In the Linux kernel, the following vulnerability has been resolved:
dmaengine: sf-pdma: Add multithread support for a DMA channel
When we get a DMA channel and try to use it in multiple threads it will cause oops and hanging the system.
% echo 64 > /sys/module/dmatest/parameters/threadsperchan % echo 10000 > /sys/module/dmatest/parameters/iterations % echo 1 > /sys/module/dmatest/parameters/run [ 89.480664] Unable to handle kernel NULL pointer dereference at virtual address 00000000000000a0 [ 89.488725] Oops [#1] [ 89.494708] CPU: 2 PID: 1008 Comm: dma0chan0-copy0 Not tainted 5.17.0-rc5 [ 89.509385] epc : vchanfinddesc+0x32/0x46 [ 89.513553] ra : sfpdmatx_status+0xca/0xd6
This happens because of data race. Each thread rewrite channels's descriptor as soon as deviceprepdma_memcpy() is called. It leads to the situation when the driver thinks that it uses right descriptor that actually is freed or substituted for other one.
With current fixes a descriptor changes its value only when it has been used. A new descriptor is acquired from vc->desc_issued queue that is already filled with descriptors that are ready to be sent. Threads have no direct access to DMA channel descriptor. Now it is just possible to queue a descriptor for further processing.