mmc: sdhci: plug DMA mapping leak on error
authorRussell King <rmk+kernel@arm.linux.org.uk>
Tue, 26 Jan 2016 13:40:42 +0000 (13:40 +0000)
committerUlf Hansson <ulf.hansson@linaro.org>
Mon, 29 Feb 2016 10:03:16 +0000 (11:03 +0100)
If we terminate a command early, we fail to properly clean up the DMA
mappings for the data part of the request.  Put this clean up to the
tasklet, which is the common path for finishing a request so we always
clean up after ourselves.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
[ Split original patch so that it now contains only the fix ]
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: stable@vger.kernel.org # v4.5+
Tested-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
drivers/mmc/host/sdhci.c

index 78359baee36935243e32874f0ab5e88897a1321a..97e5f40a831fff8d4ce662484273f471eaed02fa 100644 (file)
@@ -2207,6 +2207,22 @@ static void sdhci_tasklet_finish(unsigned long param)
 
        mrq = host->mrq;
 
+       /*
+        * Always unmap the data buffers if they were mapped by
+        * sdhci_prepare_data() whenever we finish with a request.
+        * This avoids leaking DMA mappings on error.
+        */
+       if (host->flags & SDHCI_REQ_USE_DMA) {
+               struct mmc_data *data = mrq->data;
+
+               if (data && data->host_cookie == COOKIE_MAPPED) {
+                       dma_unmap_sg(mmc_dev(host->mmc), data->sg, data->sg_len,
+                                    (data->flags & MMC_DATA_READ) ?
+                                    DMA_FROM_DEVICE : DMA_TO_DEVICE);
+                       data->host_cookie = COOKIE_UNMAPPED;
+               }
+       }
+
        /*
         * The controller needs a reset of internal state machines
         * upon error conditions.
This page took 0.026351 seconds and 5 git commands to generate.