On 24/02/27 02:39PM, Christian Brauner wrote:
> On Fri, Feb 23, 2024 at 11:41:54AM -0600, John Groves wrote:
> > Famfs works on both /dev/pmem and /dev/dax devices. This commit introduces
> > the function that opens a block (pmem) device and the struct
> > dax_holder_operations that are needed for that ABI.
> >
> > In this commit, support for opening character /dev/dax is stubbed. A
> > later commit introduces this capability.
> >
> > Signed-off-by: John Groves <john@xxxxxxxxxx>
> > ---
> > fs/famfs/famfs_inode.c | 83 ++++++++++++++++++++++++++++++++++++++++++
> > 1 file changed, 83 insertions(+)
> >
> > diff --git a/fs/famfs/famfs_inode.c b/fs/famfs/famfs_inode.c
> > index 3329aff000d1..82c861998093 100644
> > --- a/fs/famfs/famfs_inode.c
> > +++ b/fs/famfs/famfs_inode.c
> > @@ -68,5 +68,88 @@ static const struct super_operations famfs_ops = {
> > .show_options = famfs_show_options,
> > };
> >
> > +/***************************************************************************************
> > + * dax_holder_operations for block dax
> > + */
> > +
> > +static int
> > +famfs_blk_dax_notify_failure(
> > + struct dax_device *dax_devp,
> > + u64 offset,
> > + u64 len,
> > + int mf_flags)
> > +{
> > +
> > + pr_err("%s: dax_devp %llx offset %llx len %lld mf_flags %x\n",
> > + __func__, (u64)dax_devp, (u64)offset, (u64)len, mf_flags);
> > + return -EOPNOTSUPP;
> > +}
> > +
> > +const struct dax_holder_operations famfs_blk_dax_holder_ops = {
> > + .notify_failure = famfs_blk_dax_notify_failure,
> > +};
> > +
> > +static int
> > +famfs_open_char_device(
> > + struct super_block *sb,
> > + struct fs_context *fc)
> > +{
> > + pr_err("%s: Root device is %s, but your kernel does not support famfs on /dev/dax\n",
> > + __func__, fc->source);
> > + return -ENODEV;
> > +}
> > +
> > +/**
> > + * famfs_open_device()
> > + *
> > + * Open the memory device. If it looks like /dev/dax, call famfs_open_char_device().
> > + * Otherwise try to open it as a block/pmem device.
> > + */
> > +static int
> > +famfs_open_device(
>
> I'm confused why that function is added here but it's completely unclear
> in what wider context it's called. This is really hard to follow.
First, thank you for taking the time to do a thoughtful review.
I didn't factor this series correctly. The next one will be
"module-operations-up" unless you or somebody suggests a more sensible
approach.
Some background that might be useful: this work is really targeted for
/dev/dax, but it started on /dev/pmem because the iomap interface wasn't
working on /dev/dax. This patch addresses that (the dev_dax_iomap commits),
although it's likely that code will evolve.
The current famfs code base tries to support both pmem (block) and /dev/dax
(char), but I'm now thinking it should move to /dev/dax-only (no block
support).
/dev/pmem devices can converted to /dev/dax mode anyway, so I'm not sure
there is a reason to support both interfaces. (Need to think a bit more on
that...).
>
> > + struct super_block *sb,
> > + struct fs_context *fc)
> > +{
> > + struct famfs_fs_info *fsi = sb->s_fs_info;
> > + struct dax_device *dax_devp;
> > + u64 start_off = 0;
> > + struct bdev_handle *handlep;
> > +
> > + if (fsi->dax_devp) {
> > + pr_err("%s: already mounted\n", __func__);
> > + return -EALREADY;
> > + }
> > +
> > + if (strstr(fc->source, "/dev/dax")) /* There is probably a better way to check this */
> > + return famfs_open_char_device(sb, fc);
> > +
> > + if (!strstr(fc->source, "/dev/pmem")) { /* There is probably a better way to check this */
>
> Yeah, this is not just a bit ugly but also likely wrong because:
>
> sudo mount --bind /dev/pmem /opt/muhaha
>
> fsconfig(fd_fs, FSCONFIG_SET_STRING, "source", "/opt/muhaha", [...])
>
> or a simple mknod to create that device somewhere else. You likely want:
>
> lookup_bdev(fc->source, &dev);
>
> if (!DEVICE_NUMBER_SOMETHING_SOMETHING_SANE(dev))
> return invalfc(fc, "SOMETHING SOMETHING...
>
> bdev_open_by_dev(dev, ....)
>
> (This reminds me that I should get back to making it possible to specify
> "source" as a file descriptor instead of a mere string with the new
> mount api...)
All good points - sorry for the flakyness here.
I think the solution is to stop trying to support both pmem and dax. Then
I don't need to distinguish between different device types.
>
> > + pr_err("%s: primary backing dev (%s) is not pmem\n",
> > + __func__, fc->source);
> > + return -EINVAL;
> > + }
> > +
> > + handlep = bdev_open_by_path(fc->source, FAMFS_BLKDEV_MODE, fsi, &fs_holder_ops);
>
> Hm, I suspected that FAMFS_BLKDEV_MODE would be wrong based on:
> https://lore.kernel.org/r/13556dbbd8d0f51bc31e3bdec796283fe85c6baf.1708709155.git.john@xxxxxxxxxx
>
> It's defined as FMODE_READ | FMODE_WRITE which is wrong. But these
> helpers want BLOCK_OPEN_READ | BLOCK_OPEN_WRITE.
Dropping pmem/block support will also make this go away
>
> > + if (IS_ERR(handlep->bdev)) {
>
> @bdev_handle will be gone as of v6.9 so you might want to wait until
> then to resend.
And this dependency will also disappear...
Thank you!!
John
