This has theoretical bug if our collection spans different
filesystems and some filesystems support holes, and some not.
If this ever encountered we should use pathconf(2) and cache
its result for directories.
My guess the goal was to skip files with holes in them. Original commit
590f0761f4aba64c4a3b4c8f11543a929178da38, doesn't explain that. If that
is still necessary, better to add lseek(SEEK_DATA) check later.
Some effort was made to unify monitoring via kqueue and via inotify
APIs. Now both provide their implementation of add_watch() function.
I guess there are some logical bugs in vnode_process(). With this commit
it would be better provide code as is, and resolve bugs separately.
The idea is taken from the nginx web server, but much simplified and
almost no copypaste left. This will allow minidlna to use different
event dispatcher APIs, which would be defined at compile time.
My personal goal is to convert minidlna to kqueue(2) on FreeBSD. This
would later allow for kqueue based directory change notification, which
won't conflict with select(2) like the current patch does.
Other platforms will also benefit from the pluggability of the event
system, Linux can switch to epoll(2) or at least to poll(2).
Detailed list of changes:
* event.h [New]
Our internal API to unify different event dispatch systems.
* select.c [New]
Much simplified version of nginx's ngx_select_module.c.
* minidlna.c
- Split out listen socket event processing into separate function
ProcessListen(), which matches event_process_t type.
- Create and initialize struct event for the monitor socket, SSDP
socket, HTTP socket and beacon socket.
- Simplify and make more precise timeout calculation using
helper timeval functions from utils.c. Treat gettimeofday() error
as a fatal event.
- Rip out all stuff related to select(2). Just call event_module.process().
* upnpevents.c
- Embed struct event into upnp_event_notify.
- Merge upnp_event_create_notify() with upnp_event_notify_connect().
Start connecting immediately after socket creation. Garbage collect
now useless ECreated state.
- Make upnp_event_process_notify() of event_process_t type, and use it
as process callback for upnp_event_notify event.
- Looks like we always create upnp_event_notify with existing subscriber,
and never clear it later. Remove checks for obj->sub and assert that it
is never NULL. Simplifies things.
- When switching obj state, add/del it to event dispatcher accrodingly.
- Garbage collect upnpevents_selectfds().
- Garbage collect select(2) related stuff from upnpevents_processfds().
Rename function to upnpevents_gc(), since the remaining functionality
is garbage collecting, not file descriptor processing.
Actually, this can be simplified even more. We can safely close sockets
and free objects immediately, eliminating need for upnpevents_gc(). But
this change would be beyond scope of this commit.
* upnphttp.c, upnphttp.h
Embed struct event into struct upnphttp. Adjust Process_upnphttp() to match
event_process_t type. Add/del to event dispatcher once creating/closing a
socket.
* minissdp.c, minissdp.h
Make ProcessSSDPRequest() of event_process_t type.
* getifaddr.c, getifaddr.h
Make ProcessMonitorEvent() of event_process_t type.
Add the ability to force title modification for clients that do their own
alphanumeric sorting. Adding a '!' to the beginning of the
force_sort_criteria value will enable this behavior.
It is implemented by prepending all titles with a zero-padded number, and
possibly season and episode (or disc and track) numbers if they aren't
already in the title.
Fix several issues with the non-destructive rescan functionality.
Most of these issues also affected inotify scanning as well. These
include annoying debug messages, adding album art for files that we
aren't supposed to be scanning anyway, incrementing the UpdateID when no
changes were made to the database, and other smaller issues.
Some Samsung clients (at least UE40ES8000) apparently cannot handle our
standard container IDs when we turn on the DCM10 feature. So they end up
requesting an empty ObjectID, which is invalid. If we blindly return our
root container, they still have problems browsing because they only use
one object from our returned results, so they don't get a full list of
browseable categories.
It has been confirmed that we can work around this by using the same
ObjectIDs that Samsung uses for our root containers in our
X_GetFeatureList response (A, V, and I). And fortunately it's trivial to
handle these renamed ObjectIDs using our magic container system.
Fixes: SF Bug #311 (Samsung TV built-in DLNA client broken since v1.1.5)
Protect against DoS by limiting the number of subscribers we allow to a
total of 500, and immediately removing subscribers from our list if we
encounter an error communicating with them.
Bolt (Series6) seems to have a bug where it only reacts to
_tivo-videostream._tcp service types after a TiVo reboot, and ignores
_tivo-vidoes._tcp. I have no idea what the difference is supposed to be
because it's undocumented, but either type works on both Roamio (Series5)
and Bolt, so hopefully it's safe for Series 3 and 4 also.
On Darwin & *BSD binding to mcast address may force NITIFY packet send from
this mcast address, not from primary interface address. Therefore subsequent
NOTIFY packets might be ignored by some clients, including Samsung Smart TV
and they will stop & disconnect after initial max-age timeout.
SO_BROADCAST on notify packets is not needed, because there's no broadcasted
M-SEARCH packets and notify socket queue is never read.
