Rework switch statement

Typo in commit title: s/FIONIO/FIONBIO/

Rebase on new commit

The point is any nested kernel exception saves the "wrong" value in the x18 stack in the trapframe. That value isn't used during exception exit so it doesn't matter currently (though Konrad I think does depend on changing x18 in trapframes downstream). The way this is visible in kgdb is that when you unwind across a nested exception, gdb believes that x18's value is saved in the trapframe (it just always believes the trapframe layout for any exception) so if you do p $x18 in the frames "above" a nested exception it will not reflect the actual value the CPU was using. So this is true for any nested exception (e.g. a device interrupt taking while handling a system call).

In D52210#1198404, @jrtc27 wrote:

I guess my view is that OLD_LIBS is really there so that soname bumps don't break existing binaries; you need the new library to be able to build the new binary, but don't want a window between installing the new library and being able to build the new binary where it doesn't work. That isn't really important for this case, because you can just not set WITHOUT_LIBFOO until you've updated all your binaries to not use that ABI. But at the same time, if we're using OLD_LIBS for all the libraries here, then yes, that's pointless if you're using OLD_FILES for rtld itself.

I had worked on this patch a while ago due to the weird kgdb issue, but @def has some downstream changes in CheriBSD where the wrong saved value in the trapframe matters outside of just debugging I believe.

In D52403#1198184, @kib wrote:

In D52403#1198039, @jhb wrote:

In D52403#1196680, @kib wrote:

We need some file lock to prevent parallel threads from corrupting nonblock/async state vs flags.

That is an old bug though? If it is just F_SETFL I would be happy to use a pool of sx locks or the like just for F_SETFL.

This is a separate bug and should get a separate fix.

Do we want to introduce sx pool? I think a pair of bit in f_vnread_flag similar to FOFFSET_LOCKED/WANT_LOCK and the existing mtx pool would be enough.

In D52403#1196680, @kib wrote:

We need some file lock to prevent parallel threads from corrupting nonblock/async state vs flags.

The bug of always reverting back to disabling FIONBIO dates back to the very first version of fcntl() 42 years ago: https://svnweb.freebsd.org/csrg/sys/kern/kern_descrip.c?r1=12748&r2=12747&pathrev=12748

I tested this in QEMU by using an NVMeoF connection using CRC32 digests over to another x86 VM running the target side of the NVMeoF connection.

Thanks, I will update the commit log.

I suspect this was added while working on a merge from upstream early in the merge process but then made unnecessary later in the merge process. This raises an error on CheriBSD since the enum value is cast to a uintptr_t and then back to an enum inside of __DECONST and that looks like a cast to a capability from an integer which raises the error.

I need to think about this. I don't really want to have the PCI bus directly reading the config registers from the parent bridge. It's also true that you want to not release the bus until all VFs using that bus have been torn down, so you really kind of want a refcounted bus_alloc_resource here (in essence). It might be that just switching from RF_ACTIVE to RF_SHAREABLE when calling bus_alloc_resource is sufficient however.

I think it would be useful (if possible) to reformat the list of existing SRCS as a separate commit. It will be clearer what the differences from the regeneration are when it comes to reviewing those, and will also be easier to bisect if needed.

Replaces part of D52194 where there is some discussion. This was broken in the original addition of MK_LIB32==no removal in rG74c52c652633228dead1f7f5e7e46e21cb518a67.

Replaces part of D52194

For now I will split some parts of this out into separate reviews.

Simplify and add CLEANFILES