Source-Changes-HG archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
[src/kqueue]: src/sys screened the four suspicious closef() usages listed in ...
details: https://anonhg.NetBSD.org/src/rev/f6c30686c437
branches: kqueue
changeset: 512521:f6c30686c437
user: jdolecek <jdolecek%NetBSD.org@localhost>
date: Wed Sep 25 14:45:41 2002 +0000
description:
screened the four suspicious closef() usages listed in this file to find
out whether knote_fdclose() is needed or not; it's not
move the item to 'done' stuff
diffstat:
sys/kqueue-todo | 24 ++++++++++++++++--------
1 files changed, 16 insertions(+), 8 deletions(-)
diffs (45 lines):
diff -r 60095acdd248 -r f6c30686c437 sys/kqueue-todo
--- a/sys/kqueue-todo Wed Sep 25 13:29:22 2002 +0000
+++ b/sys/kqueue-todo Wed Sep 25 14:45:41 2002 +0000
@@ -1,4 +1,4 @@
-# $NetBSD: kqueue-todo,v 1.1.2.25 2002/09/25 13:29:22 jdolecek Exp $
+# $NetBSD: kqueue-todo,v 1.1.2.26 2002/09/25 14:45:41 jdolecek Exp $
items todo from bill sommerfeld/thorpej:
- multiple entry points
@@ -37,6 +37,21 @@
- section 9 man page for internal api
- use pool for everything (that it can sanely be used for)
- lib/libc/sys/kqueue.2 manpage merged
+ - need to test bits of the tree that call closef() to determine
+ if knote_fdclose() is required. (aka fdrelease()/fdfree()):
+ * ktrderef() (f)ktrace - OK [ jdolecek - references
+ struct file, no descriptor available for userland ]
+ * unp_detach() OK [ jdolecek -
+ the sockets aren't attached to any process descriptor
+ and thus can't have kevent attached to them]
+ * nfsrv_zapsock() nfsd - OK [ jdolecek - the socket
+ will never be available as descriptor for nfsd]
+ * portal_unmount() mount_portal - OK [ jdolecek - holds
+ reference to struct file of the socket the process
+ calling portal_mount() used; this DTRT WRT eventual
+ kevents, the knote would be g/c when the process
+ terminates without regard to the extra reference
+ to the struct file ]
merge in from FreeBSD:
usr.bin/tail
@@ -56,13 +71,6 @@
areas to either fix or sanity check they're ok (and remove
the XXXLUKEM comment)
- - need to test bits of the tree that call closef() to determine
- if knote_fdclose() is required. (aka fdrelease()/fdfree()):
- * ktrderef() ktrace [jdolecek - OK]
- * unp_detach()
- * nfsrv_zapsock() nfsd
- * portal_unmount() mount_portal ?
-
- kqfilter for systrace (evaluate feasibility)
not implemented:
Home |
Main Index |
Thread Index |
Old Index