From: Stefan Reiter <s.reiter@proxmox.com>
To: Proxmox Backup Server development discussion
<pbs-devel@lists.proxmox.com>,
Dietmar Maurer <dietmar@proxmox.com>
Subject: Re: [pbs-devel] [PATCH backup] src/api2/backup.rs: aquire backup lock earlier in create_locked_backup_group()
Date: Thu, 30 Jul 2020 11:31:32 +0200 [thread overview]
Message-ID: <89116eff-a7ca-a428-08d2-52432a3173e0@proxmox.com> (raw)
In-Reply-To: <20200730085031.30790-1-dietmar@proxmox.com>
Looking again, I don't think the race would have mattered (since we
checked the owner after create_backup_group() anyway), but locking early
is still superior and also works nicely for pull.rs.
Reviewed-By: Stefan Reiter <s.reiter@proxmox.com>
On 7/30/20 10:50 AM, Dietmar Maurer wrote:
> ---
> src/api2/backup.rs | 8 ++++----
> src/backup/datastore.rs | 14 +++++++++-----
> src/client/pull.rs | 2 +-
> 3 files changed, 14 insertions(+), 10 deletions(-)
>
> diff --git a/src/api2/backup.rs b/src/api2/backup.rs
> index b9dff1f..ad13faa 100644
> --- a/src/api2/backup.rs
> +++ b/src/api2/backup.rs
> @@ -88,7 +88,10 @@ async move {
> let env_type = rpcenv.env_type();
>
> let backup_group = BackupGroup::new(backup_type, backup_id);
> - let owner = datastore.create_backup_group(&backup_group, &username)?;
> +
> + // lock backup group to only allow one backup per group at a time
> + let (owner, _group_guard) = datastore.create_locked_backup_group(&backup_group, &username)?;
> +
> // permission check
> if owner != username { // only the owner is allowed to create additional snapshots
> bail!("backup owner check failed ({} != {})", username, owner);
> @@ -103,9 +106,6 @@ async move {
> }
> }
>
> - // lock backup group to only allow one backup per group at a time
> - let _group_guard = backup_group.lock(&datastore.base_path())?;
> -
> let (path, is_new) = datastore.create_backup_dir(&backup_dir)?;
> if !is_new { bail!("backup directory already exists."); }
>
> diff --git a/src/backup/datastore.rs b/src/backup/datastore.rs
> index 9c2f285..af964f6 100644
> --- a/src/backup/datastore.rs
> +++ b/src/backup/datastore.rs
> @@ -8,7 +8,7 @@ use anyhow::{bail, format_err, Error};
> use lazy_static::lazy_static;
> use chrono::{DateTime, Utc};
>
> -use super::backup_info::{BackupGroup, BackupDir, BackupInfo};
> +use super::backup_info::{BackupGroup, BackupGroupGuard, BackupDir, BackupInfo};
> use super::chunk_store::ChunkStore;
> use super::dynamic_index::{DynamicIndexReader, DynamicIndexWriter};
> use super::fixed_index::{FixedIndexReader, FixedIndexWriter};
> @@ -318,11 +318,13 @@ impl DataStore {
> Ok(())
> }
>
> - /// Create a backup group if it does not already exists.
> + /// Create (if it does not already exists) and lock a backup group
> ///
> /// And set the owner to 'userid'. If the group already exists, it returns the
> /// current owner (instead of setting the owner).
> - pub fn create_backup_group(&self, backup_group: &BackupGroup, userid: &str) -> Result<String, Error> {
> + ///
> + /// This also aquires an exclusive lock on the directory and returns the lock guard.
> + pub fn create_locked_backup_group(&self, backup_group: &BackupGroup, userid: &str) -> Result<(String, BackupGroupGuard), Error> {
>
> // create intermediate path first:
> let base_path = self.base_path();
> @@ -336,13 +338,15 @@ impl DataStore {
> // create the last component now
> match std::fs::create_dir(&full_path) {
> Ok(_) => {
> + let guard = backup_group.lock(&base_path)?;
> self.set_owner(backup_group, userid, false)?;
> let owner = self.get_owner(backup_group)?; // just to be sure
> - Ok(owner)
> + Ok((owner, guard))
> }
> Err(ref err) if err.kind() == io::ErrorKind::AlreadyExists => {
> + let guard = backup_group.lock(&base_path)?;
> let owner = self.get_owner(backup_group)?; // just to be sure
> - Ok(owner)
> + Ok((owner, guard))
> }
> Err(err) => bail!("unable to create backup group {:?} - {}", full_path, err),
> }
> diff --git a/src/client/pull.rs b/src/client/pull.rs
> index c44cb9f..fc193a8 100644
> --- a/src/client/pull.rs
> +++ b/src/client/pull.rs
> @@ -406,7 +406,7 @@ pub async fn pull_store(
> for item in list {
> let group = BackupGroup::new(&item.backup_type, &item.backup_id);
>
> - let owner = tgt_store.create_backup_group(&group, &username)?;
> + let (owner, _lock_guard) = tgt_store.create_locked_backup_group(&group, &username)?;
> // permission check
> if owner != username { // only the owner is allowed to create additional snapshots
> worker.log(format!("sync group {}/{} failed - owner check failed ({} != {})",
>
prev parent reply other threads:[~2020-07-30 9:31 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-07-30 8:50 Dietmar Maurer
2020-07-30 9:31 ` Stefan Reiter [this message]
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=89116eff-a7ca-a428-08d2-52432a3173e0@proxmox.com \
--to=s.reiter@proxmox.com \
--cc=dietmar@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.