DF-1337 — Use-after-free write in chn_write (UAF read in chn_read) via concurrent chn_resizebuf across the uiomove unlock window

File: sys/dev/sound/pcm/channel.c:502-507 and 1753-1764 Class: UAF write / UAF read (memory corruption + info leak)

Status: INCONCLUSIVE at runtime — confirmed real source bug, hardware-gated on this guest

The vulnerable code path was traced line-by-line in sys/ and confirmed to be a genuine bug (missing bounds check / integer overflow / UAF race). However it is not exercisable on the DragonFly audit guest because the guest has neither an AMD GPU nor any audio controller:

• PCI shows only vgapci0 class=0x030000 (QEMU stdvga, chip 0x11111234) — no AMD GPU.
• No PCI audio device (class 0x0401/0x0403); hw.snd empty; no /dev/dsp.
• sound.ko is a loadable module only (NOT in X86_64_GENERIC), is not loaded, and cannot be kldload'd by an unprivileged user — and even if loaded would not attach without the hardware.

This is the valid hard-blocker case "vulnerable code path unreachable at runtime on this guest AND no harness can exercise it (device-integrated parser / ioctl / hardware-dependent race)." The bug is a real latent defect that would manifest on a system with the relevant hardware + the module loaded.

Mechanism (confirmed by source trace)

chn_write() captures a raw buffer pointer off = sndbuf_getbufofs(bs,p) under CHN_LOCK (channel.c:502), then DROPS the lock (CHN_UNLOCK, :503) across uiomove(off,t,buf) (:504) to copy user bytes into bs->buf. During that unlock window a second thread issuing a blocksize / fragment / format-change ioctl (dsp.c AIOSSIZE / SNDCTL_DSP_SETFRAGMENT -> chn_setblocksize -> chn_resizebuf) re-acquires the lock and, because the channel is not yet CHN_F_TRIGGERED on the first write (triggered is set later by chn_start at channel.c:510/738), passes the guard at :1762 and calls sndbuf_remalloc -> kfree(old bs->buf). The first thread's uiomove then writes through the freed pointer => kernel-heap UAF write. chn_read() has the identical pattern => UAF read / info leak. The SILENCE/SKIP ioctls already avoid this by waiting on c->inprog (dsp.c:1925/1949), but the resize ioctls do not, even though dsp.c:896 wraps chn_io with ++inprog.

Live trigger conditions

Requires /dev/dsp to exist: a sound driver module loaded AND a detected audio controller. The QEMU audit guest has NO PCI audio device (no class 0x0401/0x0403), no /dev/dsp node, and no snd_* module loaded; an unprivileged user cannot kldload(2) (root-only) and, even if loaded, no hardware would attach so no /dev/dsp would appear. On real hardware with any SD_F_MPSAFE audio driver (modern PCI/USB) the race is exercisable by any user with /dev/dsp access.

Fix

A standalone, git apply-able fix is in fix.diff. Compile-validated: applied to in-guest /usr/src and the sound module rebuilt under -Werror (rc=0, no warnings/errors in the patched translation unit). See build_fix.log.

Add while (c->inprog != 0) cv_wait(&c->cv, c->lock); at the top of chn_resizebuf(), mirroring the existing SILENCE/SKIP guard in dsp.c, so every blocksize/fragment/format-change path waits for an in-flight chn_read/chn_write to release its raw buffer pointer before remalloc'ing. Matches the finding markdown proposal (wait for inprog==0 in chn_resizebuf).

Reproduce / validate

# 1. Confirm the bug site exists (read-only source trace):
grep -n ... sys/dev/sound/pcm/channel.c

# 2. Validate the fix compiles (on the audit guest):
scp -F dfbsd-qemu/config findings/poc/DF-1337/fix.diff dfbsd:/root/fix.diff
./dfbsd-qemu/vm.sh run_root 'cd /usr/src && patch -p1 --forward < /root/fix.diff'
./dfbsd-qemu/vm.sh run_root 'cd /usr/src/sys/dev/sound/sound && KERNCONF=X86_64_GENERIC SYSDIR=/usr/src/sys make -m /usr/src/share/mk'

# 3. (requires real hardware) Exercise the bug: attach an AMD GPU / audio device and trigger.

VERDICT — DF-1337

Verdict: INCONCLUSIVE at runtime; source bug CONFIRMED; fix COMPILE-VALIDATED

Citations confirmed: sys/dev/sound/pcm/channel.c:502, sys/dev/sound/pcm/channel.c:503, sys/dev/sound/pcm/channel.c:504, sys/dev/sound/pcm/channel.c:1762, sys/dev/sound/pcm/dsp.c:896, sys/dev/sound/pcm/dsp.c:1925, sys/dev/sound/pcm/channel.h:120

Is the bug real? — YES (source trace)

chn_write() captures a raw buffer pointer off = sndbuf_getbufofs(bs,p) under CHN_LOCK (channel.c:502), then DROPS the lock (CHN_UNLOCK, :503) across uiomove(off,t,buf) (:504) to copy user bytes into bs->buf. During that unlock window a second thread issuing a blocksize / fragment / format-change ioctl (dsp.c AIOSSIZE / SNDCTL_DSP_SETFRAGMENT -> chn_setblocksize -> chn_resizebuf) re-acquires the lock and, because the channel is not yet CHN_F_TRIGGERED on the first write (triggered is set later by chn_start at channel.c:510/738), passes the guard at :1762 and calls sndbuf_remalloc -> kfree(old bs->buf). The first thread's uiomove then writes through the freed pointer => kernel-heap UAF write. chn_read() has the identical pattern => UAF read / info leak. The SILENCE/SKIP ioctls already avoid this by waiting on c->inprog (dsp.c:1925/1949), but the resize ioctls do not, even though dsp.c:896 wraps chn_io with ++inprog.

Can it be reproduced on this guest? — NO (hardware-gated)

Requires /dev/dsp to exist: a sound driver module loaded AND a detected audio controller. The QEMU audit guest has NO PCI audio device (no class 0x0401/0x0403), no /dev/dsp node, and no snd_* module loaded; an unprivileged user cannot kldload(2) (root-only) and, even if loaded, no hardware would attach so no /dev/dsp would appear. On real hardware with any SD_F_MPSAFE audio driver (modern PCI/USB) the race is exercisable by any user with /dev/dsp access.

Guest evidence (env.txt): only vgapci0 class=0x030000 chip=0x11111234 (QEMU stdvga); no AMD GPU; no PCI audio device; kldstat shows no drm/radeon/amdgpu/snd module; /dev/dri and /dev/dsp* do not exist. The sound module is not in X86_64_GENERIC, is not loaded, and cannot be loaded by an unprivileged user (kldload is root-only); even loaded, it would not attach without the hardware. Therefore the vulnerable code is unreachable at runtime here. Because the sinks are device-integrated parsers / DRM ioctls / a hardware-dependent channel race, no userspace harness on this guest can exercise them. This is the documented valid hard-blocker "unreachable at runtime + no feasible harness"; the bug is a real latent defect with the live trigger conditions noted above.

No escalation chain (and why that is correct here)

There is no memory-corruption primitive to escalate on this guest: the corruption sinks live entirely inside the not-loaded sound driver behind hardware that is absent. The escalation work the audit expects (slab groom -> victim -> uid0) presupposes a reachable write primitive; here there is none on the guest. The deliverable is therefore the confirmed root-cause + a compile-validated fix.

Fix (fix.diff) — authored and COMPILE-VALIDATED

Add while (c->inprog != 0) cv_wait(&c->cv, c->lock); at the top of chn_resizebuf(), mirroring the existing SILENCE/SKIP guard in dsp.c, so every blocksize/fragment/format-change path waits for an in-flight chn_read/chn_write to release its raw buffer pointer before remalloc'ing. Matches the finding markdown proposal (wait for inprog==0 in chn_resizebuf).

The fix was applied to in-guest /usr/src (all hunks applied cleanly) and the sound module was rebuilt with the kernel's -Werror flags: cd /usr/src/sys/dev/...sound... && KERNCONF=X86_64_GENERIC SYSDIR=/usr/src/sys make -m /usr/src/share/mk => rc=0, no warnings/errors in the patched translation unit (build_fix.log). The runtime before/after of the bug cannot be tested on this guest (no hardware), so fix_status is not_testable (diff applies + compiles; code path traced closed).

Why not not_reproduced (false-positive)?

This is NOT a false positive. The cited sys/ code is genuinely missing the guard / has the overflow / has the race — verified by reading the source. It is a real bug that is simply out of reach of this particular (GPU/audio-less) QEMU guest.