the ioctl issued by the ICP RAID management libraries (used by the
storcon and iirconfig tools). This requires some infrastructure changes:
* Add a "service callback" mechanism that the ld driver (cache service)
and the iopsp driver (raw service) can register with the icp parent.
Right now this callback allows the children to adjust their notion of
how many command openings are available.
* Add a mutex around the icp ioctl handler, allowing only one thread
to execute an ioctl at a time.
* Add a way to freeze the controller command queue. We stop all I/O
while processing rescans (due to the semantics of icp_cmd()).
* Make icp_cmd() work when !cold.
* Add detach support to ld@icp.
Instead, wait when it's really needed, when detaching. Project
draining from spruious wakeups.
* Don't flush the controller's cache when detaching; we can't do it
reliably, as the controller probably things the array drive no
longer exists. The cache is already flushed in ldclose() so this
should be no problem.
to explicitly version syspkgs. See discussion on source-changes@netbsd.org.
If we want to lose this functionality, we should discuss this on the appropriate
lists...
thread from the continuation chain: Not only must it have released all
spinlocks, but it must have signaled completion by finishing pthread__switch
or by having stored into pt_switchto.
Together with the previous pthread_switch.S changes, this fixes a couple of
crasehes caused by race conditions in the examination and use of pt_next,
and by switching to empty pt_switchtouc's.
* In switch-away cases, write PT_SWITCHTO last (after PT_SWITCHTOUC), so
that pthread__resolve_locks() doesn't see an empty SWITCHTOUC value. This
also permits pthread__resolve_locks() to use the presence of PT_SWITCHTO
as a sign that the thread has done all of its necessary chain work.
* Make the return-point of pthread__switch global and visible, so that its
address can be compared to the PC of a thread, again as a sign that its
chain-work is done.
(other architectures in progress, after they get the *previous* asm fix...)
file entries from the syspkg hierarchy, brought up to date to include some
that had been missed out.
Add the descrs file, which is a single file containing all the DESCR
file entries from the syspkg hierarchy, brought up to date to include some
that had been missed out.
Thanks to Jim Wise for all the work that went into the syspkg sets.
Add the attrs file, which is a single file to which can be added
attributes (like preserve).
Add a regpkg script, which allows a single system package to be
registered, based simply on its setname and syspkg name. All the
information is gleaned from the lists files by means of scripts, no
source tree is needed. The version number for the system package is
calculated automatically, based on the most recent mtime of the files
that constitute a system package.
Add a regpkgset script, to register all the system packages in a set.
The set names recognised are "all", "base", "comp", "etc", "games",
"man", "misc", "text".
and also bring up to date the ELF vs a.out knowledge.
Also, sort the directories into reverse alphabetical order, and add them to
the generated PLIST file after any file entries.
Grab the comments into a single file from the individual syspkg set entries.
Grab the descriptions into a single file from the individual syspkg set entries.
Add an attributes file to which can be added attributes for system packages.
The "preserve" attribute has been implemented.
Add extra comments and descriptions for system packages which were missing.
value field can now always be used as an input, and output, or both.
This is especially handy for data movers which can e.g. calculate the
CRC32C of a buffer as it is moved (which requires both a seed [input],
and a place to stash the result [output]).
1. Use REG_NEWLINE, rather than matching "\n".
2. Fix a bug where rm_user_from_groups would mangle group entries in
/etc/group when the user to be removed is the only member of the group,
by using substring matches.
itojun