From: Andy Adamson <[email protected]>
Only resending NFSv4.1 add nfs_inode book keeping for mdsthreshold.
Responded to Trond's comments.
I have not tested the directio nor the memmap io cases. The regular file
i/o works as intended.
Andy Adamson (4):
NFSv4.1 mdsthreshold attribute xdr
NFSv4.1 cache mdsthreshold values on OPEN
NFSv4.1 add nfs_inode book keeping for mdsthreshold
NFSv4.1 test the mdsthreshold hint parameters
fs/nfs/direct.c | 2 +
fs/nfs/file.c | 1 +
fs/nfs/inode.c | 4 ++
fs/nfs/nfs4proc.c | 38 ++++++++++++--
fs/nfs/nfs4xdr.c | 125 ++++++++++++++++++++++++++++++++++++++++++++++-
fs/nfs/pnfs.c | 94 +++++++++++++++++++++++++++++++++++
fs/nfs/pnfs.h | 21 ++++++++
fs/nfs/read.c | 2 +
include/linux/nfs4.h | 7 +++
include/linux/nfs_fs.h | 4 ++
include/linux/nfs_xdr.h | 10 ++++
11 files changed, 301 insertions(+), 7 deletions(-)
--
1.7.7.6
From: Andy Adamson <[email protected]>
Keep track of the number of bytes read or written via buffered, direct, and
mem-mapped i/o for use by mdsthreshold size_io hints.
Signed-off-by: Andy Adamson <[email protected]>
---
fs/nfs/direct.c | 2 ++
fs/nfs/file.c | 1 +
fs/nfs/inode.c | 2 ++
fs/nfs/pnfs.c | 3 +++
fs/nfs/read.c | 2 ++
include/linux/nfs_fs.h | 3 +++
6 files changed, 13 insertions(+), 0 deletions(-)
diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c
index c47a46e..23d170b 100644
--- a/fs/nfs/direct.c
+++ b/fs/nfs/direct.c
@@ -447,6 +447,7 @@ static ssize_t nfs_direct_read(struct kiocb *iocb, const struct iovec *iov,
result = nfs_direct_read_schedule_iovec(dreq, iov, nr_segs, pos);
if (!result)
result = nfs_direct_wait(dreq);
+ NFS_I(inode)->read_io += result;
out_release:
nfs_direct_req_release(dreq);
out:
@@ -785,6 +786,7 @@ static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq,
pos += vec->iov_len;
}
nfs_pageio_complete(&desc);
+ NFS_I(dreq->inode)->write_io += desc.pg_bytes_written;
/*
* If no bytes were started, return the error, and let the
diff --git a/fs/nfs/file.c b/fs/nfs/file.c
index 8eda8a6..56311ca 100644
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@ -424,6 +424,7 @@ static int nfs_write_end(struct file *file, struct address_space *mapping,
if (status < 0)
return status;
+ NFS_I(mapping->host)->write_io += copied;
return copied;
}
diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
index 889f7e5..a6f5fbb 100644
--- a/fs/nfs/inode.c
+++ b/fs/nfs/inode.c
@@ -323,6 +323,8 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
inode->i_gid = -2;
inode->i_blocks = 0;
memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
+ nfsi->write_io = 0;
+ nfsi->read_io = 0;
nfsi->read_cache_jiffies = fattr->time_start;
nfsi->attr_gencount = fattr->gencount;
diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index cbcb6ae..6620606 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -395,6 +395,9 @@ mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo,
dprintk("%s:Begin lo %p\n", __func__, lo);
if (list_empty(&lo->plh_segs)) {
+ /* Reset MDS Threshold I/O counters */
+ NFS_I(lo->plh_inode)->write_io = 0;
+ NFS_I(lo->plh_inode)->read_io = 0;
if (!test_and_set_bit(NFS_LAYOUT_DESTROYED, &lo->plh_flags))
put_layout_hdr_locked(lo);
return 0;
diff --git a/fs/nfs/read.c b/fs/nfs/read.c
index 2cfdd77..86ced78 100644
--- a/fs/nfs/read.c
+++ b/fs/nfs/read.c
@@ -152,6 +152,7 @@ int nfs_readpage_async(struct nfs_open_context *ctx, struct inode *inode,
nfs_pageio_init_read(&pgio, inode, &nfs_async_read_completion_ops);
nfs_pageio_add_request(&pgio, new);
nfs_pageio_complete(&pgio);
+ NFS_I(inode)->read_io += pgio.pg_bytes_written;
return 0;
}
@@ -656,6 +657,7 @@ int nfs_readpages(struct file *filp, struct address_space *mapping,
ret = read_cache_pages(mapping, pages, readpage_async_filler, &desc);
nfs_pageio_complete(&pgio);
+ NFS_I(inode)->read_io += pgio.pg_bytes_written;
npages = (pgio.pg_bytes_written + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
nfs_add_stats(inode, NFSIOS_READPAGES, npages);
read_complete:
diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
index ca4a707..9d44860 100644
--- a/include/linux/nfs_fs.h
+++ b/include/linux/nfs_fs.h
@@ -202,6 +202,9 @@ struct nfs_inode {
/* pNFS layout information */
struct pnfs_layout_hdr *layout;
#endif /* CONFIG_NFS_V4*/
+ /* how many bytes have been written/read and how many bytes queued up */
+ __u64 write_io;
+ __u64 read_io;
#ifdef CONFIG_NFS_FSCACHE
struct fscache_cookie *fscache;
#endif
--
1.7.7.6
On May 23, 2012, at 2:19 PM, Myklebust, Trond wrote:
> On Wed, 2012-05-23 at 05:02 -0400, [email protected] wrote:
>> From: Andy Adamson <[email protected]>
>>
>> Keep track of the number of bytes read or written, including those queued
>> up to be flushed. For use by mdsthreshold i/o size hints.
>>
>> No locking needed as this is used as hint information.
>>
>> Signed-off-by: Andy Adamson <[email protected]>
>> ---
>> fs/nfs/file.c | 8 ++++++--
>> fs/nfs/inode.c | 2 ++
>> fs/nfs/pnfs.c | 3 +++
>> include/linux/nfs_fs.h | 3 +++
>> 4 files changed, 14 insertions(+), 2 deletions(-)
>>
>> diff --git a/fs/nfs/file.c b/fs/nfs/file.c
>> index 8eda8a6..c4cc096 100644
>> --- a/fs/nfs/file.c
>> +++ b/fs/nfs/file.c
>> @@ -203,8 +203,10 @@ nfs_file_read(struct kiocb *iocb, const struct iovec *iov,
>> result = nfs_revalidate_mapping(inode, iocb->ki_filp->f_mapping);
>> if (!result) {
>> result = generic_file_aio_read(iocb, iov, nr_segs, pos);
>> - if (result > 0)
>> + if (result > 0) {
>> + NFS_I(inode)->read_io += result;
>
> Should we perhaps rather do this from nfs_readpages(), nfs_readpage()
> and nfs_direct_read()?
>
> If we do it here in nfs_file_read, we miss mmaped reads, O_DIRECT reads,
> as well as splice reads.
Well that's not good.
> We also count read cache hits where we don't
> have to actually access the server.
OK.
>
>> nfs_add_stats(inode, NFSIOS_NORMALREADBYTES, result);
>> + }
>> }
>> return result;
>> }
>> @@ -613,8 +615,10 @@ static ssize_t nfs_file_write(struct kiocb *iocb, const struct iovec *iov,
>> if (err < 0)
>> result = err;
>> }
>> - if (result > 0)
>> + if (result > 0) {
>> + NFS_I(inode)->write_io += written;
>
> For the same reason, perhaps we should move this to
> nfs_direct_write_schedule_iovec(), and nfs_write_end().
>
>> nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, written);
>> + }
>> out:
>> return result;
>>
>> diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
>> index 889f7e5..a6f5fbb 100644
>> --- a/fs/nfs/inode.c
>> +++ b/fs/nfs/inode.c
>> @@ -323,6 +323,8 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
>> inode->i_gid = -2;
>> inode->i_blocks = 0;
>> memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
>> + nfsi->write_io = 0;
>> + nfsi->read_io = 0;
>>
>> nfsi->read_cache_jiffies = fattr->time_start;
>> nfsi->attr_gencount = fattr->gencount;
>> diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
>> index cbcb6ae..6620606 100644
>> --- a/fs/nfs/pnfs.c
>> +++ b/fs/nfs/pnfs.c
>> @@ -395,6 +395,9 @@ mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo,
>> dprintk("%s:Begin lo %p\n", __func__, lo);
>>
>> if (list_empty(&lo->plh_segs)) {
>> + /* Reset MDS Threshold I/O counters */
>> + NFS_I(lo->plh_inode)->write_io = 0;
>> + NFS_I(lo->plh_inode)->read_io = 0;
>> if (!test_and_set_bit(NFS_LAYOUT_DESTROYED, &lo->plh_flags))
>> put_layout_hdr_locked(lo);
>> return 0;
>> diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
>> index ca4a707..c6954ac 100644
>> --- a/include/linux/nfs_fs.h
>> +++ b/include/linux/nfs_fs.h
>> @@ -201,6 +201,9 @@ struct nfs_inode {
>>
>> /* pNFS layout information */
>> struct pnfs_layout_hdr *layout;
>> + /* how many bytes have been written/read and how many bytes queued up */
>> + __u64 write_io;
>> + __u64 read_io;
>> #endif /* CONFIG_NFS_V4*/
>
> ^^^^ This doesn't look as if it will compile without CONFIG_NFS_V4.
I'll fix and resend. Thanks for the review :)
-->Andy
>
>> #ifdef CONFIG_NFS_FSCACHE
>> struct fscache_cookie *fscache;
>
> --
> Trond Myklebust
> Linux NFS client maintainer
>
> NetApp
> [email protected]
> http://www.netapp.com
>