6 4/5 SUNRPC: kmap() the xdr pages during decode
In reply to Anna SchumakerFrom: Anna Schumaker <[email protected]>
If the pages are in HIGHMEM then we need to make sure they're mapped before trying to read data off of them, otherwise we could end up with a NULL pointer dereference.
The downside to this is that we need an extra cleanup step at the end of decode to kunmap() the last page. I introduced an xdr_finish_decode() function to do this. Right now this function only calls the unmap_current_page() function, but other generic cleanup steps could be added in the future if we come across anything else.
--- include/linux/sunrpc/xdr.h | 3 +++ net/sunrpc/clnt.c | 1 + net/sunrpc/svc.c | 2 ++ net/sunrpc/xdr.c | 27 ++++++++++++++++++++++++++- 4 files changed, 32 insertions(+), 1 deletion(-)
include/linux/sunrpc/xdr.h +3 -0
@@ -228,6 +228,7 @@ struct xdr_stream {struct kvec *iov;/* pointer to the current kvec */ struct kvec scratch;/* Scratch buffer */ struct page **page_ptr;/* pointer to the current page */ +void *page_kaddr;/* kmapped address of the current page */ unsigned int nwords;/* Remaining decode buffer length */ struct rpc_rqst *rqst;/* For debugging */
@@ -253,12 +254,14 @@ extern void xdr_truncate_decode(struct xdr_stream *xdr, size_t len);extern int xdr_restrict_buflen (struct xdr_stream *xdr,int newbuflen);extern void xdr_write_pages (struct xdr_stream *xdr,struct page **pages,unsigned int base,unsigned int len); +extern void xdr_stream_unmap_current_page (struct xdr_stream *xdr);extern unsigned int xdr_stream_pos (const struct xdr_stream *xdr);extern unsigned int xdr_page_pos (const struct xdr_stream *xdr);extern void xdr_init_decode (struct xdr_stream *xdr,struct xdr_buf *buf,__be32 *p,struct rpc_rqst *rqst);extern void xdr_init_decode_pages (struct xdr_stream *xdr,struct xdr_buf *buf,struct page **pages,unsigned int len); +extern void xdr_finish_decode (struct xdr_stream *xdr);extern __be32 *xdr_inline_decode (struct xdr_stream *xdr,size_t nbytes);extern unsigned int xdr_read_pages (struct xdr_stream *xdr,unsigned int len);extern void xdr_enter_page (struct xdr_stream *xdr,unsigned int len);
net/sunrpc/clnt.c +1 -0
@@ -2602,6 +2602,7 @@ call_decode(struct rpc_task *task)case 0 : task->tk_action = rpc_exit_task; task->tk_status =rpcauth_unwrap_resp (task, &xdr); +xdr_finish_decode (&xdr);return ;case -EAGAIN : task->tk_status =0 ;
net/sunrpc/svc.c +2 -0
@@ -1370,6 +1370,8 @@ svc_process_common(struct svc_rqst *rqstp) rc = process.dispatch (rqstp);if (procp->pc_release) procp->pc_release (rqstp); +xdr_finish_decode (xdr); +if (!rc) goto dropit;if (rqstp->rq_auth_stat != rpc_auth_ok)
net/sunrpc/xdr.c +26 -1
@@ -1306,6 +1306,14 @@ static unsigned int xdr_set_tail_base(struct xdr_stream *xdr,return xdr_set_iov (xdr, buf->tail, base, len); } +void xdr_stream_unmap_current_page (struct xdr_stream *xdr) +{ +if (xdr->page_kaddr) { +kunmap_local (xdr->page_kaddr); + xdr->page_kaddr =NULL ; + } +} +static unsigned int xdr_set_page_base (struct xdr_stream *xdr,unsigned int base,unsigned int len) {
@@ -1323,12 +1331,18 @@ static unsigned int xdr_set_page_base(struct xdr_stream *xdr,if (len > maxlen) len = maxlen; +xdr_stream_unmap_current_page (xdr);xdr_stream_page_set_pos (xdr, base); base += xdr->buf->page_base; pgnr = base >>PAGE_SHIFT ; xdr->page_ptr = &xdr->buf->pages[pgnr]; - kaddr =page_address (*xdr->page_ptr); + +if (PageHighMem (*xdr->page_ptr)) { + xdr->page_kaddr =kmap_local_page (*xdr->page_ptr); + kaddr = xdr->page_kaddr; + }else + kaddr =page_address (*xdr->page_ptr); pgoff = base & ~PAGE_MASK ; xdr->p = (__be32 *)(kaddr + pgoff);
@@ -1382,6 +1396,7 @@ void xdr_init_decode(struct xdr_stream *xdr, struct xdr_buf *buf, __be32 *p, struct rpc_rqst *rqst) { xdr->buf = buf; + xdr->page_kaddr =NULL ;xdr_reset_scratch_buffer (xdr); xdr->nwords =XDR_QUADLEN (buf->len); if (xdr_set_iov(xdr, buf->head, 0, buf->len) == 0 &&
@@ -1414,6 +1429,16 @@ void xdr_init_decode_pages(struct xdr_stream *xdr, struct xdr_buf *buf, }EXPORT_SYMBOL_GPL (xdr_init_decode_pages); +/** +* xdr_finish_decode - Clean up the xdr_stream after decoding data. +* @xdr: pointer to xdr_stream struct +*/ +void xdr_finish_decode (struct xdr_stream *xdr) +{ +xdr_stream_unmap_current_page (xdr); +} +EXPORT_SYMBOL (xdr_finish_decode); +static __be32 *__xdr_inline_decode (struct xdr_stream *xdr,size_t nbytes) {unsigned int nwords =XDR_QUADLEN (nbytes);
Signature · 1 line
2.41.0
Re: [PATCH v6 4/5] SUNRPC: kmap() the xdr pages during decode
Quoted 34 lines · Anna Schumaker
From: Anna Schumaker <[email protected]>
If the pages are in HIGHMEM then we need to make sure they're mapped before trying to read data off of them, otherwise we could end up with a NULL pointer dereference.
The downside to this is that we need an extra cleanup step at the end of decode to
kunmap()the last page. I introduced anxdr_finish_decode()function to do this. Right now this function only calls theunmap_current_page()function, but other generic cleanup steps could be added in the future if we come across anything else.Reported-by Krzysztof Kozlowski <[email protected]>Signed-off-by Anna Schumaker <[email protected]>--- include/linux/sunrpc/xdr.h | 3 +++ net/sunrpc/clnt.c | 1 + net/sunrpc/svc.c | 2 ++ net/sunrpc/xdr.c | 27 ++++++++++++++++++++++++++- 4 files changed, 32 insertions(+), 1 deletion(-)include/linux/sunrpc/xdr.h +1
-0@@ -228,6 +228,7 @@ struct xdr_stream {struct kvec *iov;/* pointer to the current kvec */ struct kvec scratch;/* Scratch buffer */ struct page **page_ptr;/* pointer to the current page */ +void *page_kaddr;/* kmapped address of the current page */ unsigned int nwords;/* Remaining decode buffer length */ struct rpc_rqst *rqst;/* For debugging */
include/linux/sunrpc/xdr.h extern void xdr_truncate_decode(struct xdr_stream *xdr, size_t len); · line 257
Quoted 4 lines
include/linux/sunrpc/xdr.h +1
-0@@ -253,12 +254,14 @@ extern void xdr_truncate_decode(struct xdr_stream *xdr, size_t len);extern int xdr_restrict_buflen (struct xdr_stream *xdr,int newbuflen);extern void xdr_write_pages (struct xdr_stream *xdr,struct page **pages,unsigned int base,unsigned int len); +extern void xdr_stream_unmap_current_page (struct xdr_stream *xdr);
Nit: I don't think this function needs to be public any more. The usual convention is to make helpers like this static until they are needed somewhere outside the source file where they are defined.
The current definition (below) of xdr_stream_unmap_current_page() has neither an EXPORT_SYMBOL nor a kdoc comment, so seems like an easy change to make it static.
I don't see other issues, so you could just make this change as you apply it rather than posting a v7, but it's your nickel.
Quoted 102 lines
include/linux/sunrpc/xdr.h +1
-0extern unsigned int xdr_stream_pos (const struct xdr_stream *xdr);extern unsigned int xdr_page_pos (const struct xdr_stream *xdr);extern void xdr_init_decode (struct xdr_stream *xdr,struct xdr_buf *buf,__be32 *p,struct rpc_rqst *rqst);extern void xdr_init_decode_pages (struct xdr_stream *xdr,struct xdr_buf *buf,struct page **pages,unsigned int len); +extern void xdr_finish_decode (struct xdr_stream *xdr);extern __be32 *xdr_inline_decode (struct xdr_stream *xdr,size_t nbytes);extern unsigned int xdr_read_pages (struct xdr_stream *xdr,unsigned int len);extern void xdr_enter_page (struct xdr_stream *xdr,unsigned int len);net/sunrpc/clnt.c +1
-0@@ -2602,6 +2602,7 @@ call_decode(struct rpc_task *task)case 0 : task->tk_action = rpc_exit_task; task->tk_status =rpcauth_unwrap_resp (task, &xdr); +xdr_finish_decode (&xdr);return ;case -EAGAIN : task->tk_status =0 ;net/sunrpc/svc.c +2
-0@@ -1370,6 +1370,8 @@ svc_process_common(struct svc_rqst *rqstp) rc = process.dispatch (rqstp);if (procp->pc_release) procp->pc_release (rqstp); +xdr_finish_decode (xdr); +if (!rc) goto dropit;if (rqstp->rq_auth_stat != rpc_auth_ok)net/sunrpc/xdr.c +26
-1@@ -1306,6 +1306,14 @@ static unsigned int xdr_set_tail_base(struct xdr_stream *xdr,return xdr_set_iov (xdr, buf->tail, base, len); } +void xdr_stream_unmap_current_page (struct xdr_stream *xdr) +{ +if (xdr->page_kaddr) { +kunmap_local (xdr->page_kaddr); + xdr->page_kaddr =NULL ; + } +} +static unsigned int xdr_set_page_base (struct xdr_stream *xdr,unsigned int base,unsigned int len) {@@ -1323,12 +1331,18 @@ static unsigned int xdr_set_page_base(struct xdr_stream *xdr,if (len > maxlen) len = maxlen; +xdr_stream_unmap_current_page (xdr);xdr_stream_page_set_pos (xdr, base); base += xdr->buf->page_base; pgnr = base >>PAGE_SHIFT ; xdr->page_ptr = &xdr->buf->pages[pgnr]; - kaddr =page_address (*xdr->page_ptr); + +if (PageHighMem (*xdr->page_ptr)) { + xdr->page_kaddr =kmap_local_page (*xdr->page_ptr); + kaddr = xdr->page_kaddr; + }else + kaddr =page_address (*xdr->page_ptr); pgoff = base & ~PAGE_MASK ; xdr->p = (__be32 *)(kaddr + pgoff);@@ -1382,6 +1396,7 @@ void xdr_init_decode(struct xdr_stream *xdr, struct xdr_buf *buf, __be32 *p, struct rpc_rqst *rqst) { xdr->buf = buf; + xdr->page_kaddr =NULL ;xdr_reset_scratch_buffer (xdr); xdr->nwords =XDR_QUADLEN (buf->len); if (xdr_set_iov(xdr, buf->head, 0, buf->len) == 0 &&@@ -1414,6 +1429,16 @@ void xdr_init_decode_pages(struct xdr_stream *xdr, struct xdr_buf *buf, }EXPORT_SYMBOL_GPL (xdr_init_decode_pages); +/** +* xdr_finish_decode - Clean up the xdr_stream after decoding data. +* @xdr: pointer to xdr_stream struct +*/ +void xdr_finish_decode (struct xdr_stream *xdr) +{ +xdr_stream_unmap_current_page (xdr); +} +EXPORT_SYMBOL (xdr_finish_decode); +static __be32 *__xdr_inline_decode (struct xdr_stream *xdr,size_t nbytes) {unsigned int nwords =XDR_QUADLEN (nbytes);Signature · 1 line
2.41.0
Signature · 1 line
Chuck Lever
Anna Schumaker
maintainerPATCH v6 5/5 NFS: Enable the READ_PLUS operation by default
In reply to Anna SchumakerFrom: Anna Schumaker <[email protected]>
Now that the remaining issues have been worked out, including some unexpected 32 bit issues, we can safely enable the feature by default.
--- fs/nfs/Kconfig | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-)
fs/nfs/Kconfig +2 -4
@@ -209,8 +209,6 @@ config NFS_DISABLE_UDP_SUPPORT config NFS_V4_2_READ_PLUS bool "NFS: Enable support for the NFSv4.2 READ_PLUS operation" depends on NFS_V4_2 - default n + default y help - This is intended for developers only. The READ_PLUS operation has - been shown to have issues under specific conditions and should not - be used in production. + Choose Y here to enable use of the NFS v4.2 READ_PLUS operation.
Signature · 1 line
2.41.0