allow a sysname list which is interated in lookup instead of a simple
namespace
====================
This delta was composed from multiple commits as part of the CVS->Git migration.
The checkin message with each commit was inconsistent.
The following are the additional commit messages.
====================
remove bit which leaked in
"locking on linux 2.4.x does not work. 2.4.x uses internally 64bit offsets
(start and end of lock) and end is set to 64 bit int max. When it converts
to struct flock, len is set to 0xffffffff (cut off high 32 bits) and afs_lockctl
decides it is byte-range locking. So on 2.4.x kernel there should be used flock64
instead of flock (flock64 is typedef to flock on 64bit systems such as sparc64
and ia64)"
====================
This delta was composed from multiple commits as part of the CVS->Git migration.
The checkin message with each commit was inconsistent.
The following are the additional commit messages.
====================
use flock64 for linux24
based on suggestion from rees@umich.edu
====================
This delta was composed from multiple commits as part of the CVS->Git migration.
The checkin message with each commit was inconsistent.
The following are the additional commit messages.
====================
update for solaris cc joy
"The afs toplevel makefile theoretically allows you to build bits and
pieces of AFS without too much other stuff you don't need. Building the
kernel code this way is hard, since it imports stuff from so many modules,
and the dependencies aren't necessarily done right. This patch fixes many
of the special "for kernel use only" targets to be "for kernel use
only" and adds a few new ones, so that the only userlevel stuff built by
make TARGET=libafs is util, pinstall, rxgen, comerr (needed for
afs/Makefile), and cmd"
"an mr-afs related patch introduced xdr_afs_int64 into the code. This is a
good thing I suppose. The same patch also re-introduced the use of
xdr_long on some platforms, which isn't needed, and isn't really
appropriate. Said patch also only included the #defines for xdr_afs_int64
into rpc_main.c and not xdr.h. (it is my opinion that
rpc_main.c:write_int32_macros should only be used if kflag is set, but
transarc didn't do it that way, so I'm not going to change it)"
"fix for rxgen so that split rpc's don't erroneously declare parameters in
the "start" function that should only be declared in the "end":
example:
for the following declaration:
Store(IN struct _index_t *descr,
IN string group<STAGE_STRINGSIZE>,
OUT string spool<STAGE_STRINGSIZE>, OUT afs_uint32 *slot) split;
before patch:
int StartDMV_Store(z_call, descr, group)
register struct rx_call *z_call;
struct _index_t * descr;
char * group, **spool;
{
[...]
after patch:
int StartDMV_Store(z_call, descr, group)
register struct rx_call *z_call;
struct _index_t * descr;
char * group;
{
[...]
"
"afsconf_GetKey() in auth/cellconfig.c doesn't check the return
value of afsconf_Check(), which can lead to the fileserver SEGV'ing if
afsconf_Check() returns an error."
"I noticed that the pdf documentation bundled with openafs doesn't include
the correct fonts to display properly (atleast with my acrobat). I've
generated afs-pdf:s with type1 fonts"
"This patch makes "df ." work in AFS directories under Solaris 8.
(df looks in /etc/mnttab, which is obtained from the kernel, for
the device number and then does a pwd-style walk up the tree..
Currently the device number gets set to 0x100000 by default which
doesn't match the 0x1234 device number of AFS vnodes.)"
fix for s390, per Carsten:
(this behaviour occurs only
on 31-bit
S/390-Systems, the "s390x"-machines should change the AFS-password also
without this patch as far as I can imagine)
put together from 1994(!) rx open source release available
at ftp://ftp.dementia.org/pub/rx, other architectures, rx_kcommon.c,
and some net research on hpux
has no chance of working, slim chance of compiling, and seriously needs
spinlock work to be useful
====================
This delta was composed from multiple commits as part of the CVS->Git migration.
The checkin message with each commit was inconsistent.
The following are the additional commit messages.
====================
shell of this file. won't work
====================
shell of osi_vfs.h
not useful yet
====================
use macros in osi_machdep.h for doing spinlocks
====================
include streams tli header
====================
do a little cleanup after looking at other architectures
still won't work until we know more about sleep/wakeup and kernel
semaphore interface crap
====================
try again. probably still not useful
====================
get this stuff compiling
This rewrite cleans up the code a bit, removes any athena specific
references (not needed anymore in this version), and adds support for
multi realm management of afs servers (you can now specify
"admin@OTHERREALM" in your userlist).
Code now checks as follows:
tname
tinst - remote user info from conn
tcell
lcell - local cell
lrealm - local realm (defaults to lcell if not avail)
if no remote cell or instance
allow localauth
if the cell of the remote connection matches local cell or local realm
if not tinst
allow if tname in UserList
if tinst
allow if tname.tinst in UserList
if cell doesn't match local cell or realm
if not tinst
allow if tname@cell in UserList
allow if tname@CELL in UserList
if tinst
allow if tname.tinst@cell in UserList
allow if tname.tinst@CELL in UserList
modified per openafs-devel discussion such that krb5 versions (/tinst
rather than .tinst) code path disabled for now
MR-AFs is making use of this to hook into logging to write multiline messages
into the log
====================
This delta was composed from multiple commits as part of the CVS->Git migration.
The checkin message with each commit was inconsistent.
The following are the additional commit messages.
====================
move forward to current version of function
