This patchset has been created for FreeBSD 11. It has not been tested nor maintened since and I currently don't have time to work on it. I think we can drop it (it can be a base for new contributors).

Remove obsolete syscalls. Add linux_send() and linux_recv()

Remove obsolete syscalls. Add linux_send() and linux_recv()

Remove obsolete syscalls. Add linux_send() and linux_recv()

Remove unnecessary cast. Remove obsolete syscalls. Add linux_send() and linux_recv()

Remove old Linux syscalls (not presents in Linux calls.S). Add dummy linux_syscall.

Remove old Linux syscalls (not presents in Linux calls.S)

Remove old Linux syscalls (not presents in Linux calls.S). Add dummy linux_syscall.

Remove old Linux syscalls (not presents in Linux calls.S)

In D7972#167400, @andrew wrote:

There are also a few syscalls that were removed when they changed to EABI. I don't see a need for us to support the syscalls userspace will never call.

Use <machine/pcb.h> instead of <arm/include/pcb.h>

Add more comments for ARM specific syscalls

Add sys/conf/files.arm

In D7970#167040, @gonzo wrote:

In D7970#166894, @soutade_gmail.com wrote:

In D7970#166887, @gonzo wrote:

Gregory,

I take it v2 reviews are the ones that are relevant and that should be reviewed/tested. And previous 16 reviews can be closed now? Is that right?

Thank you

Right. They're strictly the same (except an update needed to compile in new HEAD) but the first version is more Linux like patchset while the second is FreeBSD style.

Gregory,

Could you close all Linux-style revisions? Go to the bottom of the page and choose "Abandon" action with comment like "Overtaken by D7970 .. D7978". I was going to do it myself but it seems only revision author can abandon it.

Thank you

Overtaken by D7970 .. D7978

Overtaken by D7970 .. D7978

Overtaken by D7970 .. D7978

Overtaken by D7970 .. D7978

Overtaken by D7970 .. D7978

Overtaken by D7970 .. D7978

Overtaken by D7970 .. D7978

Overtaken by D7970 .. D7978

Overtaken by D7970 .. D7978

Overtaken by D7970 .. D7978

Overtaken by D7970 .. D7978

Overtaken by D7970 .. D7978

Overtaken by D7970 .. D7978

Overtaken by D7970 .. D7978

Overtaken by D7970 .. D7978

Overtaken by D7970 .. D7978

Overtaken by https://reviews.freebsd.org/D7970 .. https://reviews.freebsd.org/D7978

Use SV_ABI_ERRNO() macro

Typo fix

In D7970#166887, @gonzo wrote:

Gregory,

I take it v2 reviews are the ones that are relevant and that should be reviewed/tested. And previous 16 reviews can be closed now? Is that right?

Thank you

Update linux_set_tls() functions. md_tp field not present anymore for arch >= v6.

In D7971#164988, @soutade_gmail.com wrote:

These look to be only on platforms that support 32-bit linux emulation, so maybe there's a better name? maybe HAVE_LINUX32?

Sounds good to me. I'll update the patch tomorrow.

These look to be only on platforms that support 32-bit linux emulation, so maybe there's a better name? maybe HAVE_LINUX32?

In D7971#164955, @imp wrote:

I wonder, though, if there might be a more generic way to say

#if defined(i386) || defined(arm) || (defined(amd64) && defined(COMPAT_LINUX32))

since it is repeated so often.

In D7395#156870, @imp wrote:

Seems reasonable to me. I added code to do this in the normal kernel, but not in the tramp code because I didn't need it. Not sure your boot environment, but it certainly doesn't hurt anything and may help.

Update diff to current HEAD

Rework all functions with right use of ldrex and strex

Set signo and sighandler in trapframe before going into linux_sigcode() and linux_rt_sigcode()

Rework linux_locore.s. Now linux_sigcode() and linux_rt_sigcode() are simpler because all parameters all pre positioned in linux_sysvec.c while precedent implementation was just a translation.

In D6765#146757, @andrew wrote:

In D6765#146725, @soutade_gmail.com wrote:

This patch is a simple copy of i386 implementation. Modifications made to machdep.c are done in patches 14 (https://reviews.freebsd.org/D6771) and 15 (https://reviews.freebsd.org/D6772).

Those patches should be merged with this. It makes it difficult to review something like this that should be considered a complete unit of work if it's split over multiple changes.

In D6765#146644, @andrew wrote:

In D6765#146642, @andrew wrote:

Much of this seems to be copied from the i386 support, is it needed on arm? e.g. places it calls i386 specific functions.

I was talking about linux_machdep.c.

linux_sendsig() called linux_rt_sigcode() instead of linux_sigcode()

Rework linux_sigcode() and linux_rt_sigcode() after andrew comments

• Use # instead of $ (but $ seems to works too)
• Bad computation of return address (that should be the instruction following branch one)
• Frame was taken from r0, but it should be take from sp

In D6762#142437, @andrew wrote:

It looks like these should all be atomic operations. If so the flow should be:

1: ldrex old, [addr]
   operation
   strex res, new, [addr]
   cmp res, #0
   bne 1b

linux_vdso.c:33-35 : don't update style