| Message ID | 20191027221007.14317-1-navid.emamdoost@gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | ALSA: usb-audio: Fix memory leak in __snd_usbmidi_create | expand |
Thanks for the explanation, On Mon, Oct 28, 2019 at 1:27 AM Takashi Iwai <tiwai@suse.de> wrote: > > On Sun, 27 Oct 2019 23:10:06 +0100, > Navid Emamdoost wrote: > > > > In the implementation of __snd_usbmidi_create() there is a memory leak > > caused by incorrect goto destination. Go to free_midi if > > snd_usbmidi_create_endpoints_midiman() or snd_usbmidi_create_endpoints() > > fail. > > No, this will lead to double-free. After registering the rawmidi > interface at snd_usbmidi_create_rawmidi(), the common destructor will > be called via rawmidi private_free callback, and this will release the > all resources already. Now I can see how rawmidi private_free is set up to release the resources, but what concerns me is that at the moment of endpoint/port creation umidi is not yet added to the midi_list. In other words, what I see is that we still have just one local pointer to umidi if any of snd_usbmidi_create_endpoint* fail. Am I missing something? > > > thanks, > > Takashi > > > > > Fixes: 731209cc0417 ("ALSA: usb-midi: Use common error handling code in __snd_usbmidi_create()") > > Signed-off-by: Navid Emamdoost <navid.emamdoost@gmail.com> > > --- > > sound/usb/midi.c | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > diff --git a/sound/usb/midi.c b/sound/usb/midi.c > > index b737f0ec77d0..22db37fbfbbd 100644 > > --- a/sound/usb/midi.c > > +++ b/sound/usb/midi.c > > @@ -2476,7 +2476,7 @@ int __snd_usbmidi_create(struct snd_card *card, > > else > > err = snd_usbmidi_create_endpoints(umidi, endpoints); > > if (err < 0) > > - goto exit; > > + goto free_midi; > > > > usb_autopm_get_interface_no_resume(umidi->iface); > > > > -- > > 2.17.1 > >
On Mon, 28 Oct 2019 17:25:41 +0100, Navid Emamdoost wrote: > > Thanks for the explanation, > > On Mon, Oct 28, 2019 at 1:27 AM Takashi Iwai <tiwai@suse.de> wrote: > > > > On Sun, 27 Oct 2019 23:10:06 +0100, > > Navid Emamdoost wrote: > > > > > > In the implementation of __snd_usbmidi_create() there is a memory leak > > > caused by incorrect goto destination. Go to free_midi if > > > snd_usbmidi_create_endpoints_midiman() or snd_usbmidi_create_endpoints() > > > fail. > > > > No, this will lead to double-free. After registering the rawmidi > > interface at snd_usbmidi_create_rawmidi(), the common destructor will > > be called via rawmidi private_free callback, and this will release the > > all resources already. > Now I can see how rawmidi private_free is set up to release the > resources, but what concerns me is that at the moment of endpoint/port > creation umidi is not yet added to the midi_list. > In other words, what I see is that we still have just one local > pointer to umidi if any of snd_usbmidi_create_endpoint* fail. > Am I missing something? The rawmidi object that is created via snd_rawmidi_new() is managed via snd_device list, and it's traversed at snd_card_disconnect() and snd_card_free() calls. It's something like devm-stuff (but implemented in a different way). The midi_list is an explicit list for the USB MIDI driver, and it's an individual one from the device list. Takashi > > > > > > > > thanks, > > > > Takashi > > > > > > > > Fixes: 731209cc0417 ("ALSA: usb-midi: Use common error handling code in __snd_usbmidi_create()") > > > Signed-off-by: Navid Emamdoost <navid.emamdoost@gmail.com> > > > --- > > > sound/usb/midi.c | 2 +- > > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > > > diff --git a/sound/usb/midi.c b/sound/usb/midi.c > > > index b737f0ec77d0..22db37fbfbbd 100644 > > > --- a/sound/usb/midi.c > > > +++ b/sound/usb/midi.c > > > @@ -2476,7 +2476,7 @@ int __snd_usbmidi_create(struct snd_card *card, > > > else > > > err = snd_usbmidi_create_endpoints(umidi, endpoints); > > > if (err < 0) > > > - goto exit; > > > + goto free_midi; > > > > > > usb_autopm_get_interface_no_resume(umidi->iface); > > > > > > -- > > > 2.17.1 > > > > > > > -- > Navid. >
Navid Emamdoost wrote: > On Mon, Oct 28, 2019 at 1:27 AM Takashi Iwai <tiwai@suse.de> wrote: >> On Sun, 27 Oct 2019 23:10:06 +0100, >> Navid Emamdoost wrote: >>> In the implementation of __snd_usbmidi_create() there is a memory leak >>> caused by incorrect goto destination. Go to free_midi if >>> snd_usbmidi_create_endpoints_midiman() or snd_usbmidi_create_endpoints() >>> fail. >> >> No, this will lead to double-free. After registering the rawmidi >> interface at snd_usbmidi_create_rawmidi(), the common destructor will >> be called via rawmidi private_free callback, and this will release the >> all resources already. > > Now I can see how rawmidi private_free is set up to release the > resources, but what concerns me is that at the moment of endpoint/port > creation umidi is not yet added to the midi_list. > In other words, what I see is that we still have just one local > pointer to umidi if any of snd_usbmidi_create_endpoint* fail. The snd_rawmidi device is automatically registered with the card, so the line "rmidi->private_free = snd_usbmidi_rawmidi_free;" is the exact point where the ownership of umidi changes. midi_list does not matter. Regards, Clemens
On Mon, Oct 28, 2019 at 11:38 AM Takashi Iwai <tiwai@suse.de> wrote: > > On Mon, 28 Oct 2019 17:25:41 +0100, > Navid Emamdoost wrote: > > > > Thanks for the explanation, > > > > On Mon, Oct 28, 2019 at 1:27 AM Takashi Iwai <tiwai@suse.de> wrote: > > > > > > On Sun, 27 Oct 2019 23:10:06 +0100, > > > Navid Emamdoost wrote: > > > > > > > > In the implementation of __snd_usbmidi_create() there is a memory leak > > > > caused by incorrect goto destination. Go to free_midi if > > > > snd_usbmidi_create_endpoints_midiman() or snd_usbmidi_create_endpoints() > > > > fail. > > > > > > No, this will lead to double-free. After registering the rawmidi > > > interface at snd_usbmidi_create_rawmidi(), the common destructor will > > > be called via rawmidi private_free callback, and this will release the > > > all resources already. > > Now I can see how rawmidi private_free is set up to release the > > resources, but what concerns me is that at the moment of endpoint/port > > creation umidi is not yet added to the midi_list. > > In other words, what I see is that we still have just one local > > pointer to umidi if any of snd_usbmidi_create_endpoint* fail. > > Am I missing something? > > The rawmidi object that is created via snd_rawmidi_new() is managed > via snd_device list, and it's traversed at snd_card_disconnect() and > snd_card_free() calls. It's something like devm-stuff (but > implemented in a different way). The midi_list is an explicit list > for the USB MIDI driver, and it's an individual one from the device > list. Thanks for the clarification. > > > Takashi > > > > > > > > > > > > > > thanks, > > > > > > Takashi > > > > > > > > > > > Fixes: 731209cc0417 ("ALSA: usb-midi: Use common error handling code in __snd_usbmidi_create()") > > > > Signed-off-by: Navid Emamdoost <navid.emamdoost@gmail.com> > > > > --- > > > > sound/usb/midi.c | 2 +- > > > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > > > > > diff --git a/sound/usb/midi.c b/sound/usb/midi.c > > > > index b737f0ec77d0..22db37fbfbbd 100644 > > > > --- a/sound/usb/midi.c > > > > +++ b/sound/usb/midi.c > > > > @@ -2476,7 +2476,7 @@ int __snd_usbmidi_create(struct snd_card *card, > > > > else > > > > err = snd_usbmidi_create_endpoints(umidi, endpoints); > > > > if (err < 0) > > > > - goto exit; > > > > + goto free_midi; > > > > > > > > usb_autopm_get_interface_no_resume(umidi->iface); > > > > > > > > -- > > > > 2.17.1 > > > > > > > > > > > > -- > > Navid. > >
diff --git a/sound/usb/midi.c b/sound/usb/midi.c index b737f0ec77d0..22db37fbfbbd 100644 --- a/sound/usb/midi.c +++ b/sound/usb/midi.c @@ -2476,7 +2476,7 @@ int __snd_usbmidi_create(struct snd_card *card, else err = snd_usbmidi_create_endpoints(umidi, endpoints); if (err < 0) - goto exit; + goto free_midi; usb_autopm_get_interface_no_resume(umidi->iface);
In the implementation of __snd_usbmidi_create() there is a memory leak caused by incorrect goto destination. Go to free_midi if snd_usbmidi_create_endpoints_midiman() or snd_usbmidi_create_endpoints() fail. Fixes: 731209cc0417 ("ALSA: usb-midi: Use common error handling code in __snd_usbmidi_create()") Signed-off-by: Navid Emamdoost <navid.emamdoost@gmail.com> --- sound/usb/midi.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)