Commit Graph

55 Commits

Author SHA1 Message Date
Marcel Moolenaar
f2c49dd248 Revamp of the syscall path, exception and context handling. The
prime objectives are:
o  Implement a syscall path based on the epc inststruction (see
   sys/ia64/ia64/syscall.s).
o  Revisit the places were we need to save and restore registers
   and define those contexts in terms of the register sets (see
   sys/ia64/include/_regset.h).

Secundairy objectives:
o  Remove the requirement to use contigmalloc for kernel stacks.
o  Better handling of the high FP registers for SMP systems.
o  Switch to the new cpu_switch() and cpu_throw() semantics.
o  Add a good unwinder to reconstruct contexts for the rare
   cases we need to (see sys/contrib/ia64/libuwx)

Many files are affected by this change. Functionally it boils
down to:
o  The EPC syscall doesn't preserve registers it does not need
   to preserve and places the arguments differently on the stack.
   This affects libc and truss.
o  The address of the kernel page directory (kptdir) had to
   be unstaticized for use by the nested TLB fault handler.
   The name has been changed to ia64_kptdir to avoid conflicts.
   The renaming affects libkvm.
o  The trapframe only contains the special registers and the
   scratch registers. For syscalls using the EPC syscall path
   no scratch registers are saved. This affects all places where
   the trapframe is accessed. Most notably the unaligned access
   handler, the signal delivery code and the debugger.
o  Context switching only partly saves the special registers
   and the preserved registers. This affects cpu_switch() and
   triggered the move to the new semantics, which additionally
   affects cpu_throw().
o  The high FP registers are either in the PCB or on some
   CPU. context switching for them is done lazily. This affects
   trap().
o  The mcontext has room for all registers, but not all of them
   have to be defined in all cases. This mostly affects signal
   delivery code now. The *context syscalls are as of yet still
   unimplemented.

Many details went into the removal of the requirement to use
contigmalloc for kernel stacks. The details are mostly CPU
specific and limited to exception_save() and exception_restore().
The few places where we create, destroy or switch stacks were
mostly simplified by not having to construct physical addresses
and additionally saving the virtual addresses for later use.

Besides more efficient context saving and restoring, which of
course yields a noticable speedup, this also fixes the dreaded
SMP bootup problem as a side-effect. The details of which are
still not fully understood.

This change includes all the necessary backward compatibility
code to have it handle older userland binaries that use the
break instruction for syscalls. Support for break-based syscalls
has been pessimized in favor of a clean implementation. Due to
the overall better performance of the kernel, this will still
be notived as an improvement if it's noticed at all.

Approved by: re@ (jhb)
2003-05-16 21:26:42 +00:00
Poul-Henning Kamp
18b492fc8c Bump the config version to force people to upgrade their config(8)
so the fix for emitting multiple instances of .o files will prevent
link errors on LINT.
2003-04-15 21:29:11 +00:00
Ruslan Ermilov
5d1b6a85bc Standardize handling of locore.[sS] etc. files.
Submitted by:	jake, bde, ru
2003-02-28 21:59:14 +00:00
Marcel Moolenaar
52f5014e73 Build the kernel with -mconstant-gp. This means that function calls,
with the exception of indirect function calls, are assumed to be
intra load module and thus that GP will be the same. This avoids
saving, setting and restoring GP for each function call and
reduces the kernel with ~320KB. There's obviously a performance
benefit as well.

Note that since we generally don't know if calls will be intra or
inter load module when we're compiling kernel modules, -mconstant-gp
cannot be used for modules.
2002-10-13 07:07:51 +00:00
Jake Burkholder
ba37958b34 Moved the rule for locore.o from kern.post.mk to Makefile.$ARCH. 2002-07-31 14:59:05 +00:00
Bruce Evans
8365569883 Moved the setting of all profiling-related variables except the key one
(PROFLEVEL) to kern.pre.mk so that it is easier to manage.  Bumped config
version to match.

Moved the check for cputype being configured to a less bogus place in
mkmakefile.c.
2002-07-13 19:36:14 +00:00
Marcel Moolenaar
1a4a595c4b Remove CWARNFLAGS and add GCC3. We handle GCC3.x specific flags
centrally now that we have GCC3 in the tree. The GCC3 variable
is a helper during the switch.
2002-05-19 03:41:48 +00:00
Peter Wemm
6e068bd9ae Now that we can thorougly self compile the kernel, stop hard coding the
cross compiler names.
2002-04-07 04:34:21 +00:00
David E. O'Brien
25a157bc6c Remove the setting of 'FMT'. We now do ELF by default, so only bother
to increase the lenght of the command line if needed.  The setting of FMT
also gets in the cross bootstrapping way for new platforms.
2002-03-24 17:45:46 +00:00
Peter Wemm
17694298c0 Use the FreeBSD cross tools instead of the linux ones. 2002-03-19 10:52:44 +00:00
Peter Wemm
7a2eb19dbf Commit some infrastructure for turning on -Werror for kernel compiles.
It doesn't actually do it yet though.  This adds a flag to config so
that we can exclude certain vendor files from this even when the rest
of the kernel has it on.  make -DNO_WERROR would also bypass all of it.
2002-02-20 23:35:56 +00:00
Peter Wemm
f6916f666c Proper fix for old config setting maxusers to 8. 2001-12-14 09:39:29 +00:00
Warner Losh
3dfff24a19 Move kern.post.mk to last line, now that it appears safe 2001-11-11 20:33:02 +00:00
Warner Losh
41c8eb3039 Factor the common parts of the Makefile.foo files. This introduces two
new files: kern.pre.mk, which contains most of the definitions, and
kern.post.mk, which contains most of the rules.

I've tested this on i386 and pc98.  I have had feedback on the sparc64
port, but no reports from anybody on alpha, ia64 or powerpc.  I
appologize in advance if I've broken you.

Reviewed by: jake, jhb, arch@
2001-11-02 21:34:20 +00:00
Brian Somers
143281f43c Only overwrite ${DESTDIR}${KODIR}.old/${KERNEL_KO} if we haven't
booted from it when doing an installkernel.

Only change kern.bootfile from ${DESTDIR}${KODIR}/${KERNEL_KO}
to ${DESTDIR}${KODIR}.old/${KERNEL_KO}, and only when we're renaming
a booted ${DESTDIR}${KODIR}/${KERNEL_KO} kernel.
2001-11-01 02:17:49 +00:00
Doug Rabson
74ada43be9 Use -ffreestanding instead of -fno-builtin. That's the officially blessed
way of saying that its not linking with libc.

Submitted by: peter
2001-10-30 09:37:00 +00:00
Doug Rabson
dab3d5bca0 Adjust for building with gcc 3.0.1. 2001-10-30 09:09:13 +00:00
Dag-Erling Smørgrav
a4c0058691 Support the "install.debug" and "reinstall.debug" targets for kernel modules.
Small tweaks to kldxref may be necessary to avoid the surprising (but harm-
less) behaviour of 'kldload foo' loading foo.ko.debug instead of foo.ko if
it is present in the kernel directory.

Approved by:	a week of silence on -arch
MFC after:	2 weeks
2001-10-27 00:52:50 +00:00
Ruslan Ermilov
0df659af99 Finish the asbestos suited move of $mach/conf/*.$mach to conf/*.$mach.
Fix some more typos.
2001-10-26 10:33:45 +00:00
Mark Peek
43aba0eadb Make override of nm consistent across all platforms and sort the overrides.
Reminded by:	obrien
2001-10-19 22:19:06 +00:00
Brian Somers
9082264160 Change kern.bootfile when mv'ing the kernel
Not tested by: brian
2001-10-17 18:04:13 +00:00
Doug Rabson
d413603991 Allow C++ comments in headers - the EFI headers are full of them. 2001-09-15 12:28:02 +00:00
Doug Rabson
0b02d706db * Enable dynamically linked kernel. This involves adding a self-relocator
to locore to process the @fptr relocations in the dynamic executable.
* Don't initialise the timer until *after* we install the timecounter to
  avoid a race between timecounter initialisation and hardclock.
* Tidy up bootinfo somewhat including adding sanity checks for when the
  kernel is loaded without a recognisable bootinfo.
2001-09-13 12:39:15 +00:00
Doug Rabson
0026671754 Rename kernel entry point from locorestart to __start. 2001-09-10 15:34:00 +00:00
Peter Wemm
76cb0cadf1 Enable hardwiring of things like tunables from embedded enironments
that do not start from loader(8).
2001-08-27 05:11:53 +00:00
Peter Wemm
dd84e9658b In order to bring about the end of the universe, delete the lines:
# DO NOT DELETE THIS LINE -- make depend uses it
2001-08-15 00:33:55 +00:00
Sheldon Hearn
2cc33d7946 When building a debugging kernel with modules, build modules with
debugging support as well.

This relies on support added in rev 1.105 to kmod.mk.

Requested by:	peter
2001-08-02 10:56:59 +00:00
Peter Wemm
18ec8d761c param.c no longer gets special treatment. Use the source tree version. 2001-07-26 23:04:46 +00:00
Warner Losh
ab11128f3e Bump config version. 2001-07-02 05:36:32 +00:00
Warner Losh
7ef33e2e39 Move kernel compile directory from sys/compile/FOO to
sys/compile/${MACHINE}/FOO.

Reviewed by: arch, obrien, peter and
	the USENIX terminal room secret kernel cabal
2001-06-30 06:29:47 +00:00
Peter Wemm
f41325db5f With this commit, I hereby pronounce gensetdefs past its use-by date.
Replace the a.out emulation of 'struct linker_set' with something
a little more flexible.  <sys/linker_set.h> now provides macros for
accessing elements and completely hides the implementation.

The linker_set.h macros have been on the back burner in various
forms since 1998 and has ideas and code from Mike Smith (SET_FOREACH()),
John Polstra (ELF clue) and myself (cleaned up API and the conversion
of the rest of the kernel to use it).

The macros declare a strongly typed set.  They return elements with the
type that you declare the set with, rather than a generic void *.

For ELF, we use the magic ld symbols (__start_<setname> and
__stop_<setname>).  Thanks to Richard Henderson <rth@redhat.com> for the
trick about how to force ld to provide them for kld's.

For a.out, we use the old linker_set struct.

NOTE: the item lists are no longer null terminated.  This is why
the code impact is high in certain areas.

The runtime linker has a new method to find the linker set
boundaries depending on which backend format is in use.

linker sets are still module/kld unfriendly and should never be used
for anything that may be modular one day.

Reviewed by:	eivind
2001-06-13 10:58:39 +00:00
Peter Wemm
6697e35149 Move the -I../../../include or -I/usr/include to the last entry on the
cc arguments.  Otherwise ipfilter's bogus #include lines will compile
reference /usr/include/netinet/ip_frag.h etc.
2001-06-12 06:06:18 +00:00
Jonathan Lemon
88d74af548 Relocate IPFilter from sys/netinet to sys/contrib/ipfilter. 2001-06-07 04:06:21 +00:00
Mike Smith
71cb0ea9bf New files and layout for the ACPI CA 20010518 update. 2001-05-29 20:05:35 +00:00
David E. O'Brien
d856af0b1a Add back in the bits to remove any existing "schg" flags as we have
kernel's out in wild that still have these flags set.
2001-04-29 01:53:57 +00:00
David E. O'Brien
ea87b3cb61 Don't install KO's with the "schg" flag.
We are way too inconsistent with our setting of the "schg" flag, and in
our default install, it doesn't really offer any additional security.

Reviewed by:	arch@
2001-04-25 20:56:15 +00:00
Warner Losh
40afc1048e Ignore chflags errors. This makes installing to nfs mounted target
directories work.
2001-04-25 06:19:58 +00:00
Warner Losh
581a68a232 Add support for MODULES_OVERRIDE. This is a list of modules to build
instead of all of them.  You can put this in /etc/make.conf or in
makeoptions.

Reviewed by: arch@

# docs to follow.
2001-04-02 08:52:05 +00:00
David E. O'Brien
483cbcc1eb Restore the careful preservation of .depend (rev 1.78 of Makefile.i386)
that I removed in my last commit dealing with `make depend' bogons.
This commit has some races, but hopefully they are too short to matter.
Unfortuneatly, neither .newdep nor .olddep is removed by `make clean'.

Submitted by:	bde
2001-03-24 08:44:28 +00:00
Kris Kennaway
5ca7924a91 Use CPUTYPE to add appropriate compiler flags to COPTFLAGS for kernel
builds.  This may be disabled using the NO_CPU_COPTFLAGS variable.

Reviewed by:	arch
2001-03-12 07:47:09 +00:00
Peter Wemm
b1fe97412b Add a 'clobber' target.. Like 'clean' but it takes out even the config
generated stuff, leaving only 'version' in the build dir.
2001-02-27 08:13:32 +00:00
David E. O'Brien
f1a87e83ee Fix make depend' => make kernel-depend'. Using the dependancy file
when rebuilding it is just Wrong.
2001-02-12 05:55:33 +00:00
Marcel Moolenaar
136345c019 Improve kernel bootstrapping:
o  Use objdump instead of gensetdefs(1) to build the linker sets.
o  Allow overriding of nm and objdump in resp. genassym.sh and
   gensetdefs.pl for non-native toolchains.

Reviewed by: arch
Perl improvements: Jos Backus <josb@cncdsl.com>, benno
2001-01-28 06:39:56 +00:00
John Baldwin
dfea516de2 Change the reinstall.debug target to depend on
{kernel,modules}-reinstall.debug rather than {kernel,modules}-reinstall.
Otherwise, the '.debug' portion of the target is lost, and you end up
reinstalling the non debug version instead of the debug version.
2001-01-19 19:43:45 +00:00
Peter Wemm
5ae7f4d746 Corresponding version bump for config with the DEV_FOO replacement for
the NFOO count option.
2001-01-19 12:50:16 +00:00
David E. O'Brien
1bc9ef1684 Back out the modules `make obj' speedup.
I cannot find a way to do this with the more esoteric source dir schemes
I thought of.
2000-12-24 13:29:58 +00:00
David E. O'Brien
c59a148d05 Use the canonical spelling for the path to sys. 2000-12-22 10:05:07 +00:00
David E. O'Brien
b3a120de36 Greatly speed up the modules `make obj' phase. 2000-12-21 20:21:24 +00:00
Marcel Moolenaar
a107c7d1c7 Sync-up so that buils actually work again. 2000-12-10 04:31:16 +00:00
Peter Wemm
3f964411b5 Make the xxxFILES= list generation generic. This makes it easier to add
things like MFILES= or CONFFILES= without having to modify config code.
2000-11-25 03:25:34 +00:00