Douglas E. Engert wrote:
This sounds similiar to the attached e-mail from April.
It looks like it has ben fixed in the source,
http://www.openafs.org/cgi-bin/wdelta/sol10u2-rename-20060703
i.e.
replaced VN_SETPATH with vn_setpath in src/afs/SOLARIS/osi_vnodeops.c
There are other private vnode_t fields which the solaris afs client
accesses directly (see sys/vnode.h for a list)
v_path was one of them. There are others that can come back to haunt us
in the future, along with several other things that can be done to bring
the afs client's vfs ops up to solaris 10 spec. See pp. 12, section
14.3.1 of:
http://developers.sun.com/solaris/articles/solaris_internals_ch14_file_system_framework.pdf
A quick grep for a few of the private vnode fields shows:
v_pages
osi_vm.c:93: if (AFSTOV(avc)->v_pages)
Fix: ask vn_has_cached_data() instead
v_filocks
osi_vnodeops.c:1116: if (vp->v_filocks) /* if locked, disallow
mapping */
Fix: ask vn_has_flocks() instead
/dale
_______________________________________________
OpenAFS-devel mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-devel
