Tags

bsd-user: Minor fixes

These patches have been in my queue pending too long (I have a bunch of others
that haven't been reviewd, but those will be done clsoe to the end of the
release to not get in the way of the release).

The patches align the stack properly on x86_64, implements setting the tb-size
and properly setting the ts_tid for initial threads. They have all been
reviewed.

bsd-user: Misc changes for 9.1 (I hope)

V2: Add missing bsd-user/aarch64/target.h

This patch series includes two main sets of patches. To make it simple to
review, I've included the changes from my student which the later changes depend
on. I've included a change from Jessica and Doug as well. I've reviewed them,
but more eyes never hurt.

I've also included a number of 'touch up' patches needed either to get the
aarch64 building, or to implmement suggestions from prior review cycles. The
main one is what's charitably described as a kludge: force aarch64 to use 4k
pages. The qemu-project (and blitz branch) hasn't had the necessary changes to
bsd-user needed to support variable page size.

Sorry this is so late... Live has conspired to delay me.

bsd-user: Baby Steps towards eliminating qemu_host_page_size, et al

First baby-steps towards eliminating qemu_host_page_size: tackle the reserve_va
calculation (which is easier to copy from linux-user than to fix).

bsd-user mmap and exec branches from gsoc

This pull request represents the mmap and exec changes from Karim Taha
for his GSoC project.

They represent all the mmap and exec related system calls and get bsd-user to
the point that a dynamic hello-world works (at least for armv7).

There are a couple of patch check errors, but they are the lessor evil: I made
purposely bad style choices to ensure all the commits compiled (and i undid the
style choices in subsequent commits).

I pushed an earlier version to gitlab, and all but the riscv64 pipelines were
green.  Since bsd-user doesn't change anything related to ricsv64 (there's no
support in qemu-project repo, though we do have it in the bsd-user fork: coming
soon).

I think this is good to go.

https://gitlab.com/bsdimp/qemu.git

Warner

Pull request: fix ci by fixing clang-user

Pull request for bsd-user 2023 Q3 (first batch)

First batch of commits submitted by my GSoC student Karim Taha

These implement	the stat, statfs, statfh and dirents system calls.

In addition, fix a missing break statment, and submit Richard Henderson's
elf stat mmap cleansup.

2023 Q1 bsd-user upstreaming: bugfixes and sysctl

[ letter edited -- need reviews for these hunks
     bsd-user: Helper routines h2g_old_sysctl
     bsd-user: various helper routines for sysctl
]

This group of patches gets the basic framework for sysctl upstreamed. There's a
lot more to translate far too many binary blobs the kernel publishes via
sysctls, but I'm leaving those out in the name of simplicity.

There's also a bug fix from Doug Rabson that fixes a long int confusion leading
to a trunctation of addresses (oops)

There's a fix for the -static option, since clang hates -no-pie and needs only
-fno-pie.

Finally, I'm changing how I'm upstreaming a little. I'm doing a little deeper
dives into our rather chaotic repo to find a couple of authors I might have
missed. From here on out, I'll be using the original author's name as the git
author. I'll also tag the co-authors better as well when there's multiple people
that did something (other than reformat and/or move code around). I've
discovered more code moved about than I'd previously known. This seems more in
line with standard practice.

v3->pull:
o minor	tweaks in the conditional reviews around formatting
o fix all errors for check patch and am	OK with	remaining warnings for
  line length that's only slightly too long
o edited letter for changes in review process

v3:
o Removed -strict, it's not ready and needs a complete rethink.
o Add g_assert_not_reached()
o target -> guest in most places
o Use MIN() to simplify things
o Better types in many places (abi_int instead of int32_t)
o Use ARRAY_COUNT
o fix tabs copied from FreeBSD sources to spaces

v2:
o Created various helper functions to make the code a little better
o split a few patches that I thought would be approved together but
  that generated commentary. It's easier to manage 1 per patch for
  those.
o Add/delete G_GNU_UNUSED to ensure all patches compile w/o warnings
o Fix 64-bit running 32-bit binary to get a LONG or ULONG. Add a
  bounce buffer for these so we don't overflow anything on the target
  and return all the elements of arrays.
o Fixed a number of nits noticed in the review.
o Add or improve comments to explain things there were questions on
  during the review.
o fix noted typos
o fix host != target page size differences
o Add pointers to FreeBSD source code, as appropriate
o fix locking (mostly unlocking) on error paths
o Note: -strict feedback not yet applied due to large numbers of changes
  from the rest. Next round.

bsd-user: Prepare for future upstream of system calls.

This series of patches does three things.

First, it starts to give up on the idea that you can run FooBSD binaries on
BarBSD. They are too different to make that happen any time soon, though I've
kept the support for Net/OpenBSD, even though they haven't built. We'll need a
lot of work to make that happen, though, and I need to simplify to get things
upstream.

Second, it starts to move some of the ifdef trees into target.h.

Third, it starts to upstream bsd-file.h, but the remainder of the file in the
bsd-user fork had some issues that will be resolved before next quarter's
update.

Pull request

Fix FreeBSD 12 and 13 builds.

bsd-user: upstream signal implementation

Upstream the bsd-user fork signal implementation, for the most part.  This
series of commits represents nearly all of the infrastructure that surround
signals, except the actual system call glue (that was also reworked in the
fork and needs its own series). In addition, this adds the sigsegv and sigbus
code to arm. Even in the fork, we don't have good x86 signal implementation,
so there's little to upstream for that at the moment.

bsd-user's signal implementation is similar to linux-user's. The full context
can be found in the bsd-user's fork's 'blitz branch' at
https://github.com/qemu-bsd-user/qemu-bsd-user/tree/blitz which shows how these
are used to implement various system calls. Since this was built from
linux-user's stack stuff, evolved for BSD with the passage of a few years, it
no-doubt missed some bug fixes from linux-user (though nothing obvious stood out
in the quick comparison I made). After the first round of reviews, many of these
improvements have been incorporated.

Patchew history: https://patchew.org/QEMU/20220125012947.14974-1-imp@bsdimp.com/

bsd-user: arm (32-bit) support

This series of patches brings in 32-bit arm support for bsd-user.  It implements
all the bits needed to do image activation, signal handling, stack management
and threading. This allows us to get to the "Hello World" level. The arm and x86
code are now the same as in the bsd-user fork. For full context, the fork is at
https://github.com/qemu-bsd-user/qemu-bsd-user/tree/blitz (though the the recent
sig{bus,segv} needed updates are incomplete).

v5 changes:
   o Moved to using the CPUArchState typedef and move
     set_sigtramp_args, get_mcontext, set_mcontext, and
     get_ucontext_sigreturn prototypes to
     bsd-user/freebsd/target_os_ucontext.h
   o Fix issues with arm's set_mcontext related to masking
     and remove an unnecessary check.

We're down to only one hunk needing review:
    bsd-user/arm/target_arch_signal.c: arm set_mcontext

Warnings that should be ignored:
   o make checkpatch has a couple of complaints about the comments for the
     signal trampoline, since it's a false positive IMHO.
WARNING: Block comments use a leading /* on a separate line
+    /* 8 */ sys_sigreturn,
WARNING: Block comments use a leading /* on a separate line
+    /* 9 */ sys_exit

bsd-user pull request: merge dependencies for next architectures

Merge the dependencies for arm, aarch64, and riscv64 architectures. This joins
together two patch series:

[PATCH v2 00/15] bsd-user: misc cleanup for aarch64 import

Prepare for aarch64 support (the next architecture to be upstreamed). As the
aarch64 emulation is more complete, it relies on a number of different items.
In some cases, I've pulled in the full support from bsd-user fork. In other
cases I've created a simple stub (as is the case for signals, which have
independent changes pending, so I wanted to be as minimal as possible.  Since
all pre-12.2 support was purged from the bsd-user fork, go ahead and remove it
here. FreeBSD 11.x goes ouft of support at the end of the month. Remove what
little multi-version support that's in upstream.

and

[PATCH v3 0/9] bsd-user mmap fixes
This series synchronizes mmap.c with the bsd-user fork. This is a mix of old bug
fixes pulled in from linux-user, as well as some newer fixes to adress bugs
found in check-tcg and recent FreeBSD developments. There are also a couple of
style commits. Updated to migrate debugging to qemu_log.

as well as a couple of minor rebase tweaks. In addition, the next two
architectures I plan on upstreaming (arm and riscv64) also have their prereqs
satisfied with this request.

v2: Remove accidental module regression in patch 7 and try again.

This series of patches gets me to the point that I can run "Hello World" on i386
and x86_64. This is for static binaries only, that are relatively small, but
it's better than the 100% instant mmap failre that is the current state of all
things bsd-user in upstream qemu. Future patch sets will refine this, add
the missing system calls, fix bugs preventing more sophisticated programms
from running and add a bunch of new architecture support.

There's three large themes in these patches, though the changes that
represent them are interrelated making it hard to separate out further.
1. Reorganization to support multiple OS and architectures (though I've only
   tested FreeBSD, other BSDs might not even compile yet).
2. Diff reduction with the bsd-user fork for several files. These diffs include
   changes that borrowed from linux-user as well as changes to make things work
   on FreeBSD. The records keeping when this was done, however, was poor at
   best, so many of the specific borrowings are going unacknowledged here, apart
   from this general ack. These diffs also include some minor code shuffling.
   Some of the changes are done specifically to make it easier to rebase
   the bsd-user fork's changes when these land in the tree (a number of changes
   have been pushed there to make this more possible).
3. Filling in the missing pieces to make things work. There's many changes to
   elfload to make it load things in the right places, to find the interpreter
   better, etc. There's changes to mmap.c to make the mappings work better and
   there's changes to main.c that were inspired, at least, by now-ancient changes
   to linux-user's main.c.

I ran checkpatch.pl on this, and there's 350-odd errors it identifies (the vast
majoirty come from BSD's fetish for tabs), so there will need to be a V2 to fix
this at the very least. In addition, the change set is big (about +~4.5k/-~2.5k
lines), so I anticipate some iteration as well just based on its sheer
size. I've tried to keep each set small to make it easy to review in isolation,
but I've also allowed some interrelated ones to get a little bigger than I'd
normally like. I've not done the customary documentation of the expected
checkpatch.pl output because it is large, and because I wanted to get review
of the other parts rolling to get this project unstuck. Future versions of the
patch will document the expected output.

In addition, I noticed a number of places where I could modernize to make the
code match things like linux-user better. I've resisted the urge to do these at
this time, since it would complicate merging the other ~30k lines of diff that
remains after this batch. Future batches should generally be smaller once this
one has landed since they are, by and large, either a bunch of new files to
support armv7, aarch64, riscv64, mips, mipsel, mips64, ppc, ppc64 and ppc64le,
or are adding system calls, which can be done individually or small groups. I've
removed sparc and sparc64 support as they've been removed from FreeBSD and
have been near totally busted for years.

Stacey Son did the bulk of this work originally, but since I had to move things
around so much and/or retool that work in non-trivial ways, I've kept myself as
author, and added his signed-off-by line. I'm unsure of the qemu standard
practice for this, but am happy to learn if this is too far outside its current
mainstream. For a while Sean Bruno did the merges from upstream, and he's
credited using his signed-off-by in appropriate places, though for this patch
set there's only a few. I've tried to ensure that others who have work in
individual patches that I've aggregated together also are reflected in their
signed-off-by. Given the chaotic stat of the upstream repo for its early
history, this may be the best that can be reconstructed at this late date. Most
of these files are 'foundational' so have existed from the earliest days when
record keeping wasn't quite what I'd wish for in hindsight. There was only
really one change that I could easily cherry-pick (Colin's), so I did that.

This series of patches gets me to the point that I can run "Hello World" on i386
and x86_64. This is for static binaries only, that are relatively small, but
it's better than the 100% instant mmap failre that is the current state of all
things bsd-user in upstream qemu. Future patch sets will refine this, add
the missing system calls, fix bugs preventing more sophisticated programms
from running and add a bunch of new architecture support.

There's three large themes in these patches, though the changes that
represent them are interrelated making it hard to separate out further.
1. Reorganization to support multiple OS and architectures (though I've only
   tested FreeBSD, other BSDs might not even compile yet).
2. Diff reduction with the bsd-user fork for several files. These diffs include
   changes that borrowed from linux-user as well as changes to make things work
   on FreeBSD. The records keeping when this was done, however, was poor at
   best, so many of the specific borrowings are going unacknowledged here, apart
   from this general ack. These diffs also include some minor code shuffling.
   Some of the changes are done specifically to make it easier to rebase
   the bsd-user fork's changes when these land in the tree (a number of changes
   have been pushed there to make this more possible).
3. Filling in the missing pieces to make things work. There's many changes to
   elfload to make it load things in the right places, to find the interpreter
   better, etc. There's changes to mmap.c to make the mappings work better and
   there's changes to main.c that were inspired, at least, by now-ancient changes
   to linux-user's main.c.

I ran checkpatch.pl on this, and there's 350-odd errors it identifies (the vast
majoirty come from BSD's fetish for tabs), so there will need to be a V2 to fix
this at the very least. In addition, the change set is big (about +~4.5k/-~2.5k
lines), so I anticipate some iteration as well just based on its sheer
size. I've tried to keep each set small to make it easy to review in isolation,
but I've also allowed some interrelated ones to get a little bigger than I'd
normally like. I've not done the customary documentation of the expected
checkpatch.pl output because it is large, and because I wanted to get review
of the other parts rolling to get this project unstuck. Future versions of the
patch will document the expected output.

In addition, I noticed a number of places where I could modernize to make the
code match things like linux-user better. I've resisted the urge to do these at
this time, since it would complicate merging the other ~30k lines of diff that
remains after this batch. Future batches should generally be smaller once this
one has landed since they are, by and large, either a bunch of new files to
support armv7, aarch64, riscv64, mips, mipsel, mips64, ppc, ppc64 and ppc64le,
or are adding system calls, which can be done individually or small groups. I've
removed sparc and sparc64 support as they've been removed from FreeBSD and
have been near totally busted for years.

Stacey Son did the bulk of this work originally, but since I had to move things
around so much and/or retool that work in non-trivial ways, I've kept myself as
author, and added his signed-off-by line. I'm unsure of the qemu standard
practice for this, but am happy to learn if this is too far outside its current
mainstream. For a while Sean Bruno did the merges from upstream, and he's
credited using his signed-off-by in appropriate places, though for this patch
set there's only a few. I've tried to ensure that others who have work in
individual patches that I've aggregated together also are reflected in their
signed-off-by. Given the chaotic stat of the upstream repo for its early
history, this may be the best that can be reconstructed at this late date. Most
of these files are 'foundational' so have existed from the earliest days when
record keeping wasn't quite what I'd wish for in hindsight. There was only
really one change that I could easily cherry-pick (Colin's), so I did that.

v2: rejected patches dropped
    Use suggested glibc routines
    Updated to be closer to qemu style
    Disable bsd-user on netbsd and openbsd since they don't compile
    fold together a couple of related changes
    [[ tagged the review-by and acked-by from last series, but by hand...
      I think I got them all... ]]

v3: Fix a bug in refactoring load_elf_sections and is_target_elf_binary
    Fix spelling errors in commit messages
    drop copy_cpu() patch until we use that function
    reword a few commit messages to make them clearer
    fix return value of setup_sigtramp to be 0 after #ifdef elimination
    Add patch to initialize random state and implement --seed
    Fix a boatload of style issues.
    Rebase to tip of master

NOTE: checkpatch.pl will have several warning about line length > 80 and
admonishment to not use architecture specific defines. The slightly long lines
look a lot better than wrapping and the arch specific defines are basically
required (one could wrap them, and I plan to in the future once I've pruned the
obsolete ones in a future patch set: there's so many that errors would crop up
if I were to do it now).

Warner

Colin Percival (1):
  bsd-user: Add '-0 argv0' option to bsd-user/main.c

Warner Losh (42):
  bsd-user: remove sparc and sparc64
  bsd-user: add copyright header to elfload.c
  bsd-user: Add Stacey's copyright to main.c
  bsd-user: add license to bsdload.c
  bsd-user: style nits: bsdload.c whitespace to qemu standard
  bsd-user: Remove all non-x86 code from elfload.c
  bsd-user: move arch specific defines out of elfload.c
  bsd-user: pass the bsd_param into loader_exec
  bsd-user: Fix calculation of size to allocate
  bsd-user: implement path searching
  bsd-user: Eliminate elf personality
  bsd-user: remove a.out support
  bsd-user: TARGET_NGROUPS unused in this file, remove
  bsd-user: elfload: simplify bswap a bit.
  bsd-user: assume pthreads and support of __thread
  bsd-user: add host-os.h
  bsd-user: Include host-os.h from main
  bsd-user: save the path to the qemu emulator
  bsd-user: start to move target CPU functions to target_arch*
  bsd-user: Move per-cpu code into target_arch_cpu.h
  bsd-user: pull in target_arch_thread.h update target_arch_elf.h
  bsd-user: Include more things in qemu.h
  bsd-user: define max args in terms of pages
  bsd-user: Create target specific vmparam.h
  bsd-user: Add system independent stack, data and text limiting
  bsd-user: *BSD specific siginfo defintions
  bsd-user: Implement --seed and initialize random state
  bsd-user: Move stack initializtion into a per-os file.
  bsd-user: Add architecture specific signal tramp code
  bsd-user: elf cleanup
  bsd-user: Remove dead #ifdefs from elfload.c
  bsd-user: Rewrite target system call definintion glue
  bsd-user: Make cpu_model and cpu_type visible to all of main.c
  bsd-user: update debugging in mmap.c
  bsd-user: Add target_arch_reg to describe a target's register set
  bsd-user: Add target_os_user.h to capture the user/kernel structures
  bsd-user: add stubbed out core dump support
  bsd-user: elfload.c style catch up patch
  bsd-user: Refactor load_elf_sections and is_target_elf_binary
  bsd-user: move qemu_log to later in the file
  bsd-user: Implement interlock for atomic operations
  bsd-user: Update mapping to handle reserved and starting conditions

 bsd-user/bsd-mman.h                           |  121 --
 bsd-user/bsdload.c                            |  104 +-
 bsd-user/elfcore.c                            |   10 +
 bsd-user/elfload.c                            | 1469 +++++------------
 bsd-user/freebsd/host-os.h                    |   25 +
 bsd-user/freebsd/target_os_elf.h              |  137 ++
 bsd-user/freebsd/target_os_siginfo.h          |  145 ++
 bsd-user/freebsd/target_os_signal.h           |   78 +
 bsd-user/freebsd/target_os_stack.h            |  181 ++
 bsd-user/freebsd/target_os_thread.h           |   25 +
 bsd-user/freebsd/target_os_user.h             |  427 +++++
 bsd-user/freebsd/target_os_vmparam.h          |   38 +
 .../target_syscall.h => i386/target_arch.h}   |   27 +-
 bsd-user/i386/target_arch_cpu.c               |   76 +
 bsd-user/i386/target_arch_cpu.h               |  209 +++
 bsd-user/i386/target_arch_elf.h               |   35 +
 bsd-user/i386/target_arch_reg.h               |   82 +
 bsd-user/i386/target_arch_signal.h            |   94 ++
 bsd-user/i386/target_arch_sigtramp.h          |   29 +
 bsd-user/i386/target_arch_thread.h            |   47 +
 bsd-user/i386/target_arch_vmparam.h           |   46 +
 bsd-user/main.c                               |  836 ++--------
 bsd-user/mmap.c                               |  472 +++++-
 bsd-user/netbsd/host-os.h                     |   25 +
 bsd-user/netbsd/target_os_elf.h               |  146 ++
 bsd-user/netbsd/target_os_siginfo.h           |   82 +
 bsd-user/netbsd/target_os_signal.h            |   69 +
 bsd-user/netbsd/target_os_stack.h             |   56 +
 bsd-user/netbsd/target_os_thread.h            |   25 +
 bsd-user/openbsd/host-os.h                    |   25 +
 bsd-user/openbsd/target_os_elf.h              |  146 ++
 bsd-user/openbsd/target_os_siginfo.h          |   82 +
 bsd-user/openbsd/target_os_signal.h           |   69 +
 bsd-user/openbsd/target_os_stack.h            |   56 +
 bsd-user/openbsd/target_os_thread.h           |   25 +
 bsd-user/qemu.h                               |   63 +-
 bsd-user/sparc/target_arch_sysarch.h          |   52 -
 bsd-user/sparc64/target_arch_sysarch.h        |   52 -
 bsd-user/syscall.c                            |   11 -
 bsd-user/syscall_defs.h                       |  255 +--
 .../target_syscall.h => x86_64/target_arch.h} |   28 +-
 bsd-user/x86_64/target_arch_cpu.c             |   76 +
 bsd-user/x86_64/target_arch_cpu.h             |  247 +++
 bsd-user/x86_64/target_arch_elf.h             |   35 +
 bsd-user/x86_64/target_arch_reg.h             |   92 ++
 bsd-user/x86_64/target_arch_signal.h          |   94 ++
 bsd-user/x86_64/target_arch_sigtramp.h        |   29 +
 bsd-user/x86_64/target_arch_thread.h          |   40 +
 bsd-user/x86_64/target_arch_vmparam.h         |   46 +
 configure                                     |    7 +-
 meson.build                                   |    7 +-
 slirp                                         |    2 +-
 52 files changed, 4389 insertions(+), 2266 deletions(-)
 delete mode 100644 bsd-user/bsd-mman.h
 create mode 100644 bsd-user/elfcore.c
 create mode 100644 bsd-user/freebsd/host-os.h
 create mode 100644 bsd-user/freebsd/target_os_elf.h
 create mode 100644 bsd-user/freebsd/target_os_siginfo.h
 create mode 100644 bsd-user/freebsd/target_os_signal.h
 create mode 100644 bsd-user/freebsd/target_os_stack.h
 create mode 100644 bsd-user/freebsd/target_os_thread.h
 create mode 100644 bsd-user/freebsd/target_os_user.h
 create mode 100644 bsd-user/freebsd/target_os_vmparam.h
 rename bsd-user/{sparc/target_syscall.h => i386/target_arch.h} (60%)
 create mode 100644 bsd-user/i386/target_arch_cpu.c
 create mode 100644 bsd-user/i386/target_arch_cpu.h
 create mode 100644 bsd-user/i386/target_arch_elf.h
 create mode 100644 bsd-user/i386/target_arch_reg.h
 create mode 100644 bsd-user/i386/target_arch_signal.h
 create mode 100644 bsd-user/i386/target_arch_sigtramp.h
 create mode 100644 bsd-user/i386/target_arch_thread.h
 create mode 100644 bsd-user/i386/target_arch_vmparam.h
 create mode 100644 bsd-user/netbsd/host-os.h
 create mode 100644 bsd-user/netbsd/target_os_elf.h
 create mode 100644 bsd-user/netbsd/target_os_siginfo.h
 create mode 100644 bsd-user/netbsd/target_os_signal.h
 create mode 100644 bsd-user/netbsd/target_os_stack.h
 create mode 100644 bsd-user/netbsd/target_os_thread.h
 create mode 100644 bsd-user/openbsd/host-os.h
 create mode 100644 bsd-user/openbsd/target_os_elf.h
 create mode 100644 bsd-user/openbsd/target_os_siginfo.h
 create mode 100644 bsd-user/openbsd/target_os_signal.h
 create mode 100644 bsd-user/openbsd/target_os_stack.h
 create mode 100644 bsd-user/openbsd/target_os_thread.h
 delete mode 100644 bsd-user/sparc/target_arch_sysarch.h
 delete mode 100644 bsd-user/sparc64/target_arch_sysarch.h
 rename bsd-user/{sparc64/target_syscall.h => x86_64/target_arch.h} (59%)
 create mode 100644 bsd-user/x86_64/target_arch_cpu.c
 create mode 100644 bsd-user/x86_64/target_arch_cpu.h
 create mode 100644 bsd-user/x86_64/target_arch_elf.h
 create mode 100644 bsd-user/x86_64/target_arch_reg.h
 create mode 100644 bsd-user/x86_64/target_arch_signal.h
 create mode 100644 bsd-user/x86_64/target_arch_sigtramp.h
 create mode 100644 bsd-user/x86_64/target_arch_thread.h
 create mode 100644 bsd-user/x86_64/target_arch_vmparam.h

--
2.32.0

v6.1.0 release

v6.1.0-rc4 release

v6.1.0-rc3 release

v6.1.0-rc2 release

v6.1.0-rc1 release