-
Notifications
You must be signed in to change notification settings - Fork 5k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
SDHCI leaks DMA buffers leading to a full swiotlb buffer #5019
Comments
You should be able to detect the failure much earlier than this - This is on the list for next week. |
I tried playing with kmemleak today and it gave me nothing relating to mmc under the same conditions - other than some (what I assume to be spurious) warnings about setting up regmaps in v3d at boot time. Board has been left dd'ing from the dodgy SD card just in case. |
It entered the same failure mode over the weekend.
No results in kmemleak. The leak actually takes around 20k seconds of runtime to exhaust the SWIOTLB and what looks like ~2048 instances of the data interrupt error. |
It goes a lot faster - within a minute - if I apply this patch -
and set |
The problem appears to be associated with a data timeout. If I make the sdhci driver a lot more chatty I see this:
A sequence of reads are being performed so tuples of interrupts for cmd(read) complete/data complete are being received. On a dodgy read, the data timeout error bit is asserted which starts an SD controller reset cycle. Note the timestamps - usually an 8-10ms gap between the command interrupt and data interrupt but before the timeout there was just a command interrupt. I believe the timeouts are triggering the leak. Disabling card presence polling had no effect on the appearance of spurious interrupts or the leak. |
For anyone else reading this, The trace shows that for commands with a data phase you would expect to see the For a DMA buffer leak to occur, When [ To be continued ] |
See raspberrypi#5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now remove the call to mmc_blk_read_single(). Signed-off-by: Jonathan Bell <[email protected]>
See raspberrypi#5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See raspberrypi#5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See raspberrypi#5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See raspberrypi#5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See raspberrypi#5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See raspberrypi#5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See raspberrypi#5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
commit 676ce1112f4e1023ed5e49501b6279190bdcb152 from https://github.com/raspberrypi/linux.git rpi-6.6.y See raspberrypi/linux#5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]> Signed-off-by: Rajeshkumar Ramasamy <[email protected]>
commit 676ce1112f4e1023ed5e49501b6279190bdcb152 from https://github.com/raspberrypi/linux.git rpi-6.6.y See raspberrypi/linux#5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]> Signed-off-by: Rajeshkumar Ramasamy <[email protected]>
commit 676ce1112f4e1023ed5e49501b6279190bdcb152 from https://github.com/raspberrypi/linux.git rpi-6.6.y See raspberrypi/linux#5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]> Signed-off-by: Rajeshkumar Ramasamy <[email protected]>
commit 676ce1112f4e1023ed5e49501b6279190bdcb152 from https://github.com/raspberrypi/linux.git rpi-6.6.y See raspberrypi/linux#5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]> Signed-off-by: Rajeshkumar Ramasamy <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
See #5019 If an SD card has degraded performance such that IO operations time out then the MMC block layer will leak SG DMA mappings in the swiotlb during recovery. It retries the same SG and this causes the leak, as it is mapped twice - once in sdhci_pre_req() and again during single-block reads in sdhci_prepare_data(). Resetting the card (including power-cycling if a regulator for vmmc is present) ought to be enough to recover a stuck state, so for now don't try single-block reads in the recovery path. Signed-off-by: Jonathan Bell <[email protected]>
Describe the bug
When recovering data off poorly SD cards that have extremely long read access cycle times, eventually I get this splat:
Note the timestamps. I believe the prerequisite condition for the leak is hitting the spurious interrupt error message (which had been happening every ~10 seconds previously).
Given the length of time it takes for the leak to fill the bounce buffer entirely, I would guess it's related to the no-cd card presence polling mechanism as requesting the csd returns data.
Steps to reproduce the behaviour
Use a broken SD card such that you get the bad interrupt message, and read from the card continuously for about 150 hours.
Device (s)
Raspberry Pi 4 Mod. B
System
Logs
No response
Additional context
No response
The text was updated successfully, but these errors were encountered: