Date: Tue, 03 Jul 2001 19:46:23 +0200 From: Tor.Egge@fast.no To: rsowders@usgs.gov Cc: jjreynold@home.com, freebsd-stable@FreeBSD.ORG Subject: Re: Hard lockups since cvsup'ing Jul 1st. Help! Message-ID: <200107031746.TAA46553@midten.fast.no> In-Reply-To: Your message of "Tue, 3 Jul 2001 00:00:44 -0700" References: <OF9C18FD6E.59995745-ON88256A7E.00235244@wr.usgs.gov>
next in thread | previous in thread | raw e-mail | index | archive | help
> Don't know if this means anything, maybe not related at all, but from > Tor,Egge's email with patch I took a look at the cvsweb and noticed some > interesting things. Tor broke out in ddb and saw the system was hung in > lock status. I don't know why he suspected the mfs system as a cause but > he apparently patched his system to work. I used the panic command in ddb, got a memory dump, booted with old kernel, saved the memory dump, used gdb to look at the situation. I found the vnode. v_tag was VT_MFS. v_data pointed to a struct mfsnode that did not start with a struct lock. The nullfs MFC (bp 2001/06/25 21:20:13 PDT) changed the semantics of vop_noislocked() which is the default for VOP_ISLOCKED(). That change depends upon other file systems having their per-vnode private data start with a struct lock if they don't override VOP_LOCK/VOP_UNLOCK/VOP_ISLOCKED. A struct lock was added to the start of struct nfsnode in the same commit. By applying approximately the same change to mfs as to nfs, my hang was solved. The 4.3-STABLE machine didn't use devfs (struct devnode does not start with a struct lock). - Tor Egge To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200107031746.TAA46553>