CVS log for src/sys/rump/Makefile.rump

Up to [cvs.NetBSD.org] / src / sys / rump

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

Handle versions > 100

version dup3

fix tripple/quadruple "r" letter typos in comments and log messages.

Add epoll(2) from Theodore Preduta as part of GSoC 2023

Make option VNODE_LOCKDEBUG more functional.

- Add "LOCKED=EXCL" option for vnode operations needing exclusive locked nodes.

- As it is impossible to check for shared locks held by current thread avoid
  false assertions by testing for exclusive lock only in the "LOCKED=NO" case.

- New option "RUMP_VNODE_LOCKDEBUG" enables VNODE_LOCKDEBUG for the rump kernel.

Sync with HEAD.

Allow to set sanitizer flags for the rump build from the
build.sh command line, like: -V RUMP_SANITIZE=address

use MACHINE_MIPS64

Sync with HEAD.

Pass -isystem and -imacro ccp flags to lint

Sync with HEAD.

Build rump for PPC_OEA64 on evbppc64.

Remove in-kernel handling of Router Advertisements

This is much better handled by a user-land tool.
Proposed on tech-net here:
https://mail-index.netbsd.org/tech-net/2020/04/22/msg007766.html

Note that the ioctl SIOCGIFINFO_IN6 no longer sets flags. That now
needs to be done using the pre-existing SIOCSIFINFO_FLAGS ioctl.

Compat is fully provided where it makes sense, but trying to turn on
RA handling will obviously throw an error as it no longer exists.

Note that if you use IPv6 temporary addresses, this now needs to be
turned on in dhcpcd.conf(5) rather than in sysctl.conf(5).

Set -fno-delete-null-pointer-checks for RUMP/MKSANITIZER with Clang

NULL + 0 arithmetic raises LLVM UBSan warnings, specially in sys/pslist.h
in the type-safe macros _PSLIST_VALIDATE_PTRS and _PSLIST_VALIDATE_CONTAINER.

Proposes on <tech-userlevel> without objections.

Merge changes from current as of 20200406

Revert previous

Requested by <riastradh>

Build RUMP with -fno-delete-null-pointer-checks on all compilers

no-delete-null-pointer-checks is required in Clang/LLVM at least for
NULL + 0 operations that trigger LLVM UBSan and could be miscompiled.

Sync with head.

The global offset table is spelled .TOC. on PPC64, so preserve that symbol.

Disable NULL pointer check deletion like we do for the kernel in
src/sys/conf/Makefile.kern.inc. This is the same code we are compiling after
all and it has the same issues.

implement 32-bit compat support for raidframe.

convert several raidframe ioctls to be bitsize idempotent so that
they work the same in 32 and 64 bit worlds, allowing netbsd32 to
configure and query raid properly.  remove useless 'row' in a few
places.  add COMPAT_80 and put the old ioctls there.

raidframeio.h:
  RAIDFRAME_TEST_ACC
  - remove, unused
  RAIDFRAME_GET_COMPONENT_LABEL
  - convert to label not pointer to label
  RAIDFRAME_CHECK_RECON_STATUS_EXT
  RAIDFRAME_CHECK_PARITYREWRITE_STATUS_EXT
  RAIDFRAME_CHECK_COPYBACK_STATUS_EXT
  - convert to progress info not pointer to info
  RAIDFRAME_GET_INFO
  - version entirely.
raidframevar.h:
  - rf_recon_req{} has row, flags and raidPtr removed (they're
    not a useful part of this interface.)
  - RF_Config_s{} and RF_DeviceConfig_s{} have numRow/rows removed.
  - RF_RaidDisk_s{} is re-ordered slightly to fix alignment
    padding - the actual data was already OK.
  - InstallSpareTable() loses row argument

rf_compat32.c has code for RF_Config_s{} in 32 bit mode, used
by RAIDFRAME_CONFIGURE and RAIDFRAME_GET_INFO32.

rf_compat80.c has code for rf_recon_req{}, RF_RaidDisk_s{} and
RF_DeviceConfig_s{} to handle RAIDFRAME_FAIL_DISK,
RAIDFRAME_GET_COMPONENT_LABEL, RAIDFRAME_CHECK_RECON_STATUS_EXT,
RAIDFRAME_CHECK_PARITYREWRITE_STATUS_EXT,
RAIDFRAME_CHECK_COPYBACK_STATUS_EXT, RAIDFRAME_GET_INFO.

move several of the per-ioctl code blocks into separate functions.

add rf_recon_req_internal{} to replace old usage of global
rf_recon_req{} that had unused void * in the structure, ruining
it's 32/64 bit ABI.

add missing case for RAIDFRAME_GET_INFO50.

adjust raid tests to use the new .conf format, and add a case to
test the old method as well.

raidctl:
deal with lack of 'row' members in a couple of places.
fail request no longer takes row.
handle "START array" sections with just "numCol numSpare", ie
no "numRow" specified.  for now, generate old-style configuration
but update raidctl.8 to specify the new style (keeping reference
to the old style.)

note that: RF_ComponentLabel_s::{row,num_rows} and
RF_SingleComponent_s::row are obsolete but not removed yet.

update from HEAD

Honor REPROFLAGS for rump components.

Sync with HEAD

Pass down ${CPUFLAGS}

Sync with HEAD (as of 26th Dec)

describe => rumpdescribe

Add a COMMENT describing what each component roughly does.

"make describe" prints the comment.

Requested/inspired by Vincent Schwarzer on rumpkernel-users

Sync with HEAD

Remove individual opt_foo.h files, continue consolidating to opt_rumpkernel.h

remove librump/rumpnet/opt, consolidate in rump/include/opt

Remove unnecessary DPSRCS+=

-imacros automatically results in a dependency.

Use BUILDRUMP_IMACROS if specified.

Move constant cpp macros from Makefile.rump into opt_rumpkernel.h
(makes cc invocation lines a bit shorter)

Sync with HEAD

Add a "default" alias for RUMP_NBCOMPAT, in case "default" and "all"
stop being the same thing in the future.

Add "all" and "none" as acceptable RUMP_NBCOMPAT value aliases to
undefined and empty, respectively.

There are tests which rely on compat code continuing to be present,
so default RUMP_NBCOMPAT to all releases.

reported by gson

Rename RUMP_COMPAT to RUMP_NBCOMBAT to better signify what the
variable does.

For RUMP_COMPAT, accept "all" as an alias.

Also accept a comma-separated list, since that's easier to pass through
shells than a whitespace-separated list.

Default compat to 60 70, now that it hopefully generates no fallout.

Put 50 back to compat list for a while (some fallout)

Define rump_nativeabi_p() under arch, where one would logically expect
it to be defined.

default compat to 60 & 70

Generate -DCOMPAT_nn from a list of nn provided in RUMP_COMPAT

Sync with HEAD

Remove shlib_version files and just use Makefile SHLIB_MAJOR/MINOR,
with the default provided by Makefile.rump (they're all 0.0 anyway)

need -Wno-unused-parameter with -Wextra

define the warnings used for hypercall components once

RUMPCOMP_USER has been deprecated for long enough, remove it.

Rebase to HEAD as of a few days ago.

Rebase.

Implement rumpcomp_sync_icache() hyprecall for mips and add
a barebone implementation if mips cache ops to librumpkern_sljit.

Some rump kernel components for interfacing with the environment need
to access both NetBSD kernel facilities and facilities provided by the
environment.

Where the environment's headers can be #included without namespace
conflict (which can be done - with care - for some Xen mini-os
headers), it is convenient to simply allow the rump kernel code to
call directly into the environment.

This also requires not mangling the symbol names.

So provide a variable RUMP_SYM_NORENAME which can be set by a component
namefile to a series of symbols or symbol prefixes (in makefile-quoted
awk regexp syntax).  For example, in the rumpuser-xen xenbus driver:
 RUMP_SYM_NORENAME=xenbus_|HYPERVISOR_|wake$$|block$$|schedule$$|force_evtchn

Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
(proposed on rumpkernel-users; very minor modification by myself)

sync with head.

for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.

this commit was splitted into small chunks to avoid
a limitation of cvs.  ("Protocol error: too many arguments")

sync with head

Remove old HAVE_SUN_LD compat handling, it was there long enough.

Add a RUMP_USE_CTOR compile-time switch.  If defined, use
__attribute__((constructor)) to determine which rump kernel components
and kernel modules are linked into the rump kernel.  If not defined
(default), use the regular approach with link sets.

This option is aimed to fix problems with toolchains where using link
sets is not possible because the linker does not generate the requisite
__start/__stop symbols for the link set sections (e.g. GNU gold, OS X, ...).

add multi-inclusion protection

Introduce RUMP_CURLWP compile option to select curlwp scheme.
Default is still "hypercall".

Allow RUMP_LDSCRIPT itself to be used to specify which ldscript to use.
HAVE_SUN_LD will be deprecated in the future.

Allow compile-time optimizations to curlwp.  This can have a pretty
staggering impact on performance.  When running sendto() in a loop,
the improvement is 200k more calls per second with an inlined __thread
curlwp as opposed to the default.  In other words, it shaves off hundreds
of CPU cycles per call (~20%).  Even just eliminating the x86_curlwp()
call in favor of an inline gives an improvement of 60k calls per second.

Allow multiple "rumpcomp_user" source modules to be specified by
introducing RUMPCOMP_USER_SRCS.  Make RUMPCOMP_USER issue a deprecation
warning, but for compat make it set RUMPCOMP_USER_SRCS=rumpcomp_user.c
for now.

Move the "is arch capable of loading native kernel modules into
rump kernel" clauses from bsd.own.mk to Makefile.rump.  Also,
add a rump_nativeabi_p() call to determine if rump kernel is
compiled with native ABI support.

Make ktrace a compile-time option

Support ktrace for rump kernels.

Requested by Justin Cormack on rumpkernel-users.

RUMPCOMP_USEROBJ -> RUMPOBJ_NORENAME

Widens the implied usage scope of the variable.

Rename RUMPTEST_BUILDSH to RUMPKERN_ONLY.  Turns out not building any
POSIX implementations of hypercalls is useful beyond the build.sh test,
such as when building rump kernels for non-POSIX hosts.

sync with head

The linker script is the argument of -T, so keep them together.

COMPAT_60 for rump kernels.

Requested via github (buildrump.sh)

Fix pasto: BUILDRUMP_CPPFLAGS was used twice.

Apparently warning flags are not kept in CWARNFLAGS.  Compensate.

resync from head

fix "build.sh rumptest" wrt private hypercalls

Build only rumpcomp_user.c as part of the rumpcomp_user rules
(otherwise ioconf.c wants to be part of .ALLSRC for whatever reason
I'm not really interested in, but which the make gurus will surely
explain to me)

Allow to completely skip the ldscript, useful for linkers that don't need one.

fix typo

order too

uniquefy, really should fix this so there are no dup rules.

Fix the build w/o DESTDIR

one more typofix

On second thought, let's use make syntax in makefiles ...

Use correct ldscript etc. for Sun ld.

Use "-isysroot ${DESTDIR}" for RUMPCOMP_USER when run as part of build.sh
so as to catch the necessary headers, which, unlike in the buildrump.sh
case, are not supposed to be intermingled with the ones provided by
the host.

Shuffle the namespaces in the previous around a bit, and make it
rumpcomp_user instead of rumpuser_component.  This makes sure the
rumpuser namespace is used only by librumpuser.

Allow rump kernel components to define their own hypercalls.  This is
useful for example for supporting the tap device on Linux, for which
Linux-specific ioctl's are required (or at least so I was told).

resync with head

sync with (a bit old) head

Pass linker script with -Wl,-T to avoid clang incompating itself out
without flagging an error.

Instead of maintaining an evergrowing list of -U's, leave that up
to an external script.

Don't duplicate arbitrary chunks of bsd.sys.mk in here (especially
-Wsign-compare, since that doesn't even work with the kernel, and was
just previously masked by -isystem).  Reorder a bit too.

Don't unconditionally use the error-silencing magic of -isystem
(as opposed to -I) on every component, only the zfs ones depend on it
to build.  Regular modules might benefit from a similar move.

Don't use ldscript unless HAVE_BINUTILS == 219

Resync to 2012-11-19 00:00:00 UTC

add some -U's for linux builds

sync with head

Deal with optional HAVE_GCC.

Make it easy to turn -DDIAGNOSTIC off (RUMP_DIAGNOSTIC=no) and -DDEBUG
on (RUMP_DEBUG=yes).

Sync with HEAD.

sync with head

Make rump compile things with -std=gnu99 like the kernel and modules.

sync with head

Sync with HEAD

Define COMPAT_50 to be 1 just like config(8) would be opt_compat_netbsd.h

Support LOCKDEBUG.  To use it, compile sys/rump with RUMP_LOCKDEBUG=yes.

requested by martin (sparc64 gdb cannot reliably produce a stack trace)

use -isystem where appropriate

Sync with HEAD.

sync with head.

sync with head

Limit visibility of _RUMPKERNEL to prevent abuse.

Sync with HEAD.

Give a cpp hint if the native kernel abi is supported.

apply a patch from dennis fergusson:

fix the powerpc64 quirk handling by introducing a prefix quirk and
using it as appropriate.  fix the (postfix) quirk.

now rump and powerpc64 get along nicely.

add + _SYMQUIRK='|\.(rump|RUMP)' for powerpc64, as recommended by pooka.
it gets rid of many of the warnings trying to link rump_cd9960, but
there are a bunch remaining.  atomic ops seem missing, and some ktrace
points.

Move make snippets for use with config(1) ioconf functionality from
rump into share/mk.  This is to make it useful for all kernel
builders.

Note: we have waaay too many weird and wonderful ways of making
kernel code (monolithic kernel, modules, rump).  There should be
only one way to build kernel code instead of a maze of twisty little
.mk files, all not quite alike.  When that is fixed, this snippet can go
into the more generic .mk file.

sync with head

Rename IOCONFIG to IOCONF.  It matches the config(1) keyword and
aligns better in Makefiles.

Invert inverted comment and verbosify it.

Globally define -Wno-pointer-sign, as it has become a pointless
exercise of "add it to every Makefile individually".

XXX: should autosynchronize with the rest of the kernel buildflags
in sys/conf/Makefile.kern.inc.

IOCONFIG comes from .CURDIR

add a rule for locators.h too

If IOCONFIG is defined, autogenerate ioconf.c and locators.h
from ${IOCONFIG}.

Make (no pun intended) it possible to perform component symbol
protection from a shell prompt with the command line:

    make -f /sys/rump/Makefile RUMP_SYMREN=module.kobj

This is benefitial especially on i386, where rump ABI matches the
kernel ABI (sans the renaming isolation).  In the above example
module.kobj can then be loaded in a rump kernel either statically
at link-time simply by giving module.kobj as an object to the
linker, or dynamically at runtime by calling rump_sys_modctl(MODCTL_LOAD).

Use linker script to make __start/stop_link_set_modules be present
in libs built with binutils >=2.19.  This is a less error-prone
method than the previous where components had to be tagged in the
Makefile as modules (and if they weren't, things broke.  and vice
versa).

Include arch/include/Makefile.inc for some compiler flags.

Hmm, in addition to leaving out -lc, -nostdlibs also leaves out
libgcc, and causes problems on some platforms.  Currently there is
no easy way to say "link against libgcc but not against libc", so
just comment out -nostdlibs until a better way is invented.

Don't link rump kernel libs against libc.  They have no host system
dependencies apart from indirect ones via rumpuser.

sync with head

binutils 2.19 has changed the old behaviour of defining __start_SECTNAME
for orphaned sections to using PROVIDE.  What this means is that
unless a rump component internally references that symbol, it will
not be included in the component shared library, and hence cannot
be referenced when the component is loaded.  Add a workaround which
works both with 2.16 and 2.19: force a reference to the __start
symbol internally and hence retain it in the resulting library.

kernel code needs -Wno-format-zero-length

Always define __NetBSD__ (for builds on non-NetBSD)

Sync with HEAD.

Commit is split, to avoid a "too many arguments" protocol error.

sync with head.

define DEBUGPRINT

Sync with HEAD.

prettyprint long line, no functional change

More reproducible build fixes:

1) invoke objcopy with --preserve-dates or it will update all the
   dates in the .a file to the present date.
2) add an invocation of ar-as-ranlib or the object file index ends up
   with the date in it. (it might also end up with an incorrect
   index.)

remove RANLIB too.  pointed out by perry.

Reflect the change to archivebuild in bsd.lib.mk rev 1.296 for
deterministic builds, as hinted by perry.

Sync with HEAD.

Drop to WARNS?=3 as src/sys doesn't compile with -Wsign-compare yet

* enable COMPAT_50 (we need a better compat model here, but this is
  enough for now)
* _RUMPKERNEL is not going anywhere anymore, so adjust comment

Sync with HEAD.

Change the use of formally undocumented features, which have now been
made to fail.  Specifically, change
.ifdef(SYMBOL) -> .ifdef SYMBOL or .if defined(SYMBOL),
and corresponding for .ifndef.

Also correct one error in lib/libm/Makefile (.ifdef (${MKCOMPLEX} != "no")?!?).

Sync with HEAD.

Do rump kernel symbol protection for vax.  All archs support it now.

* Add adapted version of the userspace atomic_cas ops for platforms
  lacking special instructions.  We always use the spinlocked version
  (could use RAS on UP NetBSD host, but it's not portable).
* Add platform-based symbol quirk tables for selectively not renaming
  toolchain symbols for some platforms.  Although, this should really
  depend on the (toolchain,platform)-tuple and not just the platform.

This allows arm, hppa, mips and sh3 to build succesfully with an
isolated rump kernel namespace.  ... now, one arch remains: vax.
you must compile vax.  then, only then will you MI be.

Treat all of ^__ as compiler namespace and do not rename those
symbols.  This fixes e.g. fortify and sh3 millicode issues.  This
is still not perfect, however, since e.g. the MIPS _gp_disp is not
in ^__.

XXX: we provide a few ^__ symbols in the kernel, but those can be
easily renamed.  The harder problem is how to keep the kernel clean
of those afterwards.

* sparc supports rump kernel symbol rename
* plonk vax in here since it has issues with rename

Disable symbol renaming for now on arm, mips, sh3 and sparc to
allow them to build.  Some libgcc symbols are being renamed also,
plus not all atomic operations are provided by libkern on these
archs.

Add Makefile.rump to DPSRCS.  Should solve issues with UPDATE=1 builds.

Rename symbols in rump kernel libraries to always belong to a "rump"
symbol namespace.  This has several implications:
  * no symbols are accidentally linked against host platform libraries
  * all host platform access is done via rumpuser, and hence the
    portability of rump is now well defined
  * functionality such as proplib can now exist within the rump virtual
    kernel and the application using it

Thanks to uwe for originally suggesting the bulk renaming approach
over a pint of what I believe was very crisp, curious and untypical
stout.  It had an almost acidic characteristic with fruity notes
and lacking dark and roasted flavours.

And thanks to lukem for pointing me in the right direction with make.

Define -DMODULAR in Makefile.rump instead of the sys/rump/librump
Makefile.inc so that it gets picked up by UPDATE=1

disable strict aliasing to mirror kernel build flags

Purge multiple kernel opt files.

Move list of rump-uses-kmod archs to bsd.own.mk for wider consumption.

Switch amd64 rumps to the kernel ABI.

Switch i386 away from rump/include/machine.  This means that rump
on i386 now uses the native kernel ABI.  This in turn means that
rump modules and kernel modules are binary equivalent and can be
used interchangeably.

We are conseptually a module, so -D_MODULE.  Group CPPFLAGS better.

Pick PPC_OEA over PPC_IBM4XX.  I've decided I like it more.
(plus it hopefully builds without source changes)

liberate trailing whitespace

-Isys/arch is used in kernel builds

Update haad-dm branch to haad-dm-base2.

Instead of punishing all archs for a fake vmparam.h, punish only
evbppc for its insolence and pick a ppc cpu family at random in
Makefile.rump.

Sync with HEAD.

Sync with HEAD.

Compile all rump kernel portions with -DMULTIPROCESSOR.

Sync with HEAD.

Sync with wrstuden-revivesa-base-2.

Sync with head.

Prevent amd64 compiler flag -mcmodel=kernel for rump kernel code.
This allows to compile rump libraries as shared libraries also on mad64.

Problem spotted by Greg Oster.  I guess his build broke or something ...

Install rump libraries and utilities to the base system and remove the
private non-installed build infrastructure from sys/rump.

breakdown of commit:
  * install relevant headers into /usr/include/rump
  * build sys/rump/librump/rumpuser and sys/rump/librump/rumpkern
    from src/lib and install as librumpuser and librump, respectively
    + this retains the ability to test a librump build with just the
      kernel sources at hand
  * move sys/rump/fs/lib/libukfs and sys/rump/fs/lib/libp2k to src/lib
    for general consumption, they are not kernel-space dwellers anyway
  * build and install sys/rump/fs/lib/lib$fs as librumpfs_$fs
  * add chapter 3 manual pages for rump, rumpuser, ukfs and p2k
  * build and install userspace kernel file system daemons if MKPUFFS=yes
    is spexified
  * retire fsconsole for now, it will make a comeback with an actually
    implemented version shortly

Sync with head.

enable DIAGNOSTIC

Sync w/ -current. 34 merge conflicts to follow.

sync with head

Sync with HEAD.

Turn off DIAGNOSTIC so it builds.

sync with head.

sync with head.

Rename MKPRIVATELIB to LIBISPRIVATE, to make it clearer that this is
a variable that is used by in-tree Makefiles to control behaviour.
(MKsomevar variables are generally intended to be controlled by the end-user)

sync with HEAD

Sync with HEAD.

sync with head.

Use namei() etc. from kernel sources instead of a reimplementation.
To accommodate, give the rootvnode its own vnode op vector with a
simple lookup operation.  This is used for looking up the file
system's device vnode instead of doing that directly in a homesmoked
namei().

Reduce diff to HEAD by adding files forgotten in early merges.

Sync with head.

sync with head.

Sync with HEAD.

file Makefile.rump was added on branch jmcneill-pm on 2007-11-02 12:43:13 +0000

file Makefile.rump was added on branch yamt-lazymbuf on 2007-09-03 14:44:26 +0000

Don't play rename & typecast games with kauth_cred_t, but rather
declare the type in rump.h only if necessary with the help of ifdef
magic.

Sync with HEAD.

file Makefile.rump was added on branch vmlocking on 2007-08-20 22:07:08 +0000

Wrap malloc() so that we catch the kernel arguments (namely M_ZERO)
properly.  It's fairly amusing that this wasn't noticed until now.

Sync with HEAD.

file Makefile.rump was added on branch nick-csl-alignment on 2007-08-15 13:50:07 +0000

Compile kernel stuff with DIAGNOSTIC.

-I${NETBSDSRCDIR}/common/include

Introduce RUMPs - Runnable Userspace Meta-Programs

/sys/rump contains programs which run unmodified kernel code in an
emulated userspace environment.  The kernel environment is provided
by librump.  Currently supported are a number of file systems,
which by using puffs integrate seamlessly into the system and
provide a similar user experience to if the code was running as
part of the kernel.  Potential future rumpification targets include
for example parts of the networking stack and some device drivers.

This work was supported by Google Summer of Code 2007.

file Makefile.rump was added on branch matt-mips64 on 2007-08-05 22:28:02 +0000

Introduce RUMPs - Runnable Userspace Meta-Programs

/sys/rump contains programs which run unmodified kernel code in an
emulated userspace environment.  The kernel environment is provided
by librump.  Currently supported are a number of file systems,
which by using puffs integrate seamlessly into the system and
provide a similar user experience to if the code was running as
part of the kernel.  Potential future rumpification targets include
for example parts of the networking stack and some device drivers.

This work was supported by Google Summer of Code 2007.