Am Dienstag, den 17.10.2017, 12:44 +0200 schrieb Arnd Bergmann:
> lkft found a boot time regression on the Hikey board that has no
> aliases entry for spi buses. of_alias_get_highest_id() here
> returns -ENODEV, which is then used as the initial number for the
> IDR allocation, and that in turn triggers a WARN_ON:
>
> WARNING: CPU: 3 PID: 53 at include/linux/idr.h:113
> spi_register_controller+0x890/0x940
>
> This reworks the algorithm to turn any negative number into zero
> again, restoring the original behavior for the Hikey case.
>
> Fixes: 9ce70d49fa80 ("spi: fix IDR collision on systems with both
> fixed and dynamic SPI bus numbers")
> Reported-by: Naresh Kamboju <[email protected]>
> Link: https://lkft.validation.linaro.org/scheduler/job/48009#L3147
> Signed-off-by: Arnd Bergmann <[email protected]>
> ---
> Not tested yet, please review carefully
I've already sent a v2 of the borked patch, which Mark apparently
applied to his fix/idr branch.
Regards,
Lucas
> ---
> drivers/spi/spi.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/spi/spi.c b/drivers/spi/spi.c
> index 0483410e7cb3..4b6985ec962c 100644
> --- a/drivers/spi/spi.c
> +++ b/drivers/spi/spi.c
> @@ -2116,9 +2116,11 @@ int spi_register_controller(struct
> spi_controller *ctlr)
> }
> if (ctlr->bus_num < 0) {
> mutex_lock(&board_lock);
> + id = of_alias_get_highest_id("spi");
> + if (id < 0)
> + id = -1;
> id = idr_alloc(&spi_master_idr, ctlr,
> - of_alias_get_highest_id("spi") + 1,
> - 0, GFP_KERNEL);
> + id + 1, 0, GFP_KERNEL);
> mutex_unlock(&board_lock);
> if (WARN(id < 0, "couldn't get idr"))
> return id;
