[PATCH AUTOSEL 5.4 1/4] btrfs: fix data race at btrfs_use_block_rsv() when accessing block reserve
From: Filipe Manana fdmanana@suse.com
[ Upstream commit c7bb26b847e5b97814f522686068c5628e2b3646 ]
At btrfs_use_block_rsv() we read the size of a block reserve without locking its spinlock, which makes KCSAN complain because the size of a block reserve is always updated while holding its spinlock. The report from KCSAN is the following:
[653.313148] BUG: KCSAN: data-race in btrfs_update_delayed_refs_rsv [btrfs] / btrfs_use_block_rsv [btrfs]
[653.314755] read to 0x000000017f5871b8 of 8 bytes by task 7519 on cpu 0: [653.314779] btrfs_use_block_rsv+0xe4/0x2f8 [btrfs] [653.315606] btrfs_alloc_tree_block+0xdc/0x998 [btrfs] [653.316421] btrfs_force_cow_block+0x220/0xe38 [btrfs] [653.317242] btrfs_cow_block+0x1ac/0x568 [btrfs] [653.318060] btrfs_search_slot+0xda2/0x19b8 [btrfs] [653.318879] btrfs_del_csums+0x1dc/0x798 [btrfs] [653.319702] __btrfs_free_extent.isra.0+0xc24/0x2028 [btrfs] [653.320538] __btrfs_run_delayed_refs+0xd3c/0x2390 [btrfs] [653.321340] btrfs_run_delayed_refs+0xae/0x290 [btrfs] [653.322140] flush_space+0x5e4/0x718 [btrfs] [653.322958] btrfs_preempt_reclaim_metadata_space+0x102/0x2f8 [btrfs] [653.323781] process_one_work+0x3b6/0x838 [653.323800] worker_thread+0x75e/0xb10 [653.323817] kthread+0x21a/0x230 [653.323836] __ret_from_fork+0x6c/0xb8 [653.323855] ret_from_fork+0xa/0x30
[653.323887] write to 0x000000017f5871b8 of 8 bytes by task 576 on cpu 3: [653.323906] btrfs_update_delayed_refs_rsv+0x1a4/0x250 [btrfs] [653.324699] btrfs_add_delayed_data_ref+0x468/0x6d8 [btrfs] [653.325494] btrfs_free_extent+0x76/0x120 [btrfs] [653.326280] __btrfs_mod_ref+0x6a8/0x6b8 [btrfs] [653.327064] btrfs_dec_ref+0x50/0x70 [btrfs] [653.327849] walk_up_proc+0x236/0xa50 [btrfs] [653.328633] walk_up_tree+0x21c/0x448 [btrfs] [653.329418] btrfs_drop_snapshot+0x802/0x1328 [btrfs] [653.330205] btrfs_clean_one_deleted_snapshot+0x184/0x238 [btrfs] [653.330995] cleaner_kthread+0x2b0/0x2f0 [btrfs] [653.331781] kthread+0x21a/0x230 [653.331800] __ret_from_fork+0x6c/0xb8 [653.331818] ret_from_fork+0xa/0x30
So add a helper to get the size of a block reserve while holding the lock. Reading the field while holding the lock instead of using the data_race() annotation is used in order to prevent load tearing.
Signed-off-by: Filipe Manana fdmanana@suse.com Reviewed-by: David Sterba dsterba@suse.com Signed-off-by: David Sterba dsterba@suse.com Signed-off-by: Sasha Levin sashal@kernel.org --- fs/btrfs/block-rsv.c | 2 +- fs/btrfs/block-rsv.h | 16 ++++++++++++++++ 2 files changed, 17 insertions(+), 1 deletion(-)
diff --git a/fs/btrfs/block-rsv.c b/fs/btrfs/block-rsv.c index 36ef3228bac86..63205d2f4d84c 100644 --- a/fs/btrfs/block-rsv.c +++ b/fs/btrfs/block-rsv.c @@ -392,7 +392,7 @@ struct btrfs_block_rsv *btrfs_use_block_rsv(struct btrfs_trans_handle *trans,
block_rsv = get_block_rsv(trans, root);
- if (unlikely(block_rsv->size == 0)) + if (unlikely(btrfs_block_rsv_size(block_rsv) == 0)) goto try_reserve; again: ret = btrfs_block_rsv_use_bytes(block_rsv, blocksize); diff --git a/fs/btrfs/block-rsv.h b/fs/btrfs/block-rsv.h index d1428bb73fc5a..69770360917cb 100644 --- a/fs/btrfs/block-rsv.h +++ b/fs/btrfs/block-rsv.h @@ -98,4 +98,20 @@ static inline void btrfs_unuse_block_rsv(struct btrfs_fs_info *fs_info, btrfs_block_rsv_release(fs_info, block_rsv, 0); }
+/* + * Get the size of a block reserve in a context where getting a stale value is + * acceptable, instead of accessing it directly and trigger data race warning + * from KCSAN. + */ +static inline u64 btrfs_block_rsv_size(struct btrfs_block_rsv *rsv) +{ + u64 ret; + + spin_lock(&rsv->lock); + ret = rsv->size; + spin_unlock(&rsv->lock); + + return ret; +} + #endif /* BTRFS_BLOCK_RSV_H */
From: Ranjan Kumar ranjan.kumar@broadcom.com
[ Upstream commit ee0017c3ed8a8abfa4d40e42f908fb38c31e7515 ]
If the driver detects that the controller is not ready before sending the first IOC facts command, it will wait for a maximum of 10 seconds for it to become ready. However, even if the controller becomes ready within 10 seconds, the driver will still issue a diagnostic reset.
Modify the driver to avoid sending a diag reset if the controller becomes ready within the 10-second wait time.
Signed-off-by: Ranjan Kumar ranjan.kumar@broadcom.com Link: https://lore.kernel.org/r/20240221071724.14986-1-ranjan.kumar@broadcom.com Signed-off-by: Martin K. Petersen martin.petersen@oracle.com Signed-off-by: Sasha Levin sashal@kernel.org --- drivers/scsi/mpt3sas/mpt3sas_base.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/scsi/mpt3sas/mpt3sas_base.c b/drivers/scsi/mpt3sas/mpt3sas_base.c index e72f1dbc91f7f..04fa7337cb1ef 100644 --- a/drivers/scsi/mpt3sas/mpt3sas_base.c +++ b/drivers/scsi/mpt3sas/mpt3sas_base.c @@ -6040,7 +6040,9 @@ _base_wait_for_iocstate(struct MPT3SAS_ADAPTER *ioc, int timeout) return -EFAULT; }
- issue_diag_reset: + return 0; + +issue_diag_reset: rc = _base_diag_reset(ioc); return rc; }
From: Yuxuan Hu 20373622@buaa.edu.cn
[ Upstream commit 2535b848fa0f42ddff3e5255cf5e742c9b77bb26 ]
During our fuzz testing of the connection and disconnection process at the RFCOMM layer, we discovered this bug. By comparing the packets from a normal connection and disconnection process with the testcase that triggered a KASAN report. We analyzed the cause of this bug as follows:
1. In the packets captured during a normal connection, the host sends a `Read Encryption Key Size` type of `HCI_CMD` packet (Command Opcode: 0x1408) to the controller to inquire the length of encryption key.After receiving this packet, the controller immediately replies with a Command Completepacket (Event Code: 0x0e) to return the Encryption Key Size.
2. In our fuzz test case, the timing of the controller's response to this packet was delayed to an unexpected point: after the RFCOMM and L2CAP layers had disconnected but before the HCI layer had disconnected.
3. After receiving the Encryption Key Size Response at the time described in point 2, the host still called the rfcomm_check_security function. However, by this time `struct l2cap_conn *conn = l2cap_pi(sk)->chan->conn;` had already been released, and when the function executed `return hci_conn_security(conn->hcon, d->sec_level, auth_type, d->out);`, specifically when accessing `conn->hcon`, a null-ptr-deref error occurred.
To fix this bug, check if `sk->sk_state` is BT_CLOSED before calling rfcomm_recv_frame in rfcomm_process_rx.
Signed-off-by: Yuxuan Hu 20373622@buaa.edu.cn Signed-off-by: Luiz Augusto von Dentz luiz.von.dentz@intel.com Signed-off-by: Sasha Levin sashal@kernel.org --- net/bluetooth/rfcomm/core.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/net/bluetooth/rfcomm/core.c b/net/bluetooth/rfcomm/core.c index 2db9e285215c5..596fa3172642b 100644 --- a/net/bluetooth/rfcomm/core.c +++ b/net/bluetooth/rfcomm/core.c @@ -1940,7 +1940,7 @@ static struct rfcomm_session *rfcomm_process_rx(struct rfcomm_session *s) /* Get data directly from socket receive queue without copying it. */ while ((skb = skb_dequeue(&sk->sk_receive_queue))) { skb_orphan(skb); - if (!skb_linearize(skb)) { + if (!skb_linearize(skb) && sk->sk_state != BT_CLOSED) { s = rfcomm_recv_frame(s, skb); if (!s) break;
From: Takashi Sakamoto o-takashi@sakamocchi.jp
[ Upstream commit d0b06dc48fb15902d7da09c5c0861e7f042a9381 ]
When resetting the bus after a gap count error, use a long rather than short bus reset.
IEEE 1394-1995 uses only long bus resets. IEEE 1394a adds the option of short bus resets. When video or audio transmission is in progress and a device is hot-plugged elsewhere on the bus, the resulting bus reset can cause video frame drops or audio dropouts. Short bus resets reduce or eliminate this problem. Accordingly, short bus resets are almost always preferred.
However, on a mixed 1394/1394a bus, a short bus reset can trigger an immediate additional bus reset. This double bus reset can be interpreted differently by different nodes on the bus, resulting in an inconsistent gap count after the bus reset. An inconsistent gap count will cause another bus reset, leading to a neverending bus reset loop. This only happens for some bus topologies, not for all mixed 1394/1394a buses.
By instead sending a long bus reset after a gap count inconsistency, we avoid the doubled bus reset, restoring the bus to normal operation.
Signed-off-by: Adam Goldman adamg@pobox.com Link: https://sourceforge.net/p/linux1394/mailman/message/58741624/ Signed-off-by: Takashi Sakamoto o-takashi@sakamocchi.jp Signed-off-by: Sasha Levin sashal@kernel.org --- drivers/firewire/core-card.c | 14 +++++++++++++- 1 file changed, 13 insertions(+), 1 deletion(-)
diff --git a/drivers/firewire/core-card.c b/drivers/firewire/core-card.c index be195ba834632..d446a72629414 100644 --- a/drivers/firewire/core-card.c +++ b/drivers/firewire/core-card.c @@ -500,7 +500,19 @@ static void bm_work(struct work_struct *work) fw_notice(card, "phy config: new root=%x, gap_count=%d\n", new_root_id, gap_count); fw_send_phy_config(card, new_root_id, generation, gap_count); - reset_bus(card, true); + /* + * Where possible, use a short bus reset to minimize + * disruption to isochronous transfers. But in the event + * of a gap count inconsistency, use a long bus reset. + * + * As noted in 1394a 8.4.6.2, nodes on a mixed 1394/1394a bus + * may set different gap counts after a bus reset. On a mixed + * 1394/1394a bus, a short bus reset can get doubled. Some + * nodes may treat the double reset as one bus reset and others + * may treat it as two, causing a gap count inconsistency + * again. Using a long bus reset prevents this. + */ + reset_bus(card, card->gap_count != 0); /* Will allocate broadcast channel after the reset. */ goto out; }
linux-stable-mirror@lists.linaro.org
-
Sasha Levin