| Message ID | 20200106185817.640331-1-krisman@collabora.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | drivers base: transport component error propagation | expand |
On Mon, Jan 06, 2020 at 01:58:14PM -0500, Gabriel Krisman Bertazi wrote: > Hi, > > This small series improves error propagation on the transport component > to prevent an inconsistent state in the iscsi module. The bug that > motivated this patch results in a hanging iscsi connection that cannot > be used or removed by userspace, since the session is in an inconsistent > state. > > That said, I tested it using the TCP iscsi transport (and forcing errors > on the triggered function), which doesn't require a particularly complex > container structure, so it is not the best test for finding corner cases > on the atomic attribute_container_device trigger version. > > Please let me know what you think. Looks sane, feel free to take the first two patches through what ever tree iscsi patches go through. thanks, greg k-h
Gabriel, > This small series improves error propagation on the transport > component to prevent an inconsistent state in the iscsi module. The > bug that motivated this patch results in a hanging iscsi connection > that cannot be used or removed by userspace, since the session is in > an inconsistent state. > > That said, I tested it using the TCP iscsi transport (and forcing > errors on the triggered function), which doesn't require a > particularly complex container structure, so it is not the best test > for finding corner cases on the atomic attribute_container_device > trigger version. Applied to 5.6/scsi-queue, thanks!
Hi, This small series improves error propagation on the transport component to prevent an inconsistent state in the iscsi module. The bug that motivated this patch results in a hanging iscsi connection that cannot be used or removed by userspace, since the session is in an inconsistent state. That said, I tested it using the TCP iscsi transport (and forcing errors on the triggered function), which doesn't require a particularly complex container structure, so it is not the best test for finding corner cases on the atomic attribute_container_device trigger version. Please let me know what you think. Gabriel Krisman Bertazi (3): drivers: base: Support atomic version of attribute_container_device_trigger drivers: base: Propagate errors through the transport component iscsi: Fail session and connection on transport registration failure drivers/base/attribute_container.c | 103 ++++++++++++++++++++++++++++ drivers/base/transport_class.c | 11 ++- drivers/scsi/scsi_transport_iscsi.c | 18 ++++- include/linux/attribute_container.h | 7 ++ include/linux/transport_class.h | 6 +- 5 files changed, 137 insertions(+), 8 deletions(-)