On 02/20/2012 02:44 PM, Arend van Spriel wrote:
> As indicated in [1] on netdev mailing list drivers should not block
> on the init_module() syscall. This patch defers the actual driver
> registration to a workqueue so the init_module() syscall can complete
> without delay.
>
> [1] http://article.gmane.org/gmane.linux.network/217729/
>
> Cc: Kay Sievers<[email protected]>
> Cc: Larry Finger<[email protected]>
> Signed-off-by: Arend van Spriel<[email protected]>
> ---
> Here is the patch I came up with to avoid udev blocking on init_module().
> It is pretty straightforward, but maybe I am overlooking things here. So
> feel free to comment.
>
> Gr. AvS
Arend,
Despite the simplicity, this patch looks right to me. As a side benefit, I
learned about the DECLARE_WORK macro. Handy for this purpose.
Larry
> ---
> .../net/wireless/brcm80211/brcmsmac/mac80211_if.c | 17 ++++++++++-------
> 1 files changed, 10 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/net/wireless/brcm80211/brcmsmac/mac80211_if.c b/drivers/net/wireless/brcm80211/brcmsmac/mac80211_if.c
> index fec0f10..fe73451 100644
> --- a/drivers/net/wireless/brcm80211/brcmsmac/mac80211_if.c
> +++ b/drivers/net/wireless/brcm80211/brcmsmac/mac80211_if.c
> @@ -1173,21 +1173,24 @@ static struct bcma_driver brcms_bcma_driver = {
> * and if so, performs a brcms_attach() on it.
> *
> */
> -static int __init brcms_module_init(void)
> +static void brcms_driver_init(struct work_struct *work)
> {
> int error = -ENODEV;
>
> + error = bcma_driver_register(&brcms_bcma_driver);
> + pr_err("%s: register returned %d\n", __func__, error);
> +}
> +
> +DECLARE_WORK(brcms_driver_work, brcms_driver_init);
> +
> +static int __init brcms_module_init(void)
> +{
> #ifdef DEBUG
> if (msglevel != 0xdeadbeef)
> brcm_msg_level = msglevel;
> #endif /* DEBUG */
>
> - error = bcma_driver_register(&brcms_bcma_driver);
> - pr_err("%s: register returned %d\n", __func__, error);
> - if (!error)
> - return 0;
> -
> - return error;
> + return schedule_work(&brcms_driver_work);
> }
>
> /**