CVE-2024-53058

In the Linux kernel, the following vulnerability has been resolved:

net: stmmac: TSO: Fix unbalanced DMA map/unmap for non-paged SKB data

In case the non-paged data of a SKB carries protocol header and protocol payload to be transmitted on a certain platform that the DMA AXI address width is configured to 40-bit/48-bit, or the size of the non-paged data is bigger than TSOMAXBUFF_SIZE on a certain platform that the DMA AXI address width is configured to 32-bit, then this SKB requires at least two DMA transmit descriptors to serve it.

For example, three descriptors are allocated to split one DMA buffer mapped from one piece of non-paged data: dmadesc[N + 0], dmadesc[N + 1], dmadesc[N + 2]. Then three elements of txq->txskbuffdma[] will be allocated to hold extra information to be reused in stmmactxclean(): txq->txskbuffdma[N + 0], txq->txskbuffdma[N + 1], txq->txskbuffdma[N + 2]. Now we focus on txq->txskbuffdma[entry].buf, which is the DMA buffer address returned by DMA mapping call. stmmactxclean() will try to unmap the DMA buffer ONLYIF_ txq->txskbuff_dma[entry].buf is a valid buffer address.

The expected behavior that saves DMA buffer address of this non-paged data to txq->txskbuffdma[entry].buf is: txq->txskbuffdma[N + 0].buf = NULL; txq->txskbuffdma[N + 1].buf = NULL; txq->txskbuffdma[N + 2].buf = dmamapsingle(); Unfortunately, the current code misbehaves like this: txq->txskbuffdma[N + 0].buf = dmamapsingle(); txq->txskbuffdma[N + 1].buf = NULL; txq->txskbuff_dma[N + 2].buf = NULL;

On the stmmactxclean() side, when dmadesc[N + 0] is closed by the DMA engine, txq->txskbuffdma[N + 0].buf is a valid buffer address obviously, then the DMA buffer will be unmapped immediately. There may be a rare case that the DMA engine does not finish the pending dmadesc[N + 1], dmadesc[N + 2] yet. Now things will go horribly wrong, DMA is going to access a unmapped/unreferenced memory region, corrupted data will be transmited or iommu fault will be triggered :(

In contrast, the for-loop that maps SKB fragments behaves perfectly as expected, and that is how the driver should do for both non-paged data and paged frags actually.

This patch corrects DMA map/unmap sequences by fixing the array index for txq->txskbuff_dma[entry].buf when assigning DMA buffer address.

Tested and verified on DWXGMAC CORE 3.20a