On Fri 11 Oct 04:57 PDT 2019, Kalle Valo wrote:
Bjorn Andersson bjorn.andersson@linaro.org wrote:
The return value of dma_map_single() should be checked for errors using dma_mapping_error(), rather than testing for NULL. Correct this.
Fixes: 1807da49733e ("ath10k: wmi: add management tx by reference support over wmi") Cc: stable@vger.kernel.org Reported-by: Niklas Cassel niklas.cassel@linaro.org Signed-off-by: Bjorn Andersson bjorn.andersson@linaro.org
Did this fix any real bug? Or is this just something found during code review?
CONFIG_DMA_API_DEBUG screamed at us for calling dma_unmap_single() without ever having called dma_mapping_error() on the return value.
But Govind just pointed out to me that I hastily missed the fact that this code path leaks the dequeued skb. So I'll respin the patch to fix both issues at once.
Regards, Bjorn
-- https://patchwork.kernel.org/patch/11183923/
https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatch...
ath10k mailing list ath10k@lists.infradead.org http://lists.infradead.org/mailman/listinfo/ath10k