2013-09-18 00:56:49

by Stephen Rothwell

[permalink] [raw]
Subject: linux-next: manual merge of the vfs tree with the aio-direct tree

Hi Al,

Today's linux-next merge of the vfs tree got conflicts in fs/nfs/direct.c
and fs/nfs/file.c between commits b9517433d65d ("dio: Convert direct_IO
to use iov_iter"), a8431c667ae8 ("nfs: add support for read_iter,
write_iter") and a1b8ec384b73 ("nfs: simplify swap") from the aio-direct
tree and commit c18d1ec44f7a ("nfs: use %p[dD] instead of open-coded (and
often racy) equivalents") from the vfs tree.

I fixed it up (see below) and can carry the fix as necessary (no action
is required).

--
Cheers,
Stephen Rothwell [email protected]

diff --cc fs/nfs/direct.c
index 239c2fe,d71d66c..0000000
--- a/fs/nfs/direct.c
+++ b/fs/nfs/direct.c
@@@ -118,18 -117,26 +118,17 @@@ static inline int put_dreq(struct nfs_d
* @nr_segs: size of iovec array
*
* The presence of this routine in the address space ops vector means
- * the NFS client supports direct I/O. However, for most direct IO, we
- * shunt off direct read and write requests before the VFS gets them,
- * so this method is only ever called for swap.
+ * the NFS client supports direct I/O. However, we shunt off direct
+ * read and write requests before the VFS gets them, so this method
+ * should never be called.
*/
-ssize_t nfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, loff_t pos, unsigned long nr_segs)
+ssize_t nfs_direct_IO(int rw, struct kiocb *iocb, struct iov_iter *iter,
+ loff_t pos)
{
- dprintk("NFS: nfs_direct_IO (%s) off/no(%Ld/%lu) EINVAL\n",
- iocb->ki_filp->f_path.dentry->d_name.name,
- (long long) pos, iter->nr_segs);
-#ifndef CONFIG_NFS_SWAP
+ dprintk("NFS: nfs_direct_IO (%pD) off/no(%Ld/%lu) EINVAL\n",
- iocb->ki_filp, (long long) pos, nr_segs);
++ iocb->ki_filp, (long long) pos, iter->nr_segs);

return -EINVAL;
-#else
- VM_BUG_ON(iocb->ki_nbytes != PAGE_SIZE);
-
- if (rw == READ || rw == KERNEL_READ)
- return nfs_file_direct_read(iocb, iov, nr_segs, pos,
- rw == READ ? true : false);
- return nfs_file_direct_write(iocb, iov, nr_segs, pos,
- rw == WRITE ? true : false);
-#endif /* CONFIG_NFS_SWAP */
}

static void nfs_direct_release_pages(struct page **pages, unsigned int npages)
@@@ -1010,13 -905,11 +1009,11 @@@ ssize_t nfs_file_direct_read(struct kio
struct address_space *mapping = file->f_mapping;
size_t count;

- count = iov_length(iov, nr_segs);
+ count = iov_iter_count(iter);
nfs_add_stats(mapping->host, NFSIOS_DIRECTREADBYTES, count);

- dfprintk(FILE, "NFS: direct read(%s/%s, %zd@%Ld)\n",
- file->f_path.dentry->d_parent->d_name.name,
- file->f_path.dentry->d_name.name,
- count, (long long) pos);
+ dfprintk(FILE, "NFS: direct read(%pD2, %zd@%Ld)\n",
+ file, count, (long long) pos);

retval = 0;
if (!count)
@@@ -1065,13 -959,11 +1062,11 @@@ ssize_t nfs_file_direct_write(struct ki
struct address_space *mapping = file->f_mapping;
size_t count;

- count = iov_length(iov, nr_segs);
+ count = iov_iter_count(iter);
nfs_add_stats(mapping->host, NFSIOS_DIRECTWRITTENBYTES, count);

- dfprintk(FILE, "NFS: direct write(%s/%s, %zd@%Ld)\n",
- file->f_path.dentry->d_parent->d_name.name,
- file->f_path.dentry->d_name.name,
- count, (long long) pos);
+ dfprintk(FILE, "NFS: direct write(%pD2, %zd@%Ld)\n",
+ file, count, (long long) pos);

retval = generic_write_checks(file, &pos, &count, 0);
if (retval)
diff --cc fs/nfs/file.c
index 19ac4fd,e2fcacf..0000000
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@@ -174,18 -165,18 +165,17 @@@ nfs_file_flush(struct file *file, fl_ow
EXPORT_SYMBOL_GPL(nfs_file_flush);

ssize_t
-nfs_file_read(struct kiocb *iocb, const struct iovec *iov,
- unsigned long nr_segs, loff_t pos)
+nfs_file_read_iter(struct kiocb *iocb, struct iov_iter *iter, loff_t pos)
{
- struct dentry * dentry = iocb->ki_filp->f_path.dentry;
- struct inode * inode = dentry->d_inode;
+ struct inode *inode = file_inode(iocb->ki_filp);
ssize_t result;

if (iocb->ki_filp->f_flags & O_DIRECT)
- return nfs_file_direct_read(iocb, iov, nr_segs, pos, true);
+ return nfs_file_direct_read(iocb, iter, pos);

- dprintk("NFS: read_iter(%s/%s, %lu@%lu)\n",
- dentry->d_parent->d_name.name, dentry->d_name.name,
- dprintk("NFS: read(%pD2, %lu@%lu)\n",
++ dprintk("NFS: read_iter(%pD2, %lu@%lu)\n",
+ iocb->ki_filp,
- (unsigned long) iov_length(iov, nr_segs), (unsigned long) pos);
+ (unsigned long) iov_iter_count(iter), (unsigned long) pos);

result = nfs_revalidate_mapping(inode, iocb->ki_filp->f_mapping);
if (!result) {
@@@ -655,25 -634,24 +633,24 @@@ static int nfs_need_sync_write(struct f
return 0;
}

-ssize_t nfs_file_write(struct kiocb *iocb, const struct iovec *iov,
- unsigned long nr_segs, loff_t pos)
+ssize_t nfs_file_write_iter(struct kiocb *iocb, struct iov_iter *iter,
+ loff_t pos)
{
- struct dentry * dentry = iocb->ki_filp->f_path.dentry;
- struct inode * inode = dentry->d_inode;
+ struct file *file = iocb->ki_filp;
+ struct inode *inode = file_inode(file);
unsigned long written = 0;
ssize_t result;
- size_t count = iov_length(iov, nr_segs);
+ size_t count = iov_iter_count(iter);

- result = nfs_key_timeout_notify(iocb->ki_filp, inode);
+ result = nfs_key_timeout_notify(file, inode);
if (result)
return result;

- if (iocb->ki_filp->f_flags & O_DIRECT)
+ if (file->f_flags & O_DIRECT)
- return nfs_file_direct_write(iocb, iov, nr_segs, pos, true);
+ return nfs_file_direct_write(iocb, iter, pos);

- dprintk("NFS: write_iter(%s/%s, %lu@%lld)\n",
- dentry->d_parent->d_name.name, dentry->d_name.name,
- (unsigned long) count, (long long) pos);
- dprintk("NFS: write(%pD2, %lu@%Ld)\n",
++ dprintk("NFS: write_iter(%pD2, %lu@%Ld)\n",
+ file, (unsigned long) count, (long long) pos);

result = -EBUSY;
if (IS_SWAPFILE(inode))


Attachments:
(No filename) (5.72 kB)
(No filename) (836.00 B)
Download all attachments

2013-09-18 02:00:39

by Al Viro

[permalink] [raw]
Subject: Re: linux-next: manual merge of the vfs tree with the aio-direct tree

On Wed, Sep 18, 2013 at 10:56:38AM +1000, Stephen Rothwell wrote:
> Hi Al,
>
> Today's linux-next merge of the vfs tree got conflicts in fs/nfs/direct.c
> and fs/nfs/file.c between commits b9517433d65d ("dio: Convert direct_IO
> to use iov_iter"), a8431c667ae8 ("nfs: add support for read_iter,
> write_iter") and a1b8ec384b73 ("nfs: simplify swap") from the aio-direct
> tree and commit c18d1ec44f7a ("nfs: use %p[dD] instead of open-coded (and
> often racy) equivalents") from the vfs tree.
>
> I fixed it up (see below) and can carry the fix as necessary (no action
> is required).

Hrm... FWIW, I would greatly prefer to offload these printk patches to
NFS and NFSD trees. Didn't get around to that yet, but...

As for aio-direct... Two questions:
* had anybody tried to measure the effect on branch predictor from
introducing that method vector? Commit d6afd4c4 ("iov_iter: hide iovec
details behind ops function pointers")
* WTF does aforementioned commit lack its author's s-o-b? The same
goes for a lot of zab's commits in there...

2013-09-18 03:25:30

by Dave Kleikamp

[permalink] [raw]
Subject: Re: linux-next: manual merge of the vfs tree with the aio-direct tree

On 09/17/2013 09:00 PM, Al Viro wrote:
> On Wed, Sep 18, 2013 at 10:56:38AM +1000, Stephen Rothwell wrote:
>> Hi Al,
>>
>> Today's linux-next merge of the vfs tree got conflicts in fs/nfs/direct.c
>> and fs/nfs/file.c between commits b9517433d65d ("dio: Convert direct_IO
>> to use iov_iter"), a8431c667ae8 ("nfs: add support for read_iter,
>> write_iter") and a1b8ec384b73 ("nfs: simplify swap") from the aio-direct
>> tree and commit c18d1ec44f7a ("nfs: use %p[dD] instead of open-coded (and
>> often racy) equivalents") from the vfs tree.
>>
>> I fixed it up (see below) and can carry the fix as necessary (no action
>> is required).
>
> Hrm... FWIW, I would greatly prefer to offload these printk patches to
> NFS and NFSD trees. Didn't get around to that yet, but...
>
> As for aio-direct... Two questions:
> * had anybody tried to measure the effect on branch predictor from
> introducing that method vector? Commit d6afd4c4 ("iov_iter: hide iovec
> details behind ops function pointers")

Any suggestions for a good benchmark?

> * WTF does aforementioned commit lack its author's s-o-b? The same
> goes for a lot of zab's commits in there...

That would be my bad. Zach's original patches had his s-o-b (w/an Oracle
email address). I'll add them back.

Shaggy

2013-09-18 17:06:42

by Zach Brown

[permalink] [raw]
Subject: Re: linux-next: manual merge of the vfs tree with the aio-direct tree

> As for aio-direct... Two questions:
> * had anybody tried to measure the effect on branch predictor from
> introducing that method vector? Commit d6afd4c4 ("iov_iter: hide iovec
> details behind ops function pointers")

FWIW, I never did. I only went that route to begin with because the few
alternatives I tried before that were even more ugly :(. It's been long
enough that I can't quite remember what I'd tried, though.

- z