[PATCH] dm: Add check for NULL dev_part_str

[Heinrich Schuchardt]

On 5/15/21 6:36 PM, Sean Anderson wrote:
> Some callers (e.g. cmd/fs.c) of fs_set_blk_dev may use a NULL dev_part_str.
> While blk_get_device_part_str handles this fine,
> part_get_info_by_dev_and_name does not. This fixes commands crashing when
> implicitly using bootdevice.
>
> Fixes: 7194527b6a ("cmd: fs: Use part_get_info_by_dev_and_name_or_num to parse partitions")
> Reported-by: Heinrich Schuchardt<xypron.glpk at gmx.de>
> Signed-off-by: Sean Anderson<seanga2 at gmail.com>

With this patch the reported error cannot be reproduced:

=> host bind 0 ../sandbox.img
=> setenv bootdevice 0:1
=> ls host
       148   boot.scr
...

But the unit test fails after issuing:

setenv bootdevice 0:1

=> ut dm dm_test_part
Test: dm_test_part: part.c
test/dm/part.c:47, dm_test_part(): -19 ==
part_get_info_by_dev_and_name_or_num("mmc", "", &mmc_dev_desc,
&part_info, 1): Expected 0xffffffed (-19), got 0xffffffa3 (-93)
Test: dm_test_part: part.c (flat tree)
test/dm/part.c:47, dm_test_part(): -19 ==
part_get_info_by_dev_and_name_or_num("mmc", "", &mmc_dev_desc,
&part_info, 1): Expected 0xffffffed (-19), got 0xffffffa3 (-93)
Failures: 2
=>

We get 'No such device' instead of 'Protocol not supported'.

As the environment variable bootdevice influences the test result it
should be unset when setting up the test and set to its original value
when tearing down the test.

Best regards

Heinrich