[PATCH v1] src/device: Fix stored gatt cache DB Hash value not update

[PATCH v1] src/device: Fix stored gatt cache DB Hash value not update

[Mengshi Wu]

There is an asymmetry in behavior: when services are added during
the same connection (via Service Changed indication), the persistent
storage (disk) is not updated with the new DB hash, but when services
are removed, it is updated.

During the same connection, We check DB hash value stored at
/var/lib/bluetooth/<adaptor addr>/cache/<remote addr>.
When established connection, the stored DB Hash value is A.Then we
add new services, the stored DB Hash value is still A which should
change to B. However, if we remove the existing services, the stored
DB Hash value changed to C.

When performing addition, it goes like this:

discover_primary_cb()
  └─> gatt_db_insert_service()      ← NEW service inserted into db
        └─> gatt_service_added()    ← callback fires immediately
              └─> store_gatt_db()   ← SAVED TO DISK (hash still OLD)
  ...
  └─> discovery_op_complete(success=true)
        └─> read_db_hash(op)             ← sends ATT Read By Type
              └─> [ATT response arrives]
                    └─> db_hash_read_cb()
                          ├─> gatt_db_attribute_write(op->hash, ...)
                          │     └─> hash UPDATED IN MEMORY
                          └─> discovery_op_complete(true, 0)
                                ├─> [no services to remove, no
                                │    store_gatt_db called]
                                └─> service_changed_complete()

Whereas removal perform like this:
discovery_op_complete(success=true)  [1st call]
  └─> read_db_hash(op)
      └─> op->hash is NULL → sends ATT request → early return
...
[ATT response arrives]
db_hash_read_cb()
  └─> gatt_db_attribute_write(op->hash, ) ← hash UPDATED IN MEMORY
  └─> discovery_op_complete(true, 0)          [2nd call]
      └─> read_db_hash(op)  → op->hash already set → returns false
      └─> gatt_db_remove_service()
          └─> gatt_service_removed()
              └─> store_gatt_db() ← SAVED TO DISK (hash is NEW)

There is a timing issue to update DB Hash value.

The gatt_client_service_changed() callback in src/device.c is called
from service_changed_complete() in gatt-client.c, which is invoked
after db_hash_read_cb() has already updated the hash. Adding
store_gatt_db(device) here guarantees the db is persisted with the
correct, up-to-date hash for both the addition and removal cases.

Signed-off-by: Mengshi Wu <mengshi.wu@oss.qualcomm.com>
---
 src/device.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/src/device.c b/src/device.c
index 3ea683667..cfbde307b 100644
--- a/src/device.c
+++ b/src/device.c
@@ -6267,7 +6267,11 @@ static void gatt_client_service_changed(uint16_t start_handle,
 							uint16_t end_handle,
 							void *user_data)
 {
+	struct btd_device *device = user_data;
+
 	DBG("start 0x%04x, end: 0x%04x", start_handle, end_handle);
+
+	store_gatt_db(device);
 }
 
 static void gatt_debug(const char *str, void *user_data)
-- 
2.34.1

RE: [v1] src/device: Fix stored gatt cache DB Hash value not update

[bluez.test.bot]

[-- Attachment #1: Type: text/plain, Size: 1311 bytes --]

This is automated email and please do not reply to this email!

Dear submitter,

Thank you for submitting the patches to the linux bluetooth mailing list.
This is a CI test results with your patch series:
PW Link:https://patchwork.kernel.org/project/bluetooth/list/?series=1075787

---Test result---

Test Summary:
CheckPatch                    PENDING   0.43 seconds
GitLint                       PENDING   0.39 seconds
BuildEll                      PASS      20.18 seconds
BluezMake                     PASS      641.23 seconds
MakeCheck                     PASS      18.36 seconds
MakeDistcheck                 PASS      246.04 seconds
CheckValgrind                 PASS      295.39 seconds
CheckSmatch                   PASS      350.55 seconds
bluezmakeextell               PASS      184.26 seconds
IncrementalBuild              PENDING   0.41 seconds
ScanBuild                     PASS      1028.84 seconds

Details
##############################
Test: CheckPatch - PENDING
Desc: Run checkpatch.pl script
Output:

##############################
Test: GitLint - PENDING
Desc: Run gitlint
Output:

##############################
Test: IncrementalBuild - PENDING
Desc: Incremental build with the patches in the series
Output:



https://github.com/bluez/bluez/pull/1998/checks

---
Regards,
Linux Bluetooth

Re: [PATCH v1] src/device: Fix stored gatt cache DB Hash value not update

[patchwork-bot+bluetooth]

Hello:

This patch was applied to bluetooth/bluez.git (master)
by Luiz Augusto von Dentz <luiz.von.dentz@intel.com>:

On Wed,  1 Apr 2026 19:39:30 +0800 you wrote:
> There is an asymmetry in behavior: when services are added during
> the same connection (via Service Changed indication), the persistent
> storage (disk) is not updated with the new DB hash, but when services
> are removed, it is updated.
> 
> During the same connection, We check DB hash value stored at
> /var/lib/bluetooth/<adaptor addr>/cache/<remote addr>.
> When established connection, the stored DB Hash value is A.Then we
> add new services, the stored DB Hash value is still A which should
> change to B. However, if we remove the existing services, the stored
> DB Hash value changed to C.
> 
> [...]

Here is the summary with links:
  - [v1] src/device: Fix stored gatt cache DB Hash value not update
    https://git.kernel.org/pub/scm/bluetooth/bluez.git/?id=0fd01e98cf94

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html