Re: [Patch v4 1/3] CIFS: Add support for direct I/O read

From: Pavel Shilovsky
Date: Thu Nov 29 2018 - 18:23:22 EST


ÑÑ, 29 ÐÐÑÐ. 2018 Ð. Ð 14:48, Long Li <longli@xxxxxxxxxxxxx>:
>
> > Subject: Re: [Patch v4 1/3] CIFS: Add support for direct I/O read
> >
> > ÑÑ, 28 ÐÐÑÐ. 2018 Ð. Ð 15:43, Long Li <longli@xxxxxxxxxxxxx>:
> > >
> > > > Subject: Re: [Patch v4 1/3] CIFS: Add support for direct I/O read
> > > >
> > > > Hi Long,
> > > >
> > > > Please find my comments below.
> > > >
> > > >
> > > > ÑÑ, 31 ÐÐÑ. 2018 Ð. Ð 15:14, Long Li <longli@xxxxxxxxxxxxxxxxx>:
> > > > >
> > > > > From: Long Li <longli@xxxxxxxxxxxxx>
> > > > >
> > > > > With direct I/O read, we transfer the data directly from transport
> > > > > layer to the user data buffer.
> > > > >
> > > > > Change in v3: add support for kernel AIO
> > > > >
> > > > > Change in v4:
> > > > > Refactor common read code to __cifs_readv for direct and non-direct
> > I/O.
> > > > > Retry on direct I/O failure.
> > > > >
> > > > > Signed-off-by: Long Li <longli@xxxxxxxxxxxxx>
> > > > > ---
> > > > > fs/cifs/cifsfs.h | 1 +
> > > > > fs/cifs/cifsglob.h | 5 ++
> > > > > fs/cifs/file.c | 219
> > +++++++++++++++++++++++++++++++++++++++++++--
> > > > --------
> > > > > 3 files changed, 186 insertions(+), 39 deletions(-)
> > > > >
> > > > > diff --git a/fs/cifs/cifsfs.h b/fs/cifs/cifsfs.h index
> > > > > 5f02318..7fba9aa 100644
> > > > > --- a/fs/cifs/cifsfs.h
> > > > > +++ b/fs/cifs/cifsfs.h
> > > > > @@ -102,6 +102,7 @@ extern int cifs_open(struct inode *inode,
> > > > > struct file *file); extern int cifs_close(struct inode *inode,
> > > > > struct file *file); extern int cifs_closedir(struct inode *inode,
> > > > > struct file *file); extern ssize_t cifs_user_readv(struct kiocb
> > > > > *iocb, struct iov_iter *to);
> > > > > +extern ssize_t cifs_direct_readv(struct kiocb *iocb, struct
> > > > > +iov_iter *to);
> > > > > extern ssize_t cifs_strict_readv(struct kiocb *iocb, struct
> > > > > iov_iter *to); extern ssize_t cifs_user_writev(struct kiocb
> > > > > *iocb, struct iov_iter *from); extern ssize_t
> > > > > cifs_strict_writev(struct kiocb *iocb, struct iov_iter *from);
> > > > > diff --git a/fs/cifs/cifsglob.h b/fs/cifs/cifsglob.h index
> > > > > 7f62c98..52248dd 100644
> > > > > --- a/fs/cifs/cifsglob.h
> > > > > +++ b/fs/cifs/cifsglob.h
> > > > > @@ -1146,6 +1146,11 @@ struct cifs_aio_ctx {
> > > > > unsigned int len;
> > > > > unsigned int total_len;
> > > > > bool should_dirty;
> > > > > + /*
> > > > > + * Indicates if this aio_ctx is for direct_io,
> > > > > + * If yes, iter is a copy of the user passed iov_iter
> > > > > + */
> > > > > + bool direct_io;
> > > > > };
> > > > >
> > > > > struct cifs_readdata;
> > > > > diff --git a/fs/cifs/file.c b/fs/cifs/file.c index
> > > > > 87eece6..daab878
> > > > > 100644
> > > > > --- a/fs/cifs/file.c
> > > > > +++ b/fs/cifs/file.c
> > > > > @@ -2965,7 +2965,6 @@ cifs_uncached_readdata_release(struct kref
> > > > *refcount)
> > > > > kref_put(&rdata->ctx->refcount, cifs_aio_ctx_release);
> > > > > for (i = 0; i < rdata->nr_pages; i++) {
> > > > > put_page(rdata->pages[i]);
> > > > > - rdata->pages[i] = NULL;
> > > > > }
> > > > > cifs_readdata_release(refcount); } @@ -3092,6 +3091,63 @@
> > > > > cifs_uncached_copy_into_pages(struct TCP_Server_Info *server,
> > > > > return uncached_fill_pages(server, rdata, iter,
> > > > > iter->count); }
> > > > >
> > > > > +static int cifs_resend_rdata(struct cifs_readdata *rdata,
> > > > > + struct list_head *rdata_list,
> > > > > + struct cifs_aio_ctx *ctx) {
> > > > > + int wait_retry = 0;
> > > > > + unsigned int rsize, credits;
> > > > > + int rc;
> > > > > + struct TCP_Server_Info *server =
> > > > > +tlink_tcon(rdata->cfile->tlink)->ses->server;
> > > > > +
> > > > > + /*
> > > > > + * Try to resend this rdata, waiting for credits up to 3 seconds.
> > > > > + * Note: we are attempting to resend the whole rdata not in
> > segments
> > > > > + */
> > > > > + do {
> > > > > + rc = server->ops->wait_mtu_credits(server, rdata->bytes,
> > > > > + &rsize, &credits);
> > > > > +
> > > > > + if (rc)
> > > > > + break;
> > > > > +
> > > > > + if (rsize < rdata->bytes) {
> > > > > + add_credits_and_wake_if(server, credits, 0);
> > > > > + msleep(1000);
> > > > > + wait_retry++;
> > > > > + }
> > > > > + } while (rsize < rdata->bytes && wait_retry < 3);
> > > > > +
> > > > > + /*
> > > > > + * If we can't find enough credits to send this rdata
> > > > > + * release the rdata and return failure, this will pass
> > > > > + * whatever I/O amount we have finished to VFS.
> > > > > + */
> > > > > + if (rsize < rdata->bytes) {
> > > > > + rc = -EBUSY;
> > > >
> > > > We don't have enough credits and return EBUSY here...
> > > >
> > > > > + goto out;
> > > > > + }
> > > > > +
> > > > > + rc = -EAGAIN;
> > > > > + while (rc == -EAGAIN)
> > > > > + if (!rdata->cfile->invalidHandle ||
> > > > > + !(rc = cifs_reopen_file(rdata->cfile, true)))
> > > > > + rc = server->ops->async_readv(rdata);
> > > > > +
> > > > > + if (!rc) {
> > > > > + /* Add to aio pending list */
> > > > > + list_add_tail(&rdata->list, rdata_list);
> > > > > + return 0;
> > > > > + }
> > > > > +
> > > > > + add_credits_and_wake_if(server, rdata->credits, 0);
> > > > > +out:
> > > > > + kref_put(&rdata->refcount,
> > > > > + cifs_uncached_readdata_release);
> > > > > +
> > > > > + return rc;
> > > > > +}
> > > > > +
> > > > > static int
> > > > > cifs_send_async_read(loff_t offset, size_t len, struct cifsFileInfo
> > *open_file,
> > > > > struct cifs_sb_info *cifs_sb, struct
> > > > > list_head *rdata_list, @@ -3103,6 +3159,9 @@
> > > > > cifs_send_async_read(loff_t offset,
> > > > size_t len, struct cifsFileInfo *open_file,
> > > > > int rc;
> > > > > pid_t pid;
> > > > > struct TCP_Server_Info *server;
> > > > > + struct page **pagevec;
> > > > > + size_t start;
> > > > > + struct iov_iter direct_iov = ctx->iter;
> > > > >
> > > > > server = tlink_tcon(open_file->tlink)->ses->server;
> > > > >
> > > > > @@ -3111,6 +3170,9 @@ cifs_send_async_read(loff_t offset, size_t
> > > > > len,
> > > > struct cifsFileInfo *open_file,
> > > > > else
> > > > > pid = current->tgid;
> > > > >
> > > > > + if (ctx->direct_io)
> > > > > + iov_iter_advance(&direct_iov, offset - ctx->pos);
> > > > > +
> > > > > do {
> > > > > rc = server->ops->wait_mtu_credits(server, cifs_sb->rsize,
> > > > > &rsize,
> > > > > &credits); @@ -3118,20 +3180,56 @@ cifs_send_async_read(loff_t
> > > > > offset, size_t len,
> > > > struct cifsFileInfo *open_file,
> > > > > break;
> > > > >
> > > > > cur_len = min_t(const size_t, len, rsize);
> > > > > - npages = DIV_ROUND_UP(cur_len, PAGE_SIZE);
> > > > >
> > > > > - /* allocate a readdata struct */
> > > > > - rdata = cifs_readdata_alloc(npages,
> > > > > + if (ctx->direct_io) {
> > > > > +
> > > > > + cur_len = iov_iter_get_pages_alloc(
> > > > > + &direct_iov, &pagevec,
> > > > > + cur_len, &start);
> > > > > + if (cur_len < 0) {
> > > > > + cifs_dbg(VFS,
> > > > > + "couldn't get user pages (cur_len=%zd)"
> > > > > + " iter type %d"
> > > > > + " iov_offset %zd count %zd\n",
> > > > > + cur_len, direct_iov.type, direct_iov.iov_offset,
> > > > > + direct_iov.count);
> > > > > + dump_stack();
> > > > > + break;
> > > > > + }
> > > > > + iov_iter_advance(&direct_iov, cur_len);
> > > > > +
> > > > > + rdata = cifs_readdata_direct_alloc(
> > > > > + pagevec, cifs_uncached_readv_complete);
> > > > > + if (!rdata) {
> > > > > + add_credits_and_wake_if(server, credits, 0);
> > > > > + rc = -ENOMEM;
> > > > > + break;
> > > > > + }
> > > > > +
> > > > > + npages = (cur_len + start + PAGE_SIZE-1) / PAGE_SIZE;
> > > > > + rdata->page_offset = start;
> > > > > + rdata->tailsz = npages > 1 ?
> > > > > + cur_len-(PAGE_SIZE-start)-(npages-2)*PAGE_SIZE :
> > > > > + cur_len;
> > > > > +
> > > > > + } else {
> > > > > +
> > > > > + npages = DIV_ROUND_UP(cur_len, PAGE_SIZE);
> > > > > + /* allocate a readdata struct */
> > > > > + rdata = cifs_readdata_alloc(npages,
> > > > > cifs_uncached_readv_complete);
> > > > > - if (!rdata) {
> > > > > - add_credits_and_wake_if(server, credits, 0);
> > > > > - rc = -ENOMEM;
> > > > > - break;
> > > > > - }
> > > > > + if (!rdata) {
> > > > > + add_credits_and_wake_if(server, credits, 0);
> > > > > + rc = -ENOMEM;
> > > > > + break;
> > > > > + }
> > > > >
> > > > > - rc = cifs_read_allocate_pages(rdata, npages);
> > > > > - if (rc)
> > > > > - goto error;
> > > > > + rc = cifs_read_allocate_pages(rdata, npages);
> > > > > + if (rc)
> > > > > + goto error;
> > > > > +
> > > > > + rdata->tailsz = PAGE_SIZE;
> > > > > + }
> > > > >
> > > > > rdata->cfile = cifsFileInfo_get(open_file);
> > > > > rdata->nr_pages = npages; @@ -3139,7 +3237,6 @@
> > > > > cifs_send_async_read(loff_t offset, size_t len, struct cifsFileInfo
> > *open_file,
> > > > > rdata->bytes = cur_len;
> > > > > rdata->pid = pid;
> > > > > rdata->pagesz = PAGE_SIZE;
> > > > > - rdata->tailsz = PAGE_SIZE;
> > > > > rdata->read_into_pages = cifs_uncached_read_into_pages;
> > > > > rdata->copy_into_pages = cifs_uncached_copy_into_pages;
> > > > > rdata->credits = credits; @@ -3153,9 +3250,11 @@
> > > > > cifs_send_async_read(loff_t offset, size_t len, struct cifsFileInfo
> > *open_file,
> > > > > if (rc) {
> > > > > add_credits_and_wake_if(server, rdata->credits, 0);
> > > > > kref_put(&rdata->refcount,
> > > > > - cifs_uncached_readdata_release);
> > > > > - if (rc == -EAGAIN)
> > > > > + cifs_uncached_readdata_release);
> > > > > + if (rc == -EAGAIN) {
> > > > > + iov_iter_revert(&direct_iov,
> > > > > + cur_len);
> > > > > continue;
> > > > > + }
> > > > > break;
> > > > > }
> > > > >
> > > > > @@ -3211,45 +3310,62 @@ collect_uncached_read_data(struct
> > > > cifs_aio_ctx *ctx)
> > > > > * reading.
> > > > > */
> > > > > if (got_bytes && got_bytes < rdata->bytes) {
> > > > > - rc = cifs_readdata_to_iov(rdata, to);
> > > > > + rc = 0;
> > > > > + if (!ctx->direct_io)
> > > > > + rc =
> > > > > + cifs_readdata_to_iov(rdata, to);
> > > > > if (rc) {
> > > > > kref_put(&rdata->refcount,
> > > > > - cifs_uncached_readdata_release);
> > > > > +
> > > > > + cifs_uncached_readdata_release);
> > > > > continue;
> > > > > }
> > > > > }
> > > > >
> > > > > - rc = cifs_send_async_read(
> > > > > + if (ctx->direct_io) {
> > > > > + /*
> > > > > + * Re-use rdata as this is a
> > > > > + * direct I/O
> > > > > + */
> > > > > + rc = cifs_resend_rdata(
> > > > > + rdata,
> > > > > + &tmp_list, ctx);
> > > >
> > > > ...so we set rc to -EBUSY...
> > > >
> > > >
> > > > > + } else {
> > > > > + rc = cifs_send_async_read(
> > > > > rdata->offset + got_bytes,
> > > > > rdata->bytes - got_bytes,
> > > > > rdata->cfile, cifs_sb,
> > > > > &tmp_list, ctx);
> > > > >
> > > > > + kref_put(&rdata->refcount,
> > > > > + cifs_uncached_readdata_release);
> > > > > + }
> > > > > +
> > > > > list_splice(&tmp_list,
> > > > > &ctx->list);
> > > > >
> > > > > - kref_put(&rdata->refcount,
> > > > > - cifs_uncached_readdata_release);
> > > > > goto again;
> > > > > } else if (rdata->result)
> > > > > rc = rdata->result;
> > > > > - else
> > > > > + else if (!ctx->direct_io)
> > > > > rc = cifs_readdata_to_iov(rdata,
> > > > > to);
> > > > >
> > > > > /* if there was a short read -- discard anything left */
> > > > > if (rdata->got_bytes && rdata->got_bytes < rdata->bytes)
> > > > > rc = -ENODATA;
> > > > > +
> > > > > + ctx->total_len += rdata->got_bytes;
> > > > > }
> > > > > list_del_init(&rdata->list);
> > > > > kref_put(&rdata->refcount, cifs_uncached_readdata_release);
> > > > > }
> > > > >
> > > > > - for (i = 0; i < ctx->npages; i++) {
> > > > > - if (ctx->should_dirty)
> > > > > - set_page_dirty(ctx->bv[i].bv_page);
> > > > > - put_page(ctx->bv[i].bv_page);
> > > > > - }
> > > > > + if (!ctx->direct_io) {
> > > > > + for (i = 0; i < ctx->npages; i++) {
> > > > > + if (ctx->should_dirty)
> > > > > + set_page_dirty(ctx->bv[i].bv_page);
> > > > > + put_page(ctx->bv[i].bv_page);
> > > > > + }
> > > > >
> > > > > - ctx->total_len = ctx->len - iov_iter_count(to);
> > > > > + ctx->total_len = ctx->len - iov_iter_count(to);
> > > > > + }
> > > > >
> > > > > cifs_stats_bytes_read(tcon, ctx->total_len);
> > > > >
> > > > > @@ -3267,18 +3383,27 @@ collect_uncached_read_data(struct
> > > > cifs_aio_ctx *ctx)
> > > > > complete(&ctx->done);
> > > >
> > > > and then ctx->rc is set to -EBUSY too.
> > > >
> > > > > }
> > > > >
> > > > > -ssize_t cifs_user_readv(struct kiocb *iocb, struct iov_iter *to)
> > > > > +ssize_t __cifs_readv(struct kiocb *iocb, struct iov_iter *to,
> > > > > +bool
> > > > > +direct)
> > > > > {
> > > > > - struct file *file = iocb->ki_filp;
> > > > > - ssize_t rc;
> > > > > size_t len;
> > > > > - ssize_t total_read = 0;
> > > > > - loff_t offset = iocb->ki_pos;
> > > > > + struct file *file = iocb->ki_filp;
> > > > > struct cifs_sb_info *cifs_sb;
> > > > > - struct cifs_tcon *tcon;
> > > > > struct cifsFileInfo *cfile;
> > > > > + struct cifs_tcon *tcon;
> > > > > + ssize_t rc, total_read = 0;
> > > > > + loff_t offset = iocb->ki_pos;
> > > > > struct cifs_aio_ctx *ctx;
> > > > >
> > > > > + /*
> > > > > + * iov_iter_get_pages_alloc() doesn't work with ITER_KVEC,
> > > > > + * fall back to data copy read path
> > > > > + * this could be improved by getting pages directly in ITER_KVEC
> > > > > + */
> > > > > + if (direct && to->type & ITER_KVEC) {
> > > > > + cifs_dbg(FYI, "use non-direct cifs_user_readv for kvec I/O\n");
> > > > > + direct = false;
> > > > > + }
> > > > > +
> > > > > len = iov_iter_count(to);
> > > > > if (!len)
> > > > > return 0;
> > > > > @@ -3305,14 +3430,20 @@ ssize_t cifs_user_readv(struct kiocb
> > > > > *iocb,
> > > > struct iov_iter *to)
> > > > > if (to->type == ITER_IOVEC)
> > > > > ctx->should_dirty = true;
> > > > >
> > > > > - rc = setup_aio_ctx_iter(ctx, to, READ);
> > > > > - if (rc) {
> > > > > - kref_put(&ctx->refcount, cifs_aio_ctx_release);
> > > > > - return rc;
> > > > > + if (direct) {
> > > > > + ctx->pos = offset;
> > > > > + ctx->direct_io = true;
> > > > > + ctx->iter = *to;
> > > > > + ctx->len = len;
> > > > > + } else {
> > > > > + rc = setup_aio_ctx_iter(ctx, to, READ);
> > > > > + if (rc) {
> > > > > + kref_put(&ctx->refcount, cifs_aio_ctx_release);
> > > > > + return rc;
> > > > > + }
> > > > > + len = ctx->len;
> > > > > }
> > > > >
> > > > > - len = ctx->len;
> > > > > -
> > > > > /* grab a lock here due to read response handlers can access ctx */
> > > > > mutex_lock(&ctx->aio_mutex);
> > > > >
> > > > > @@ -3354,6 +3485,16 @@ ssize_t cifs_user_readv(struct kiocb *iocb,
> > > > struct iov_iter *to)
> > > > > return rc;
> > > >
> > > > In case of a blocking read we will return -EBUSY to the caller but
> > > > read man page doesn't say anything about a possibility to return
> > > > this error code. Is it being mapped somehow by VFS or is there
> > > > another consideration? The same question applies to the write patch as
> > well.
> > >
> > > Thanks Pavel, Yes this is a bug.
> > >
> > > In practice, the retry logic will rarely get hit. If getting hit, normally the
> > server will give us far more credits for resending this whole packet.
> > >
> > > How about just retrying forever (instead of trying 3 times) on this resend
> > packet? This will change the code to the same behavior before direct I/O
> > patches.
> > >
> > > e.g.:
> > >
> > > /*
> > > * Try to resend this wdata, waiting for credits before sending
> > > * Note: we are attempting to resend the whole wdata not in
> > segments
> > > */
> > > do {
> > > rc = server->ops->wait_mtu_credits(
> > > server, wdata->bytes, &wsize, &credits);
> > >
> > > if (rc)
> > > break;
> > >
> > > if (wsize < wdata->bytes) {
> > > add_credits_and_wake_if(server, credits, 0);
> > > msleep(1000);
> > > }
> > > } while (wsize < wdata->bytes);
> > >
> >
> > We can do that but it won't be the same behavior because we were using a
> > partial send:
>
> Yes it's a little different.
>
> I don't see any ill effects of doing this, other than we may wait for a little longer on a full credit to send the whole packet. In practice the server always offers a full credit on the first call to wait_mtu_credits(), I am yet to see the while loop actually get looped.
>
> The resend path is rarely getting hit in stress tests on TCP, I estimate maybe less than 0.001% on my test setup, the while loop for waiting credits is ever harder to hit that I never see one.
>
> On RDMA, the resend path is never used.

Ok. Another way to fix this is to break the request into several
smaller ones consuming 1 credit per request - see
cifs_writev_requeue() function that does the same thing for writes.

--
Best regards,
Pavel Shilovsky