"value":"In the Linux kernel, the following vulnerability has been resolved:\n\nALSA: pcm: Fix potential AB/BA lock with buffer_mutex and mmap_lock\n\nsyzbot caught a potential deadlock between the PCM\nruntime->buffer_mutex and the mm->mmap_lock. It was brought by the\nrecent fix to cover the racy read/write and other ioctls, and in that\ncommit, I overlooked a (hopefully only) corner case that may take the\nrevert lock, namely, the OSS mmap. The OSS mmap operation\nexceptionally allows to re-configure the parameters inside the OSS\nmmap syscall, where mm->mmap_mutex is already held. Meanwhile, the\ncopy_from/to_user calls at read/write operations also take the\nmm->mmap_lock internally, hence it may lead to a AB/BA deadlock.\n\nA similar problem was already seen in the past and we fixed it with a\nrefcount (in commit b248371628aa). The former fix covered only the\ncall paths with OSS read/write and OSS ioctls, while we need to cover\nthe concurrent access via both ALSA and OSS APIs now.\n\nThis patch addresses the problem above by replacing the buffer_mutex\nlock in the read/write operations with a refcount similar as we've\nused for OSS. The new field, runtime->buffer_accessing, keeps the\nnumber of concurrent read/write operations. Unlike the former\nbuffer_mutex protection, this protects only around the\ncopy_from/to_user() calls; the other codes are basically protected by\nthe PCM stream lock. The refcount can be a negative, meaning blocked\nby the ioctls. If a negative value is seen, the read/write aborts\nwith -EBUSY. In the ioctl side, OTOH, they check this refcount, too,\nand set to a negative value for blocking unless it's already being\naccessed."
