1. Fix the bogon I introduced that made some root filesystems fail to be
made with `unable to make filesystem on sd1s1a' or some such error
message.
2. Latest installment in the never-ending chapter on making FTP do the right
thing on a cruel internet.
3. Make `express install' a truly express install and split off what it used
to do into a `novice install' that's even more geared to the novice
by asking questions in order, rather than forcing them through the menu
interface.
4. Make anonymous FTP setup truly DTRT.
5. Build lndir directly into sysinstall and make the novice install offer
to set up the ports tree. Also make the ports setup truly explanatory
about what it's doing, and always link the ports destination to /usr/ports
if necessary so that bsd.port.mk doesn't need to be tweaked.
This causes:
1: inetd to clear it's getlogin() name at startup (in case the sysadmin
logged in and su'ed to root and restarted inetd)
2: inetd to start each spawned process in it's own session.
3: inetd to call setlogin() on non-root processes (eg: uucp for uucico)
4: log failures more extensively
This means that root spawned processes from inetd remain responsible for
setting their login name if they change their uid. (eg: rshd, login, etc).
If they do not do so, it is safer for them to have no "login name" than a
wrong one (like "root") because the getlogin() system call is documented
as "secure" on 4.4BSD. inetd when started from /etc/rc would have no login
name anyway, so this isn't really a change - it's making it consistant with
the bootup state...
The setsid() change *may* cause something to break that is doing a setsid()
itself and checking the result - it will fail now because it's already been
done. The consensis seems to be that this is unlikely. David G. thinks
this is acceptable as it is cleaner from an architectural point of view.
vector. Now it is called the "symbol caching" vector. This was made
possible and unconfusing by other changes that allowed me to localize
everything having to do with the caching vector in the function
reloc_map().
Switched to alloca() for allocating the caching vector, and eliminated
the special mmap-based allocation routines. Although this was motivated
by performance reasons, it led to significant simplification of the
code, and made it possible to confine the symbol caching code to the
single function reloc_map().
Got rid of the unnecessary and inefficient division loop at the
beginning of rtld().
Reduced the number of calls to getenv("LD_LIBRARY_PATH") to just 1, on
suggestion from <davidg@root.com>.
Added breaks out of the relocation loops when the relocation address is
found to be 0. A relocation address of 0 is caused by an unused
relocation entry. Unused relocation entries are caused by linking a
shared object with the "-Bsymbolic" switch. The runtime linker itself
is linked that way, and the last 40% of its relocation entries are
unused. Thus, breaking out of the loop on the first such entry is a
performance win when ld.so relocates itself. As a side benefit, it
permits removing a test from md_relocate_simple() in
../i386/md-static-funcs.c.
Unused relocation entries in other shared objects (linked with
"-Bsymbolic") caused even bigger problems in previous versions of the
runtime linker. The runtime linker interpreted the unused entries as if
they were valid. That caused it to perform repeated relocations of the
first byte of the shared object. In order to do that, it had to remap
the text segment writable. Breaking out of the loop on the first unused
relocation entry solves that.
Submitted by: John Polstra <jdp@polstra.com>
civilised manner than panicing. This only happens as a result of another
state botch somewhere else, eg: from a tty driver calling putc or b_to_q
on a closed device. Apparently, it's also been implicated in a panic
with a status (^T) event on ptys.
This change should pretty well be in it's final form now.
net.inet.ip.intr-queue-maxlen (=== ipintrq.ifq_maxlen)
and net.inet.ip.intr-queue-drops (=== ipintrq.ifq_drops)
There should probably be a standard way of getting the same information
going the other way.
to the description in the manpage. g flag means "replace every occurence
in each word", and its absence means "replace first occurence in each word".
Previously, absence of the g flag was implemented to mean "replace first
occurence found in all words, and then stop replacing", which was incorrect.
the kernel. ppp_tty.c goes to some lengths to minimise the inter-layer
calling (including a soft ISR). ppp_tty.c takes care of the soft masking
that was needed still.
(I've discovered that bugs in this area show up within an hour if the
masking was not correct.. :-} This combination has proven stable on
specialix serial ports, although there was some concern about the softtty
parts of sio/cy and netisr colliding - but Bruce has fixed that now)
set in open() when CLOCAL is set unless carrier is present.
Fixed initialization of line discipline. It lived across opens.
Lines that started with the wrong discipline probably didn't work
at all, because TS_ISOPEN is only set by TTYDISC.
if_sl and if_ppp (from ppp-2.2), eliminating the nearly identical
pppcompress.[ch] code. Add maximum VJ compression states argument to
sl_compress_init().
if_sl: call sl_compress_init() with the extra argument.
non-fatal. I've make it return an appropriate error to the caller instead
of panic()ing.
Handling an error condition is inherently more friendly than exploding
the kernel.. :-) The new behavior is a little closer to traditional
clists, potentially making porting a little simpler.
Suggested by: bde (many months ago, I've been using this for a while..)
