On 10.01.22 09:56, Sascha Hauer wrote:
> On Sat, Jan 08, 2022 at 06:14:26PM +0100, Ahmad Fatoum wrote:
>> Memory corruption around device removal may go unnoticed, because
>> barebox is shutting down anyway and doing no new allocations.
>>
>> Add a new devunbind command that should help with debugging such issues
>> by allowing selective unbinding and removal of devices.
>>
>> Signed-off-by: Ahmad Fatoum <[email protected]>
>> ---
>>  commands/Kconfig      | 12 +++++++
>>  commands/Makefile     |  1 +
>>  commands/devunbind.c  | 74 +++++++++++++++++++++++++++++++++++++++++++
>>  drivers/base/driver.c |  7 ++--
>>  include/driver.h      |  4 +++
>>  5 files changed, 95 insertions(+), 3 deletions(-)
>>  create mode 100644 commands/devunbind.c
>>
>> diff --git a/commands/Kconfig b/commands/Kconfig
>> index e2c36949347e..9abd97271952 100644
>> --- a/commands/Kconfig
>> +++ b/commands/Kconfig
>> @@ -68,6 +68,18 @@ config CMD_DEVINFO
>>        If called with a device path being the argument, devinfo shows more
>>        default information about this device and its parameters.
>>  
>> +config CMD_DEVUNBIND
>> +    tristate
>> +    default y
>> +    prompt "devunbind"
>> +    help
>> +      Debugging aid to unbind device from driver at runtime
>> +
>> +      devunbind [-f] DEVICE
>> +
>> +      Options:
>> +        -f   unbind driver and force removal of device and children
>> +
>>  config CMD_DMESG
>>      tristate
>>      prompt "dmesg"
>> diff --git a/commands/Makefile b/commands/Makefile
>> index 0b7c1563b534..875826743ffe 100644
>> --- a/commands/Makefile
>> +++ b/commands/Makefile
>> @@ -107,6 +107,7 @@ obj-$(CONFIG_CMD_MIITOOL)        += miitool.o
>>  obj-$(CONFIG_CMD_DETECT)    += detect.o
>>  obj-$(CONFIG_CMD_BOOT)              += boot.o
>>  obj-$(CONFIG_CMD_DEVINFO)   += devinfo.o
>> +obj-$(CONFIG_CMD_DEVUNBIND) += devunbind.o
>>  obj-$(CONFIG_CMD_DRVINFO)   += drvinfo.o
>>  obj-$(CONFIG_CMD_READF)             += readf.o
>>  obj-$(CONFIG_CMD_MENUTREE)  += menutree.o
>> diff --git a/commands/devunbind.c b/commands/devunbind.c
>> new file mode 100644
>> index 000000000000..4bebb27e8e68
>> --- /dev/null
>> +++ b/commands/devunbind.c
>> @@ -0,0 +1,74 @@
>> +// SPDX-License-Identifier: GPL-2.0-only
>> +// SPDX-FileCopyrightText: © 2021 Ahmad Fatoum <[email protected]>, 
>> Pengutronix
>> +
>> +#include <command.h>
>> +#include <common.h>
>> +#include <complete.h>
>> +#include <driver.h>
>> +#include <getopt.h>
>> +
>> +static int do_devunbind(int argc, char *argv[])
>> +{
>> +    bool unregister = false;
>> +    struct device_d *dev;
>> +    int ret = COMMAND_SUCCESS, i, opt;
>> +
>> +    while ((opt = getopt(argc, argv, "fl")) > 0) {
>> +            switch (opt) {
>> +            case 'f':
>> +                    unregister = true;
>> +                    break;
>> +            case 'l':
>> +                    list_for_each_entry(dev, &active_device_list, active) {
>> +                            BUG_ON(!dev->driver);
>> +                            if (dev->bus->remove)
>> +                                    printf("%pS(%s, %s)\n", 
>> dev->bus->remove,
>> +                                           dev->driver->name, 
>> dev_name(dev));
>> +                    }
>> +                    return 0;
>> +            default:
>> +                    return COMMAND_ERROR_USAGE;
>> +            }
>> +    }
>> +
>> +    if (!argv[optind])
>> +            return COMMAND_ERROR_USAGE;
>> +
>> +    for (i = optind; i < argc; i++) {
>> +            dev = get_device_by_name(argv[i]);
>> +            if (!dev)
>> +                    return -ENODEV;
>> +
>> +            if (unregister)
>> +                    return unregister_device(dev);
> 
> Shouldn't you continue the loop here?

Yes. Fixed in v2.

> 
> Sascha
> 


-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

Reply via email to