re: kern/54495: stopping axe(4) may locks up if NET_MPSAFE is enabled

[matthew green <mrg%eterna.com.au@localhost>]

The following reply was made to PR kern/54495; it has been noted by GNATS.

From: matthew green <mrg%eterna.com.au@localhost>
To: gnats-bugs%netbsd.org@localhost, skrll%netbsd.org@localhost
Cc: kern-bug-people%netbsd.org@localhost, gnats-admin%netbsd.org@localhost,
    netbsd-bugs%netbsd.org@localhost
Subject: re: kern/54495: stopping axe(4) may locks up if NET_MPSAFE is enabled
Date: Thu, 29 Aug 2019 08:23:47 +1000

 > The command `ifconfig axe0 down' often locks up on NET_MPSAFE kernel.
 
 skrll added this in rev 1.76, and it comes with this comment:
 
         /*
          * softnet_lock can be taken when NET_MPAFE is not defined when ca=
 lling
          * if_addr_init -> if_init.  This doesn't mix well with the
          * usbd_delay_ms calls in the init routines as things like nd6_slo=
 wtimo
          * can fire during the wait and attempt to take softnet_lock and t=
 hen
          * block the softclk thread meaning the wait never ends.
          */
 #ifndef NET_MPSAFE
         /* XXX What to reset? */
 
         /* Wait a little while for the chip to get its brains in order. */
         DELAY(1000);
 #else
         axe_ax_init(un);
 #endif
 
 
 i think this is two problems.  (1) the above issue probably
 occurs with NET_MPSAFE as well, just less often, and is what
 you see here, and (2) we shouldn't reset so completely for
 stop, just just disable/turn off stuff.
 
 
 (fortunately, usbnet appears to be unrelated to this :-)
 
 
 .mrg.