Panic: vm_fault: fault on nofault entry

[Anton Yuzhaninov]

Hi All,

I have panics when I try to use nginx on my notebook. When sendfile is
enabled it is easy to trigger panic, after disabling sendfile I've got
no panics so far and I think that panic is triggered by using sendfile.

Root filesystem - UFS on GELI whith 8k sectorsize.

GELI with sectorsize large than page size is non-standard configuration,
and I've got problem because of this in the past:
https://lists.freebsd.org/pipermail/freebsd-fs/2016-October/023957.html

but it was fixed in:
https://svnweb.freebsd.org/changeset/base/307626

Probably I've not used sendfile on this filesystem since then.

Panic string is:
"vm_fault: fault on nofault entry, addr: fffffe0220ab6000"

0xfffffe0220ab6000 is the address for bzero in
vnode_pager_generic_getpages_done

(kgdb) f 17
#17 0xffffffff807ee290 in vnode_pager_generic_getpages_done
(bp=0xfffffe01ea816490) at /spool/src/sys/vm/vnode_pager.c:1049
1049			bzero(bp->b_data + bp->b_bcount,
(kgdb) l
1044			if (!buf_mapped(bp)) {
1045				bp->b_data = bp->b_kvabase;
1046				pmap_qenter((vm_offset_t)bp->b_data, bp->b_pages,
1047				    bp->b_npages);
1048			}
1049			bzero(bp->b_data + bp->b_bcount,
1050			    PAGE_SIZE * bp->b_npages - bp->b_bcount);
1051		}
1052		if (buf_mapped(bp)) {
1053			pmap_qremove((vm_offset_t)bp->b_data, bp->b_npages);
(kgdb) p bp->b_data + bp->b_bcount
$4 = 0xfffffe0220ab6000 <Address 0xfffffe0220ab6000 out of bounds>

bp->b_data is a pointer to data from a file served using sendfile.
and bp->b_bcount is 8192 (GELI sector size is also 8192)

Full backtrace: http://termbin.com/dqpl
OS version: 11.1-STABLE #0 r322995

-- 
Best Regards,
Anton Yuzhaninov