ALSA: aw2-alsa: select SND_PCM
PATCH 1/1 ALSA: aw2-alsa: select SND_PCM
Several stanzas in sound/pci/Kconfig select SND_PCM. SND_AW2 selects nothing at all.
aw2-alsa calls snd_pcm_new, which sound/core/pcm.c provides under SND_PCM.
aw2-alsa.c:(.text+0x554): undefined reference to `snd_pcm_new'
4 symbols are undefined in all.
The driver can be built into vmlinux while SND_PCM is a module, in which case the symbols are not there at link time.
Select rather than depend on: SND_PCM has no prompt of its own and is only ever enabled by a select, so a depends on would leave SND_AW2 unselectable.
Reproduced on v7.3, arm64, with CONFIG_SND_AW2=y and CONFIG_SND_PCM=m.
Found with kreach, a static check that compares what Kconfig allows against the symbols the compiled objects actually reference.
--- sound/pci/Kconfig | 1 + 1 file changed, 1 insertion(+)
sound/pci/Kconfig +1 -0
@@ -141,6 +141,7 @@ config SND_AU8830 config SND_AW2 tristate "Emagic Audiowerk 2" + select SND_PCM help Say Y here to include support for Emagic Audiowerk 2 soundcards.
Signature · 1 line
2.43.0
Quoted 17 lines · Jude Gyimah
Several stanzas in
sound/pci/KconfigselectSND_PCM.SND_AW2selects nothing at all.aw2-alsa calls
snd_pcm_new, whichsound/core/pcm.cprovides underSND_PCM.aw2-alsa.c:(.text+0x554): undefined reference to `snd_pcm_new'
4 symbols are undefined in all.
The driver can be built into vmlinux while
SND_PCMis a module, in which case the symbols are not there at link time.Select rather than depend on:
SND_PCMhas no prompt of its own and is only ever enabled by a select, so a depends on would leaveSND_AW2unselectable.
Quoted 7 lines
Reproduced on v7.3, arm64, with
CONFIG_SND_AW2=yandCONFIG_SND_PCM=m.Found with kreach, a static check that compares what
Kconfigallows against the symbols the compiled objects actually reference.Fixes 98f2a97f207a ("[ALSA] Emagic Audiowerk 2 ALSA driver.")Signed-off-by Jude Gyimah <[email protected]>
Applied now. It's interesting why this wasn't caught by build tests for long time...
thanks,
Takashi
Quoted 2 lines
Applied now. It's interesting why this wasn't caught by build tests for long time...
Thanks for applying it.
Yes i agree. Quite interesting that this flew under the radar. The short answer is that none of the stock configurations can express the combination, and each one misses it for a different reason. On arm64 at v7.3:
defconfig SND_AW2 not set SND_PCM=m caller absent allmodconfig SND_AW2=m SND_PCM=m caller is a module too allyesconfig SND_AW2=y SND_PCM=y provider built in allnoconfig neither set
The failure needs SND_AW2=y while SND_PCM=m, and allmodconfig and allyesconfig are each wrong in exactly one coordinate.
randconfig can express it but is unlikely to find it. And the reason is kinda counter-intuitive i.e., 110 stanzas in 44 Kconfig files select SND_PCM. Any one of them at =y pins SND_PCM=y and hides every caller that does not select it. So the configuration has to put all 110 at m or n and this driver at y. Treating the tristates as independent and uniform (which randconfig and tristates are not, since most sit behind PCI or a platform) that is on the order of (2/3)^110. The real figure will be different, but the shape is right. The more drivers select a provider, the better hidden the one driver that forgot to.
So SND_PCM being selected by almost everything is what protected this bug rather than what would have exposed it. A sampler would have to be lucky in 110 places at once. Solving for the assignment instead takes it in one step, which is what turned this up.
It is not a one-off. The same y/m shape found 14 missing selects in sound/, all of them now applied i.e., five Rockchip, three fsl, sunxi, four Amlogic, and this. Reproducer configurations and the failing vmlinux link log for each are published at
https://github.com/JGyimah/kreach under results/, if the ALSA side is ever interested in running it over sound/ before a merge window.
Best,
Jude