casaroli opened a new pull request, #21912:
URL: https://github.com/apache/nuttx/pull/21912
## Summary
`virtqueue_add_buffer()` of OpenAMP checks the number of free descriptors
only with `VQUEUE_DEBUG` (`CONFIG_OPENAMP_DEBUG`). Without it, a request that
needs more descriptors than the queue has free walks past the end of the free
list: it overwrites descriptors that other requests still use, and the free
count wraps. QEMU then stops the device ("Desc next is 32768"), and with it the
block device. A virtio-blk read or write uses 3 descriptors and a flush 2, and
nothing limits the requests in flight: each open of `/dev/virtblkN` has its own
block-to-character driver (and its own lock), and file systems on the device
send their own requests. With a queue of 16 descriptors, 6 concurrent requests
fill it. This PR makes a request wait until the queue has enough free
descriptors; `virtio_blk_done()` wakes the waiting requests after the replies
give descriptors back. Only a cancellation ends the wait early. In an interrupt
handler or after a panic, where a request already polls for its re
ply, the request polls for finished requests instead of a wait. This is the
same fix as #21911 for the 9P transport.
## Impact
`drivers/virtio/virtio-blk.c` only. A request that finds the queue full now
waits for the replies of other requests instead of corrupting the queue.
Nothing changes while the queue has room. No API change. No Documentation
change. The same code in flat, protected and kernel builds.
## Testing
- Test program `blkring.c`: 8 threads each open `/dev/virtblk0` (so each has
its own block-to-character driver) and read random runs of 1 to 64 sectors for
10 s; the disk image has a known pattern and every byte is checked.
qemu-armv8a:nsh_smp (flat build, 2 CPUs, arm64, QEMU with HVF) with
`CONFIG_DRIVERS_VIRTIO_BLK`, `CONFIG_BCH` and
`CONFIG_DRIVERS_VIRTIO_MMIO_QUEUE_LEN=16`, a 4 MiB raw image on
`virtio-blk-device`. master: QEMU stops the device at once ("Desc next is
32768"), the threads never return, FAIL. This PR: PASS, 263098 reads, no bad
data, no error.
- ostest: qemu-armv8a:knsh (kernel build) exit 0, qemu-armv8a:nsh (flat
build) exit 0, with `CONFIG_DRIVERS_VIRTIO_BLK` and `CONFIG_BCH`, arm64, QEMU
with HVF.
- Not run: the interrupt and panic path (polling) with a full queue.
- rv-virt:nsh64 (flat build, riscv64) with the driver: `virtio-blk.c`
compiles without warnings; the link fails in libmetal (a flat risc-v build has
no `up_addrenv_va_to_pa` or `up_addrenv_pa_to_va`, on master too), not changed
here.
- checkpatch: `tools/checkpatch.sh -c -u -m -g` passes.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]