Date: Wed, 28 Sep 2022 10:13:21 +0000 From: bugzilla-noreply@freebsd.org To: net@FreeBSD.org Subject: [Bug 252165] usb network and mii bus media status race condition Message-ID: <bug-252165-7501-LwDI1PenyE@https.bugs.freebsd.org/bugzilla/> In-Reply-To: <bug-252165-7501@https.bugs.freebsd.org/bugzilla/> References: <bug-252165-7501@https.bugs.freebsd.org/bugzilla/>
next in thread | previous in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D252165 --- Comment #24 from Ali Abdallah <ali.abdallah@suse.com> --- Created attachment 236901 --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=3D236901&action= =3Dedit dtrace ure usb traces I got another machine on which I didn't apply the workaround, running 13.1, with a USB Lan adapter that uses if_ure, I used the following dtrace probes. dtrace -n 'fbt:miibus:mii_pollstat:*,fbt:uether:ue_tick_task:*,fbt:if_ure:ure_tick:*,= fbt:if_ure:ure_ifmedia_sts:*,fbt:miibus:mii_tick:*,fbt:miibus:miibus_linkch= g:*, fbt:usb:* { printf("%p", curthread); }' You can see from the attached debug log (numbered lines) 1 48573 ure_ifmedia_sts:entry fffffe00dc75c720 1 48475 mii_pollstat:entry fffffe00dc75c720 But before exit entry of re_ifmedia_sts:return of thread fffffe00dc75c720, = you can see: 1052 0 48573 ure_ifmedia_sts:entry fffffe00e00271e0 1053 2 27791 usb_command_wrapper:entry fffffe0082a17ac0 1054 2 27738 usbd_callback_wrapper:entry fffffe0082a17ac0 1055 0 48475 mii_pollstat:entry fffffe00e00271e0 And that exactly when the link up/down happened. So thread fffffe00e00271e0 managed to go ahead and call mii_pollstat, acqui= ring the sc mutex, despite that thread fffffe00dc75c720 didn't release it yet, y= ou can see that the return from "polling the media state" on thread fffffe00dc75c720 comes way after the ure_ifmedia_sts:entry of thread fffffe00e00271e0 1517 3 48476 mii_pollstat:return fffffe00dc75c720 1518 3 48574 ure_ifmedia_sts:return fffffe00dc75c720 __snip if_ure__=20 ure_ifmedia_sts { ... URE_LOCK(sc);=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20 mii_pollstat(mii); ifmr->ifm_active =3D mii->mii_media_active; ifmr->ifm_status =3D mii->mii_media_status; URE_UNLOCK(sc); } __snip__ The mutex locked/released by URE_LOCK, is the same mutex used in if_ure, USB xfer and USB transfer, so I guess something is not playing well in the release/acquire sequence... Will try if time permits to dig more into the USB transfer/process code, bu= t I have to say that I'm extremely surprised that this bug didn't attract much attention, it actually renders any USB Ethernet device unusable on FreeBSD.= .. --=20 You are receiving this mail because: You are the assignee for the bug.=
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-252165-7501-LwDI1PenyE>