All of lore.kernel.org
 help / color / mirror / Atom feed
From: Simon Mikuda <simon.mikuda@streamunlimited.com>
To: Luiz Augusto von Dentz <luiz.dentz@gmail.com>
Cc: linux-bluetooth@vger.kernel.org
Subject: Re: [PATCH] advertising: Fix crash when advertisement is unregistered while MGMT ADD_ADVERTISEMENT call
Date: Mon, 4 Nov 2019 10:50:23 +0100	[thread overview]
Message-ID: <f4a64217-102e-3504-3aac-44dc062bc2e8@streamunlimited.com> (raw)
In-Reply-To: <CABBYNZ+SvYgxdr51rr_cYOWczLWLfRUz_xMJkJJrEv2gWSer5w@mail.gmail.com>

Hi Luiz!

Using mgmt_cancel didn't came to my mind, but it makes more sense.

Advantage about using refcount is that you will always have 
advertisement instance_id from kernel callback so you can use 
advertising MGMT directly from e.g. btmgmt and it will still work ok. 
Also I could remove g_idle_add(client_free_idle_cb, client); which is 
little bit tricky to use and it is easy to make double free when editing 
- there is also a bug about this but it is harder to reproduce.

I will send new patches with updated description.

Thank you!

BR,
Simon

On 01/11/2019 15:49, Luiz Augusto von Dentz wrote:
> Hi Simon,
>
> On Thu, Oct 31, 2019 at 9:45 AM Simon Mikuda
> <simon.mikuda@streamunlimited.com> wrote:
>
> Subject should be less than 72 characters so it first nicely on git
> shortlog, you can add a short description before the backtrace.
>
>> bluetoothd[29698]: src/advertising.c:register_advertisement() RegisterAdvertisement
>> bluetoothd[29698]: src/advertising.c:client_create() Adding proxy for /org/bluez/example/advertisement0
>> bluetoothd[29698]: src/advertising.c:register_advertisement() Registered advertisement at path /org/bluez/example/advertisement0
>> bluetoothd[29698]: src/advertising.c:parse_service_uuids() Adding ServiceUUID: 180D
>> bluetoothd[29698]: src/advertising.c:parse_service_uuids() Adding ServiceUUID: 180F
>> bluetoothd[29698]: src/advertising.c:parse_manufacturer_data() Adding ManufacturerData for ffff
>> bluetoothd[29698]: src/advertising.c:parse_service_data() Adding ServiceData for 9999
>> bluetoothd[29698]: src/advertising.c:parse_data() Adding Data for type 0x26 len 3
>> bluetoothd[29698]: src/advertising.c:refresh_adv() Refreshing advertisement: /org/bluez/example/advertisement0
>> bluetoothd[29698]: src/advertising.c:unregister_advertisement() UnregisterAdvertisement
>> bluetoothd[29698]: src/advertising.c:add_adv_callback() Advertisement registered: �
>> Segmentation fault (core dumped)
>> ---
>>   src/advertising.c | 53 +++++++++++++++++++++++++++++------------------------
>>   1 file changed, 29 insertions(+), 24 deletions(-)
>>
>> diff --git a/src/advertising.c b/src/advertising.c
>> index 3ed1376..7d8cadf 100644
>> --- a/src/advertising.c
>> +++ b/src/advertising.c
>> @@ -63,6 +63,8 @@ struct btd_adv_manager {
>>   #define AD_TYPE_PERIPHERAL 1
>>
>>   struct btd_adv_client {
>> +       int ref_count;
>> +
>>          struct btd_adv_manager *manager;
>>          char *owner;
>>          char *path;
>> @@ -136,13 +138,6 @@ static void client_free(void *data)
>>          free(client);
>>   }
>>
>> -static gboolean client_free_idle_cb(void *data)
>> -{
>> -       client_free(data);
>> -
>> -       return FALSE;
>> -}
>> -
>>   static void client_release(void *data)
>>   {
>>          struct btd_adv_client *client = data;
>> @@ -175,12 +170,19 @@ static void remove_advertising(struct btd_adv_manager *manager,
>>                          manager->mgmt_index, sizeof(cp), &cp, NULL, NULL, NULL);
>>   }
>>
>> -static void client_remove(void *data)
>> +static struct btd_adv_client *client_ref(struct btd_adv_client *client)
>> +{
>> +       __sync_fetch_and_add(&client->ref_count, 1);
>> +
>> +       return client;
>> +}
>> +
>> +static void client_unref(struct btd_adv_client *client)
>>   {
>> -       struct btd_adv_client *client = data;
>>          struct mgmt_cp_remove_advertising cp;
>>
>> -       g_dbus_client_set_disconnect_watch(client->client, NULL, NULL);
>> +       if (__sync_sub_and_fetch(&client->ref_count, 1))
>> +               return;
> While the idea of using refcount to protect might seem good I think we
> would be better off using mgmt_cancel for this case so the callback is
> not called, or you are not using it on purpose?
>
>>          cp.instance = client->instance;
>>
>> @@ -188,10 +190,6 @@ static void client_remove(void *data)
>>                          client->manager->mgmt_index, sizeof(cp), &cp,
>>                          NULL, NULL, NULL);
>>
>> -       queue_remove(client->manager->clients, client);
>> -
>> -       g_idle_add(client_free_idle_cb, client);
>> -
>>          g_dbus_emit_property_changed(btd_get_dbus_connection(),
>>                                  adapter_get_path(client->manager->adapter),
>>                                  LE_ADVERTISING_MGR_IFACE, "SupportedInstances");
>> @@ -199,13 +197,18 @@ static void client_remove(void *data)
>>          g_dbus_emit_property_changed(btd_get_dbus_connection(),
>>                                  adapter_get_path(client->manager->adapter),
>>                                  LE_ADVERTISING_MGR_IFACE, "ActiveInstances");
>> +
>> +       queue_remove(client->manager->clients, client);
>> +
>> +       client_free(client);
>> +
>>   }
>>
>>   static void client_disconnect_cb(DBusConnection *conn, void *user_data)
>>   {
>>          DBG("Client disconnected");
>>
>> -       client_remove(user_data);
>> +       client_unref(user_data);
>>   }
>>
>>   static bool parse_type(DBusMessageIter *iter, struct btd_adv_client *client)
>> @@ -564,7 +567,7 @@ static gboolean client_timeout(void *user_data)
>>          client->to_id = 0;
>>
>>          client_release(client);
>> -       client_remove(client);
>> +       client_unref(client);
>>
>>          return FALSE;
>>   }
>> @@ -963,8 +966,6 @@ static void add_client_complete(struct btd_adv_client *client, uint8_t status)
>>                                                  mgmt_errstr(status), status);
>>                  reply = btd_error_failed(client->reg,
>>                                          "Failed to register advertisement");
>> -               queue_remove(client->manager->clients, client);
>> -               g_idle_add(client_free_idle_cb, client);
>>
>>          } else
>>                  reply = dbus_message_new_method_return(client->reg);
>> @@ -972,6 +973,8 @@ static void add_client_complete(struct btd_adv_client *client, uint8_t status)
>>          g_dbus_send_message(btd_get_dbus_connection(), reply);
>>          dbus_message_unref(client->reg);
>>          client->reg = NULL;
>> +
>> +       client_unref(client);
>>   }
>>
>>   static void add_adv_callback(uint8_t status, uint16_t length,
>> @@ -1060,8 +1063,11 @@ static DBusMessage *parse_advertisement(struct btd_adv_client *client)
>>          }
>>
>>          err = refresh_adv(client, add_adv_callback);
>> -       if (!err)
>> +       if (!err) {
>> +               /* Hold reference to client until add_adv_callback is finished */
>> +               client_ref(client);
>>                  return NULL;
>> +       }
>>
>>   fail:
>>          return btd_error_failed(client->reg, "Failed to parse advertisement.");
>> @@ -1082,14 +1088,12 @@ static void client_proxy_added(GDBusProxy *proxy, void *data)
>>                  return;
>>
>>          /* Failed to publish for some reason, remove. */
>> -       queue_remove(client->manager->clients, client);
>> -
>> -       g_idle_add(client_free_idle_cb, client);
>> -
>>          g_dbus_send_message(btd_get_dbus_connection(), reply);
>>
>>          dbus_message_unref(client->reg);
>>          client->reg = NULL;
>> +
>> +       client_unref(client);
>>   }
>>
>>   static struct btd_adv_client *client_create(struct btd_adv_manager *manager,
>> @@ -1189,6 +1193,7 @@ static DBusMessage *register_advertisement(DBusConnection *conn,
>>          DBG("Registered advertisement at path %s", match.path);
>>
>>          queue_push_tail(manager->clients, client);
>> +       client_ref(client);
>>
>>          return NULL;
>>   }
>> @@ -1218,7 +1223,7 @@ static DBusMessage *unregister_advertisement(DBusConnection *conn,
>>          if (!client)
>>                  return btd_error_does_not_exist(msg);
>>
>> -       client_remove(client);
>> +       client_unref(client);
>>
>>          return dbus_message_new_method_return(msg);
>>   }
>> --
>> 2.7.4
>>
>
-- 
StreamUnlimited Engineering GmbH
Levocska 9, 851 01 Bratislava, Slovakia
Office: +42 1 2 6720 0088
simon.mikuda@streamunlimited.com,www.streamunlimited.com


      reply	other threads:[~2019-11-04  9:50 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-10-31  6:52 [PATCH] advertising: Fix crash when advertisement is unregistered while MGMT ADD_ADVERTISEMENT call Simon Mikuda
2019-11-01 14:49 ` Luiz Augusto von Dentz
2019-11-04  9:50   ` Simon Mikuda [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=f4a64217-102e-3504-3aac-44dc062bc2e8@streamunlimited.com \
    --to=simon.mikuda@streamunlimited.com \
    --cc=linux-bluetooth@vger.kernel.org \
    --cc=luiz.dentz@gmail.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.