public inbox for pbs-devel@lists.proxmox.com
 help / color / mirror / Atom feed
From: Christian Ebner <c.ebner@proxmox.com>
To: pbs-devel@lists.proxmox.com
Subject: [pbs-devel] [PATCH proxmox-backup v2 18/19] GC: lock chunk marker before cleanup in phase 3 on s3 backends
Date: Tue,  4 Nov 2025 14:06:58 +0100	[thread overview]
Message-ID: <20251104130659.435139-19-c.ebner@proxmox.com> (raw)
In-Reply-To: <20251104130659.435139-1-c.ebner@proxmox.com>

To make sure there is no race between atime check and deletion with
possible re-insertion.

By only acquiring the file lock if the chunk marker would be removed
and double stating, the file locking penalty is avoided for the other
cases.

Signed-off-by: Christian Ebner <c.ebner@proxmox.com>
---
Changes since version 1:
- no changes

 pbs-datastore/src/chunk_store.rs | 28 +++++++++++++++++++++++++++-
 pbs-datastore/src/datastore.rs   |  2 ++
 2 files changed, 29 insertions(+), 1 deletion(-)

diff --git a/pbs-datastore/src/chunk_store.rs b/pbs-datastore/src/chunk_store.rs
index 49687b2fa..08519fe2b 100644
--- a/pbs-datastore/src/chunk_store.rs
+++ b/pbs-datastore/src/chunk_store.rs
@@ -5,6 +5,7 @@ use std::sync::{Arc, Mutex};
 use std::time::Duration;
 
 use anyhow::{bail, format_err, Context, Error};
+use hex::FromHex;
 use tracing::{info, warn};
 
 use pbs_api_types::{DatastoreFSyncLevel, GarbageCollectionStatus};
@@ -22,7 +23,7 @@ use crate::data_blob::DataChunkBuilder;
 use crate::file_formats::{
     COMPRESSED_BLOB_MAGIC_1_0, ENCRYPTED_BLOB_MAGIC_1_0, UNCOMPRESSED_BLOB_MAGIC_1_0,
 };
-use crate::DataBlob;
+use crate::{DataBlob, LocalDatastoreLruCache};
 
 /// File system based chunk store
 pub struct ChunkStore {
@@ -366,6 +367,7 @@ impl ChunkStore {
         min_atime: i64,
         status: &mut GarbageCollectionStatus,
         worker: &dyn WorkerTaskContext,
+        cache: Option<&LocalDatastoreLruCache>,
     ) -> Result<(), Error> {
         // unwrap: only `None` in unit tests
         assert!(self.locker.is_some());
@@ -419,6 +421,30 @@ impl ChunkStore {
                         bad,
                         status,
                         || {
+                            if let Some(cache) = cache {
+                                // never lock bad chunks
+                                if filename.to_bytes().len() == 64 {
+                                    let digest = <[u8; 32]>::from_hex(filename.to_bytes())?;
+                                    match self.lock_chunk(&digest, Duration::from_secs(0)) {
+                                        Ok(_guard) => {
+                                            // don't remove if changed since locking
+                                            match fstatat(
+                                                Some(dirfd),
+                                                filename,
+                                                nix::fcntl::AtFlags::AT_SYMLINK_NOFOLLOW,
+                                            ) {
+                                                Ok(stat) if stat.st_atime < min_atime => {
+                                                    let _ = cache.remove(&digest);
+                                                    return Ok(());
+                                                }
+                                                _ => return Ok(()),
+                                            }
+                                        }
+                                        Err(_) => return Ok(()),
+                                    }
+                                }
+                            }
+
                             unlinkat(Some(dirfd), filename, UnlinkatFlags::NoRemoveDir).map_err(
                                 |err| {
                                     format_err!(
diff --git a/pbs-datastore/src/datastore.rs b/pbs-datastore/src/datastore.rs
index 8b7333d80..df344974a 100644
--- a/pbs-datastore/src/datastore.rs
+++ b/pbs-datastore/src/datastore.rs
@@ -1761,6 +1761,7 @@ impl DataStore {
                 min_atime,
                 &mut tmp_gc_status,
                 worker,
+                self.cache(),
             )?;
         } else {
             self.inner.chunk_store.sweep_unused_chunks(
@@ -1768,6 +1769,7 @@ impl DataStore {
                 min_atime,
                 &mut gc_status,
                 worker,
+                None,
             )?;
         }
 
-- 
2.47.3



_______________________________________________
pbs-devel mailing list
pbs-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pbs-devel


  parent reply	other threads:[~2025-11-04 13:07 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-11-04 13:06 [pbs-devel] [PATCH proxmox-backup v2 00/19] fix chunk upload/insert, rename corrupt chunks and GC race conditions for s3 backend Christian Ebner
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 01/19] sync: pull: instantiate backend only once per sync job Christian Ebner
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 02/19] api/datastore: move group notes setting to the datastore Christian Ebner
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 03/19] api/datastore: move snapshot deletion into dedicated datastore helper Christian Ebner
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 04/19] api/datastore: move backup log upload by implementing " Christian Ebner
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 05/19] api: backup: use datastore add_blob helper for backup session Christian Ebner
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 06/19] api/datastore: add dedicated datastore helper to set snapshot notes Christian Ebner
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 07/19] api/datastore: move s3 index upload helper to datastore backend Christian Ebner
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 08/19] datastore: refactor chunk insert based on backend Christian Ebner
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 09/19] verify: rename corrupted to corrupt in log output and function names Christian Ebner
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 10/19] verify/datastore: make rename corrupt chunk a datastore helper method Christian Ebner
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 11/19] datastore: refactor rename_corrupt_chunk error handling Christian Ebner
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 12/19] chunk store: implement per-chunk file locking helper for s3 backend Christian Ebner
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 13/19] datastore: acquire chunk store mutex lock when renaming corrupt chunk Christian Ebner
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 14/19] datastore: get per-chunk file lock for chunk rename on s3 backend Christian Ebner
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 15/19] fix #6961: datastore: verify: evict corrupt chunks from in-memory LRU cache Christian Ebner
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 16/19] datastore: add locking to protect against races on chunk insert for s3 Christian Ebner
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 17/19] GC: fix race with chunk upload/insert on s3 backends Christian Ebner
2025-11-04 13:06 ` Christian Ebner [this message]
2025-11-04 13:06 ` [pbs-devel] [PATCH proxmox-backup v2 19/19] datastore: GC: drop overly verbose info message during s3 chunk sweep Christian Ebner

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=20251104130659.435139-19-c.ebner@proxmox.com \
    --to=c.ebner@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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
Service provided by Proxmox Server Solutions GmbH | Privacy | Legal