mirror of
https://github.com/u-boot/u-boot.git
synced 2025-05-08 10:39:08 +00:00
dm: usb: Deal with USB keyboard persisting across tests
Clear any USB-keyboard devices before running a unit test, to avoid using a stale udevice pointer in stdio. Add a long comment to explain this situation and why this solution seems best, at least for now. Signed-off-by: Simon Glass <sjg@chromium.org>
This commit is contained in:
parent
010c449263
commit
bc624321dc
5 changed files with 97 additions and 0 deletions
|
@ -12,6 +12,7 @@
|
|||
#include <net.h>
|
||||
#include <of_live.h>
|
||||
#include <os.h>
|
||||
#include <usb.h>
|
||||
#include <dm/ofnode.h>
|
||||
#include <dm/root.h>
|
||||
#include <dm/test.h>
|
||||
|
@ -289,6 +290,43 @@ static int test_pre_run(struct unit_test_state *uts, struct unit_test *test)
|
|||
{
|
||||
ut_assertok(event_init());
|
||||
|
||||
/*
|
||||
* Remove any USB keyboard, so that we can add and remove USB devices
|
||||
* in tests.
|
||||
*
|
||||
* For UT_TESTF_DM tests, the old driver model state is saved and
|
||||
* restored across each test. Within in each test there is therefore a
|
||||
* new driver model state, which means that any USB keyboard device in
|
||||
* stdio points to the old state.
|
||||
*
|
||||
* This is fine in most cases. But if a non-UT_TESTF_DM test starts up
|
||||
* USB (thus creating a stdio record pointing to the USB keyboard
|
||||
* device) then when the test finishes, the new driver model state is
|
||||
* freed, meaning that there is now a stale pointer in stdio.
|
||||
*
|
||||
* This means that any future UT_TESTF_DM test which uses stdin will
|
||||
* cause the console system to call tstc() on the stale device pointer,
|
||||
* causing a crash.
|
||||
*
|
||||
* We don't want to fix this by enabling UT_TESTF_DM for all tests as
|
||||
* this causes other problems. For example, bootflow_efi relies on
|
||||
* U-Boot going through a proper init - without that we don't have the
|
||||
* TCG measurement working and get an error
|
||||
* 'tcg2 measurement fails(0x8000000000000007)'. Once we tidy up how EFI
|
||||
* runs tests (e.g. get rid of all the restarting of U-Boot) we could
|
||||
* potentially make the bootstd tests set UT_TESTF_DM, but other tests
|
||||
* might do the same thing.
|
||||
*
|
||||
* We could add a test flag to declare that USB is being used, but that
|
||||
* seems unnecessary, at least for now. We could detect USB being used
|
||||
* in a test, but there is no obvious drawback to clearing out stale
|
||||
* pointers always.
|
||||
*
|
||||
* So just remove any USB keyboards from the console tables. This allows
|
||||
* UT_TESTF_DM and non-UT_TESTF_DM tests to coexist happily.
|
||||
*/
|
||||
usb_kbd_remove_for_test();
|
||||
|
||||
if (test->flags & UTF_DM)
|
||||
ut_assertok(dm_test_pre_run(uts));
|
||||
|
||||
|
|
Loading…
Add table
Add a link
Reference in a new issue