appearing as a single argument passed to devmatch(8).
Don't depend on "sort" utility from usr/bin which might not be
available when devd is started.
Sponsored by: Mellanox Technologies
Instead of passing flags (which describe a type of nvlist)
every send/recv we remember them in channel.
It's enough for use to extract them only during unwrap.
This simplify use of Casper.
Reviewed by: bruffer@, bcr@ (both man page)
Differential Revision: https://reviews.freebsd.org/D14196 (man page)
With the option used to compile the kernel both sx and rw shared ops would
always go to the slow path which added avoidable overhead even when the
facility is disabled.
Furthermore the increased time spent doing uncontested shared lock acquire
would be bogusly added to total wait time, somewhat skewing the results.
Restore old behaviour of going there only when profiling is enabled.
This change is a no-op for kernels without LOCK_PROFILING (which is the
default).
Older versions of GCC don't allow flexible array members in a union.
Use a zero length array instead.
MFC after: 1 week
Reported by: jbeich@
Sponsored by: Mellanox Technologies
In testing, it's often useful to copy a few files into a directory and
kldxref them to ensure that particular cases are handled correctly.
Add --hints (-h) to facilitate this testing and enable future
automated testing.
Sponsored by: Netflix
In devd/devmatch.conf, we need to pass the event to the devmatch
serivce. It gets passed to devmatch -p for matching. We always pass
this, unlike hps' original patch, so we kill two birds with one stone
and only match modules to the event passed in.
Submitted by: hps@
Sponsored by: Netflix
'I' was omitting 'zero' values. This is not quite correct, and was put
in as a hack but not documented. Remove it. If we find what the hack
was really needed for, we'll either fix the need for it, or invent a
new flagged value type.
Submitted by: hps@
Sponsored by: Netflix
Implement 'T' field matching. This is needed to prevent false
positives. However, it's not general enough. It only handles one field
and there's a ton of edge cases even with that it likely wouldn't
handle. To do it more generally and also eliminate a lot of the
hackiness that's in this program now, we'd need to creating
directories for lookups ala awk, pearl, python, etc. It appears to be
sufficient, though, to get my keyboard loaded on boot.
Sponsored by: Netflix
to parse rather than searching for all events. Pass with new -p arg to
devmatch. devmatch will use that one event rather than walking the
entire tree.
kldload will stop at the first failure. So we need to loop. Also,
symbolic links may confused kldload into trying (and failing) to load
multiple modules at once, so guard against that.
Noticed by: hps (with similar patch)
Sponsored by: Netflix
artificial NOMATCH usb does in lieu of creating a device_t for devices
with no drivers. Also, correct bus to be 'uhub' since where USB
devices attach, even though 'usb' is more logical, we need the
physical bus here.
Submitted by: hps@
These are the style points that I'd like to try and maintain in our lua
scripts:
- Parentheses around conditionals
- Trailing semicolons, except on block terminators
- s:method(...) instead of string.method(s, ...) where applicable
There's likely more, but that'll get hammered out as we continue.
Prompt for GELI passphrase when geom_eli_passphrase_prompt has been set to
"YES" in loader.conf(5).
This entailed breaking out the password prompt into its own function that
can be reused between the password compare bits and this prompt that simply
takes the entered password and passes it along in the environment as
kern.geom.eli.passphrase.
I've also added a TODO to re-evaluate later if we want the "password
masking" -- it is currently not functional, so one still can't observe the
length of the password typed at the prompt.
This is the procedure that config.loadkernel tries to go through, but
reloading kernel config didn't use this function. Amend config.loadkernel to
take an optional other_kernel.
While here, be a little more verbose ("Trying to load kernel") so that it's
easy to follow where we've gone wrong.
1.) Instead of string.function(s, ...), use s:function(...)
2.) Don't try to concatenate `res`, it was just tested to be nil
3.) Note that "Loading configuration" is configured modules, and be a little
more precise in mentioning what failed ("loading of one or more modules")
An empty module_path to start with isn't ideal, but if all modules are
contained within a kernel directory (which is what we just tested) then it
isn't strictly an error. Don't assume that module_path has a value already.
When we fail to load the kernel, printing the result (which is guaranteed to
be nil) is not intended; print the name of the kernel.
autoboot_delay=NO is documented to wait for input and *not* autoboot, which
is the exact opposite of the current behavior.
Additionally, autoboot_delay=-1 is documented to disallow the user from
interrupting the boot (i.e. autoboot immediately), which was not previously
honored.
This also fixes the case insensitive comparison to be truly case
insensitive. This is kind of nit-picky, but the previous version would only
accept "no" and "NO".
Don't move this into config.reload because we may want to force reloads if
/boot changes out from under us later.
As a caution: changing kernels in lualoader at the moment might not be
loading all of your modules (in my testing, at least) from loader.conf(5).
This is a known problem.
Detect ancient GCC specifically, rather than using target architecture as a
crude heuristic.
Side note: compilers should really ignore -Wno- and -Wno-error= flags they
don't recognize. Seems like modern compilers produce warnings instead of
errors. Though, with -Werror they turn into errors. Clang's error can be
disabled with -Wno-error=unknown-warning-option, but GCC doesn't seem to
have a named method to disable the specific warning.
Submitted by: rpokala@ (earlier version)
Suggested by: rpokala@
Reviewed by: tinderbox
Sponsored by: Dell EMC Isilon
