kernel.git/fs/quota, branch linux-6.10.y

quota: avoid missing put_quota_format when DQUOT_SUSPENDED is passed

2024-10-04T14:33:03Z

[ Upstream commit d16a5f852025be546b6e4ceef15899db3490f4d7 ] Avoid missing put_quota_format when DQUOT_SUSPENDED is passed to dquot_load_quota_sb. Link: https://patch.msgid.link/20240715130534.2112678-2-shikemeng@huaweicloud.com Signed-off-by: Kemeng Shi Fixes: d44c57663723 ("quota: Remove BUG_ON in dquot_load_quota_sb()") Reviewed-by: Joseph Qi Signed-off-by: Jan Kara Signed-off-by: Sasha Levin

quota: fix to propagate error of mark_dquot_dirty() to caller

2024-04-12T12:52:29Z

in order to let caller be aware of failure of mark_dquot_dirty(). Signed-off-by: Chao Yu Signed-off-by: Jan Kara Message-Id: <20240412094942.2131243-1-chao@kernel.org>

fs: quota: use group allocation of per-cpu counters API

2024-03-25T19:09:17Z

Use group allocation of per-cpu counters api to accelerate dquot_init() and simplify code. Signed-off-by: Kefeng Wang Signed-off-by: Jan Kara Message-Id: <20240325041240.53537-1-wangkefeng.wang@huawei.com>

quota: remove SLAB_MEM_SPREAD flag usage

2024-03-05T14:15:50Z

The SLAB_MEM_SPREAD flag is already a no-op after removal of SLAB allocator and in [1] it was fully deprecated. Remove its usage so we can delete it from slab. No functional change. Signed-off-by: Chengming Zhou Signed-off-by: Jan Kara Link: https://lore.kernel.org/all/20240223-slab-cleanup-flags-v2-1-02f1753e8303@suse.cz/ Message-Id: <20240224135118.830073-1-chengming.zhou@linux.dev>

quota: Detect loops in quota tree

2024-02-12T13:14:49Z

Syzbot has found that when it creates corrupted quota files where the quota tree contains a loop, we will deadlock when tryling to insert a dquot. Add loop detection into functions traversing the quota tree. Signed-off-by: Jan Kara

quota: Properly annotate i_dquot arrays with __rcu

2024-02-08T11:04:59Z

Dquots pointed to from i_dquot arrays in inodes are protected by dquot_srcu. Annotate them as such and change .get_dquots callback to return properly annotated pointer to make sparse happy. Fixes: b9ba6f94b238 ("quota: remove dqptr_sem") Signed-off-by: Jan Kara

quota: Fix rcu annotations of inode dquot pointers

2024-02-08T11:04:59Z

Dquot pointers in i_dquot array in the inode are protected by dquot_srcu. Annotate the array pointers with __rcu, perform the locked dereferences with srcu_dereference_check() instead of plain reads, and set the array elements with rcu_assign_pointer(). Fixes: b9ba6f94b238 ("quota: remove dqptr_sem") Reported-by: kernel test robot Closes: https://lore.kernel.org/oe-kbuild-all/202402061900.rTuYDlo6-lkp@intel.com/ Signed-off-by: Jan Kara

quota: Fix potential NULL pointer dereference

2024-02-05T14:57:34Z

Below race may cause NULL pointer dereference P1 P2 dquot_free_inode quota_off drop_dquot_ref remove_dquot_ref dquots = i_dquot(inode) dquots = i_dquot(inode) srcu_read_lock dquots[cnt]) != NULL (1) dquots[type] = NULL (2) spin_lock(&dquots[cnt]->dq_dqb_lock) (3) .... If dquot_free_inode(or other routines) checks inode's quota pointers (1) before quota_off sets it to NULL(2) and use it (3) after that, NULL pointer dereference will be triggered. So let's fix it by using a temporary pointer to avoid this issue. Signed-off-by: Wang Jianjian Signed-off-by: Jan Kara Message-Id: <20240202081852.2514092-1-wangjianjian3@huawei.com>

quota: Drop GFP_NOFS instances under dquot->dq_lock and dqio_sem

2024-01-23T18:21:11Z

Quota code acquires dquot->dq_lock whenever reading / writing dquot. When reading / writing quota info we hold dqio_sem. Since these locks can be acquired during inode reclaim (through dquot_drop() -> dqput() -> dquot_release()) we are setting nofs allocation context whenever acquiring these locks. Hence there's no need to use GFP_NOFS allocations in quota code doing IO. Just switch it to GFP_KERNEL. Signed-off-by: Jan Kara

quota: Set nofs allocation context when acquiring dqio_sem

2024-01-23T18:21:11Z

dqio_sem can be acquired during inode reclaim through dquot_drop() -> dqput() -> dquot_release() -> write_file_info() context. In some cases (most notably through dquot_get_next_id()) it can be acquired without holding dquot->dq_lock (which already sets nofs allocation context). So we need to set nofs allocation context when acquiring it as well. Signed-off-by: Jan Kara