From: Gabriel Goller <g.goller@proxmox.com>
To: pbs-devel@lists.proxmox.com
Subject: [pbs-devel] [PATCH] fix: use fragmented block size for space calculation
Date: Mon, 22 Jan 2024 15:23:32 +0100 [thread overview]
Message-ID: <20240122142333.215210-1-g.goller@proxmox.com> (raw)
We currently calculate the size of a datastore using `statfs64`, which
returns the number of blocks in the fs and the two block sizes:
fragemented block size(f_frsize) and block size (f_bsize). To calculate
eg the total space in a datastore we use total_blocks * f_bsize, which
is not always correct.
`f_frsize` is the minimum unit of allocation on the filesystem (in
bytes) and in 99% of the cases equal to `f_bsize`, but in some cases
it differs. For example some filesystems allow smaller blocks for small
files, in case f_frsize < f_bsize. In that case, f_frsize * total_blocks
returns (mostly) the correct result (ceph also did some weird stuff, which is
now being fixed though [0][1]). `statvfs` also documents this as the
recommended way ('fsblkcnt_t f_blocks; /* Size of fs in f_frsize units */')[2].
This patch aligns the the behavior with the libc utilities (also used by
`df`) [3].
Motivation: [4] (Forum post)
[0]: https://tracker.ceph.com/issues/3793
[1]: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=92a49fb0f79f3300e6e50ddf56238e70678e4202
[2]: https://www.man7.org/linux/man-pages/man3/statvfs.3.html
[3]: https://git.savannah.gnu.org/cgit/gnulib.git/tree/lib/fsusage.c#n147
[4]: https://forum.proxmox.com/threads/pbs-3-1-2-wrong-datastore-information-sshfs.139875/#post-626959
Signed-off-by: Gabriel Goller <g.goller@proxmox.com>
---
proxmox-sys/src/fs/mod.rs | 12 ++++++++----
1 file changed, 8 insertions(+), 4 deletions(-)
diff --git a/proxmox-sys/src/fs/mod.rs b/proxmox-sys/src/fs/mod.rs
index 8fb677c..44c3da5 100644
--- a/proxmox-sys/src/fs/mod.rs
+++ b/proxmox-sys/src/fs/mod.rs
@@ -121,12 +121,16 @@ pub fn fs_info<P: ?Sized + nix::NixPath>(path: &P) -> nix::Result<FileSystemInfo
let res = path.with_nix_path(|cstr| unsafe { libc::statfs64(cstr.as_ptr(), &mut stat) })?;
nix::errno::Errno::result(res)?;
- let bsize = stat.f_bsize as u64;
+ let block_size = if stat.f_frsize == 0 {
+ stat.f_bsize as u64
+ } else {
+ stat.f_frsize as u64
+ };
Ok(FileSystemInformation {
- total: stat.f_blocks * bsize,
- used: (stat.f_blocks - stat.f_bfree) * bsize,
- available: stat.f_bavail * bsize,
+ total: stat.f_blocks * block_size,
+ used: (stat.f_blocks - stat.f_bfree) * block_size,
+ available: stat.f_bavail * block_size,
total_inodes: stat.f_files,
free_inodes: stat.f_ffree,
fs_type: stat.f_type,
--
2.43.0
next reply other threads:[~2024-01-22 14:24 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-01-22 14:23 Gabriel Goller [this message]
2024-03-25 15:59 ` [pbs-devel] applied: " Thomas Lamprecht
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20240122142333.215210-1-g.goller@proxmox.com \
--to=g.goller@proxmox.com \
--cc=pbs-devel@lists.proxmox.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.