tech-net archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
m_copyback issue
Hi:
There's a case in m_copyback() where a new segment
(n) can be inserted where n->m_next = n. This can
arise when the m_split() is skipped and code local
to m_copyback() which emulates m_split() is called.
The attached diff removes this emulated code and
always uses m_split() when applicable. Any concerns
before I commit?
Thanks,
-seanb
Index: kern/uipc_mbuf.c
===================================================================
RCS file: /cvsroot/src/sys/kern/uipc_mbuf.c,v
retrieving revision 1.136
diff -u -r1.136 uipc_mbuf.c
--- kern/uipc_mbuf.c 11 May 2010 20:21:56 -0000 1.136
+++ kern/uipc_mbuf.c 21 Oct 2010 16:11:18 -0000
@@ -1346,7 +1346,7 @@
* if we're going to write into the middle of
* a mbuf, split it first.
*/
- if (off > 0 && len < mlen) {
+ if (off > 0) {
n = m_split0(m, off, how, 0);
if (n == NULL)
goto enobufs;
@@ -1390,19 +1390,6 @@
else
datap = NULL;
eatlen = n->m_len;
- KDASSERT(off == 0 || eatlen >= mlen);
- if (off > 0) {
- KDASSERT(len >= mlen);
- m->m_len = off;
- m->m_next = n;
- if (datap) {
- m_copydata(m, off, mlen, datap);
- datap += mlen;
- }
- eatlen -= mlen;
- mp = &m->m_next;
- m = m->m_next;
- }
while (m != NULL && M_READONLY(m) &&
n->m_type == m->m_type && eatlen > 0) {
mlen = min(eatlen, m->m_len);
Home |
Main Index |
Thread Index |
Old Index