]> Zhao Yanbai Git Server - minix.git/log
minix.git
12 years agoext2: resolve Coverity warnings
David van Moolenbroek [Mon, 13 Aug 2012 16:37:26 +0000 (16:37 +0000)]
ext2: resolve Coverity warnings

12 years agodp8390: resolve Coverity warnings (encore)
David van Moolenbroek [Mon, 13 Aug 2012 16:33:58 +0000 (16:33 +0000)]
dp8390: resolve Coverity warnings (encore)

12 years agoTTY: resolve Coverity warnings
David van Moolenbroek [Mon, 13 Aug 2012 16:32:59 +0000 (16:32 +0000)]
TTY: resolve Coverity warnings

12 years agorandom: resolve Coverity warnings
David van Moolenbroek [Mon, 13 Aug 2012 16:23:48 +0000 (16:23 +0000)]
random: resolve Coverity warnings

12 years agoMake memory driver arch-neutral
Arun Thomas [Mon, 13 Aug 2012 15:21:47 +0000 (17:21 +0200)]
Make memory driver arch-neutral

-Use PAGE_SIZE instead of I386_PAGE_SIZE
-No need to grant ioport access for ARM

12 years agoRename sys_vmctl_get_cr3_i386
Arun Thomas [Sat, 11 Aug 2012 15:45:35 +0000 (17:45 +0200)]
Rename sys_vmctl_get_cr3_i386

12 years agopm: update for ARM
Arun Thomas [Sat, 11 Aug 2012 15:45:06 +0000 (17:45 +0200)]
pm: update for ARM

12 years agocommon/libc: build for ARM
Arun Thomas [Sat, 11 Aug 2012 15:44:21 +0000 (17:44 +0200)]
common/libc: build for ARM

12 years agocsu: build for ARM
Arun Thomas [Sat, 11 Aug 2012 15:43:30 +0000 (17:43 +0200)]
csu: build for ARM

12 years agolibexec: add load_offset feature, used for ld.so
Ben Gras [Fri, 10 Aug 2012 16:27:23 +0000 (18:27 +0200)]
libexec: add load_offset feature, used for ld.so

. ld.so is linked at 0 but it can relocate itself; we
  wish to load ld.so higher though to trap NULL dereferences.
  if we know we have to execute ld.so, vfs tells libexec to put it
  higher.

12 years agosprofdiff for comparing sprofile results
Erik van der Kouwe [Sat, 11 Aug 2012 22:09:42 +0000 (22:09 +0000)]
sprofdiff for comparing sprofile results

This patch adds the sprofdiff tool, which compares two sets of profiling
output files. It sorts processes and symbols by difference in average
number of samples, placing those that took more time on the left first
and those that took more time on the right last. If multiple runs are
combined, a standard deviation is computed and this is used to compute
the significance level, which gives an indication of which differences
are likely to be due to chance.

This tool is run not on the raw profiling files, but on the output of
sprofalyze -d (a new option). Though having to use two tools and an
intermediate file seems a bit awkward, the advantage is that the
original source tree is not needed to resolve the symbols. For
comparisons, this is very useful. Also, the intermediate file is in a
text format that can easily be processed by scripts, which may be useful
for other purposes as well.

12 years agolibc/mallocdebug: Make arch-generic
Arun Thomas [Fri, 10 Aug 2012 20:41:28 +0000 (22:41 +0200)]
libc/mallocdebug: Make arch-generic

12 years agobrksize: Remove non-ELF code
Arun Thomas [Fri, 10 Aug 2012 20:36:02 +0000 (22:36 +0200)]
brksize: Remove non-ELF code

12 years agoFix minimal ISO installation
pikpik [Fri, 10 Aug 2012 02:10:29 +0000 (02:10 +0000)]
Fix minimal ISO installation
 - Minimal ISO's have little in /usr/src
 - Copy boot.cfg from /etc instead

12 years agoRemoved legacy floppy boot support
pikpik [Fri, 10 Aug 2012 01:54:04 +0000 (01:54 +0000)]
Removed legacy floppy boot support
 - fdbootparams and fdboot are removed
 - fdboot manual references are removed

12 years agodrivers.h: fix machine includes
Arun Thomas [Fri, 10 Aug 2012 20:12:13 +0000 (22:12 +0200)]
drivers.h: fix machine includes

12 years agoFix prototypes for getdents, read, readlink
Arun Thomas [Fri, 10 Aug 2012 20:09:43 +0000 (22:09 +0200)]
Fix prototypes for getdents, read, readlink

12 years agoClean up arch Makefiles for libminc
Arun Thomas [Fri, 10 Aug 2012 17:02:30 +0000 (19:02 +0200)]
Clean up arch Makefiles for libminc

12 years agoARM headers
Arun Thomas [Fri, 10 Aug 2012 16:05:05 +0000 (18:05 +0200)]
ARM headers

12 years agoWrite raw sprofile traces to preserve all information as well as save space
Erik van der Kouwe [Fri, 10 Aug 2012 18:49:40 +0000 (18:49 +0000)]
Write raw sprofile traces to preserve all information as well as save space

12 years agoRestore and use C version of oneC_sum
Arun Thomas [Fri, 10 Aug 2012 12:47:56 +0000 (14:47 +0200)]
Restore and use C version of oneC_sum

12 years agolibnetsock: resolve Coverity warnings
David van Moolenbroek [Tue, 7 Aug 2012 11:10:36 +0000 (13:10 +0200)]
libnetsock: resolve Coverity warnings

12 years agolibsys: resolve Coverity warnings
David van Moolenbroek [Tue, 7 Aug 2012 11:10:16 +0000 (13:10 +0200)]
libsys: resolve Coverity warnings

12 years agolibsys: remove unused stacktrace variant
David van Moolenbroek [Tue, 7 Aug 2012 11:10:05 +0000 (13:10 +0200)]
libsys: remove unused stacktrace variant

12 years agoPM: resolve Coverity warnings
David van Moolenbroek [Tue, 7 Aug 2012 11:09:20 +0000 (13:09 +0200)]
PM: resolve Coverity warnings

12 years agoIS: resolve Coverity warnings
David van Moolenbroek [Tue, 7 Aug 2012 11:09:09 +0000 (13:09 +0200)]
IS: resolve Coverity warnings

12 years agoIPC: resolve Coverity warnings
David van Moolenbroek [Tue, 7 Aug 2012 11:08:52 +0000 (13:08 +0200)]
IPC: resolve Coverity warnings

12 years agoprocfs: resolve Coverity warnings
David van Moolenbroek [Tue, 7 Aug 2012 11:08:26 +0000 (13:08 +0200)]
procfs: resolve Coverity warnings

12 years agovm: ignore RS pin (pre-allocate) requests for now
Ben Gras [Wed, 8 Aug 2012 13:47:45 +0000 (15:47 +0200)]
vm: ignore RS pin (pre-allocate) requests for now

. done by RS to reduce/remove dependency on VM for recovery
. RS has the default stack size of 64MB since the nosegments
  change, using a huge amount of unused memory to pre-allocate
. ignore these requests until actually required (i.e. being able
  to survive VM crashes)

Thanks to pikpik for investigating why RS was so huge.

12 years agolibc: ARM generateasm
Arun Thomas [Tue, 7 Aug 2012 15:01:25 +0000 (17:01 +0200)]
libc: ARM generateasm

12 years agoRe-write sprofalyze in C for better performance, more options and to fix problems...
Erik van der Kouwe [Sat, 4 Aug 2012 11:16:54 +0000 (11:16 +0000)]
Re-write sprofalyze in C for better performance, more options and to fix problems with the gap the kernel now has

12 years agopci: resolve Coverity warnings
David van Moolenbroek [Fri, 3 Aug 2012 17:00:17 +0000 (17:00 +0000)]
pci: resolve Coverity warnings

Or rather, slightly improve the code based on a false positive.

12 years agolance: resolve Coverity warnings
David van Moolenbroek [Fri, 3 Aug 2012 16:55:23 +0000 (16:55 +0000)]
lance: resolve Coverity warnings

12 years agofxp: resolve Coverity warnings
David van Moolenbroek [Fri, 3 Aug 2012 16:49:06 +0000 (16:49 +0000)]
fxp: resolve Coverity warnings

This removes some unused and unsupported features, and leaves in support
for another currently unused feature that might be used in the future.

12 years agofloppy: resolve Coverity warnings
David van Moolenbroek [Fri, 3 Aug 2012 16:14:37 +0000 (16:14 +0000)]
floppy: resolve Coverity warnings

12 years agoe1000: resolve Coverity warnings
David van Moolenbroek [Fri, 3 Aug 2012 16:10:24 +0000 (16:10 +0000)]
e1000: resolve Coverity warnings

12 years agodpeth: resolve Coverity warnings
David van Moolenbroek [Fri, 3 Aug 2012 15:48:54 +0000 (15:48 +0000)]
dpeth: resolve Coverity warnings

12 years agodp8390: resolve Coverity warnings
David van Moolenbroek [Thu, 2 Aug 2012 16:54:47 +0000 (16:54 +0000)]
dp8390: resolve Coverity warnings

12 years agodec21140A: resolve Coverity warnings
David van Moolenbroek [Thu, 2 Aug 2012 16:53:01 +0000 (16:53 +0000)]
dec21140A: resolve Coverity warnings

12 years agoRemove ACK EM_WSIZE/EM_PSIZE macro usage
Arun Thomas [Mon, 6 Aug 2012 11:00:21 +0000 (13:00 +0200)]
Remove ACK EM_WSIZE/EM_PSIZE macro usage

12 years agoReplace MACHINE/CHIP macros with compiler macros
Arun Thomas [Mon, 6 Aug 2012 10:55:50 +0000 (12:55 +0200)]
Replace MACHINE/CHIP macros with compiler macros

12 years agoNetBSD ARM headers
Arun Thomas [Tue, 3 Jul 2012 22:21:37 +0000 (00:21 +0200)]
NetBSD ARM headers

12 years agoMFS: fix block estimation computation
David van Moolenbroek [Thu, 2 Aug 2012 13:03:23 +0000 (13:03 +0000)]
MFS: fix block estimation computation

12 years agoLWIP - fix, removed segments
Tomas Hruby [Fri, 3 Aug 2012 07:20:34 +0000 (00:20 -0700)]
LWIP - fix, removed segments

- compiles again

12 years agoext2: remove -Werror from CFLAGS
David van Moolenbroek [Wed, 1 Aug 2012 12:39:14 +0000 (12:39 +0000)]
ext2: remove -Werror from CFLAGS

This breaks the Coverity build at the moment.

12 years agolibvtreefs: resolve Coverity warnings
David van Moolenbroek [Mon, 30 Jul 2012 14:00:22 +0000 (14:00 +0000)]
libvtreefs: resolve Coverity warnings

12 years agortl8139: resolve Coverity warnings
David van Moolenbroek [Mon, 30 Jul 2012 13:59:43 +0000 (13:59 +0000)]
rtl8139: resolve Coverity warnings

12 years agortl8169: resolve Coverity warnings
David van Moolenbroek [Mon, 30 Jul 2012 12:09:13 +0000 (12:09 +0000)]
rtl8169: resolve Coverity warnings

12 years agolibsffs: resolve Coverity warnings
David van Moolenbroek [Fri, 27 Jul 2012 16:30:58 +0000 (16:30 +0000)]
libsffs: resolve Coverity warnings

12 years agolibchardriver: resolve Coverity warnings
David van Moolenbroek [Fri, 27 Jul 2012 15:02:12 +0000 (15:02 +0000)]
libchardriver: resolve Coverity warnings

12 years agolibbdev: resolve Coverity warnings
David van Moolenbroek [Fri, 27 Jul 2012 14:55:10 +0000 (14:55 +0000)]
libbdev: resolve Coverity warnings

12 years agofilter: resolve Coverity warnings
David van Moolenbroek [Fri, 27 Jul 2012 14:52:23 +0000 (14:52 +0000)]
filter: resolve Coverity warnings

12 years agoatl2: resolve Coverity warnings
David van Moolenbroek [Fri, 27 Jul 2012 14:51:52 +0000 (14:51 +0000)]
atl2: resolve Coverity warnings

12 years agoat_wini: resolve Coverity warnings
David van Moolenbroek [Fri, 27 Jul 2012 14:49:55 +0000 (14:49 +0000)]
at_wini: resolve Coverity warnings

12 years agoahci: resolve Coverity warnings
David van Moolenbroek [Fri, 27 Jul 2012 14:46:05 +0000 (14:46 +0000)]
ahci: resolve Coverity warnings

12 years agobtrace(8): resolve Coverity warnings
David van Moolenbroek [Fri, 27 Jul 2012 14:44:12 +0000 (14:44 +0000)]
btrace(8): resolve Coverity warnings

12 years agoVFS: fix locking bug in clone_opcl
Thomas Veerman [Thu, 26 Jul 2012 12:28:03 +0000 (12:28 +0000)]
VFS: fix locking bug in clone_opcl

When VFS runs out of vnodes after closing a vnode in opcl, common_open
will try to unlock a vnode through unlock_filp that has already been
unlocked in clone_opcl. By first obtaining and locking a new vnode this
situation is prevented; if there are no free vnodes, common_open will
unlock a still locked vnode.

12 years agoVFS: check path components for NAME_MAX length
Thomas Veerman [Thu, 26 Jul 2012 15:22:16 +0000 (15:22 +0000)]
VFS: check path components for NAME_MAX length

12 years agoext2: use new secondary cache method
Thomas Veerman [Thu, 26 Jul 2012 15:20:16 +0000 (15:20 +0000)]
ext2: use new secondary cache method

This gets rid of the ! emitted by VM when using ext2

12 years agoEXT2: various fixes
Thomas Veerman [Thu, 26 Jul 2012 15:19:08 +0000 (15:19 +0000)]
EXT2: various fixes

.enable all compile time warnings and make them errors
.refactor functions with unused parameters
.fix null pointer dereference before checking for null
.proper variable initialization
.use safe string copy functions
.fix massive memory corruption bug in fs_getdents

12 years agoPM: a few Coverity inspired fixes
Thomas Veerman [Thu, 26 Jul 2012 15:18:21 +0000 (15:18 +0000)]
PM: a few Coverity inspired fixes

.initialize variable to prevent negative array indexing
.remove dead code

12 years agoprocfs: use safe string copy
Thomas Veerman [Thu, 26 Jul 2012 15:17:25 +0000 (15:17 +0000)]
procfs: use safe string copy

12 years agoisofs: fixes for coverity defects
Thomas Veerman [Thu, 26 Jul 2012 15:16:50 +0000 (15:16 +0000)]
isofs: fixes for coverity defects

.use safe string copy functions
.CD-ROM are always mounted read-only

12 years agoMFS: fixes for defects reported by Coverity
Thomas Veerman [Thu, 26 Jul 2012 15:16:11 +0000 (15:16 +0000)]
MFS: fixes for defects reported by Coverity

.use safe string copy
.fix (potential) int overflow in function return

12 years agoMFS: getdents fixes
Thomas Veerman [Thu, 26 Jul 2012 12:42:55 +0000 (12:42 +0000)]
MFS: getdents fixes

.Use a bigger buffer to hold results
.Do not try to store more data than user buffer can hold

12 years agokernel: facility for user-visible memory
Ben Gras [Wed, 18 Jul 2012 16:53:20 +0000 (18:53 +0200)]
kernel: facility for user-visible memory

. map all objects named usermapped_*.o with globally visible
  pages; usermapped_glo_*.o with the VM 'global' bit on, i.e.
  permanently in tlb (very scarce resource!)
. added kinfo, machine, kmessages and loadinfo for a start
. modified log, tty to make use of the shared messages struct

12 years agokernel: few more objects required in unpaged mode
Ben Gras [Sat, 28 Jul 2012 20:26:52 +0000 (22:26 +0200)]
kernel: few more objects required in unpaged mode

. for 64-bit printf()

12 years agokernel: clean up objects
Ben Gras [Sat, 28 Jul 2012 20:21:53 +0000 (22:21 +0200)]
kernel: clean up objects

. objects left over from libraries are left around

12 years agokernel: give kernel tasks their process name back
David van Moolenbroek [Thu, 26 Jul 2012 14:35:08 +0000 (14:35 +0000)]
kernel: give kernel tasks their process name back

12 years agolibsys: 64-bit numbers support for printf()
David van Moolenbroek [Tue, 24 Jul 2012 16:14:48 +0000 (16:14 +0000)]
libsys: 64-bit numbers support for printf()

Change some drivers accordingly.

12 years agoUpgrade to NetBSD: tic.c,v 1.19 05/5/1
Kees Jongenburger [Mon, 23 Jul 2012 17:13:18 +0000 (17:13 +0000)]
Upgrade to NetBSD: tic.c,v 1.19

The upgrade of tic to v 1.19 fixes a memory corruptions that  was
visible when cross building (specially when using  long  path
names in the output file argument).

12 years agoVFS: call req_inhibread again
David van Moolenbroek [Thu, 19 Jul 2012 14:36:51 +0000 (14:36 +0000)]
VFS: call req_inhibread again

12 years agoVFS: do not resolve symlinks in rename(2)
David van Moolenbroek [Wed, 18 Jul 2012 14:59:45 +0000 (14:59 +0000)]
VFS: do not resolve symlinks in rename(2)

12 years agokernel: keep serial info in system env
Ben Gras [Wed, 18 Jul 2012 13:16:57 +0000 (15:16 +0200)]
kernel: keep serial info in system env

. otherwise serial hardware gets reinitialized by tty

12 years agoPFS: fix negative array index
Thomas Veerman [Wed, 18 Jul 2012 10:05:50 +0000 (10:05 +0000)]
PFS: fix negative array index

12 years agoVFS: fix coverity defects
Thomas Veerman [Tue, 17 Jul 2012 10:29:22 +0000 (10:29 +0000)]
VFS: fix coverity defects

12 years agoVFS: remove support for sync FS communication
Thomas Veerman [Tue, 17 Jul 2012 10:12:34 +0000 (10:12 +0000)]
VFS: remove support for sync FS communication

12 years agoVFS: prevent buffer overflow
Thomas Veerman [Tue, 17 Jul 2012 08:47:51 +0000 (08:47 +0000)]
VFS: prevent buffer overflow

If an FS returns faulty struct dirent data, VFS could overflow
a buffer that holds this data.

12 years agovarious coverity-inspired fixes
Ben Gras [Mon, 16 Jul 2012 11:17:11 +0000 (13:17 +0200)]
various coverity-inspired fixes

. some strncpy/strcpy to strlcpy conversions
. new <minix/param.h> to avoid including other minix headers
  that have colliding definitions with library and commands code,
  causing parse warnings
. removed some dead code / assignments

12 years agoVFS: Use safe string copy functions
Thomas Veerman [Fri, 13 Jul 2012 16:08:06 +0000 (16:08 +0000)]
VFS: Use safe string copy functions

12 years agosegmentless smp fixes
Ben Gras [Thu, 12 Jul 2012 22:54:27 +0000 (00:54 +0200)]
segmentless smp fixes

adjust the smp booting procedure for segmentless operation. changes are
mostly due to gdt/idt being dependent on paging, because of the high
location, and paging being on much sooner because of that too.

also smaller fixes: redefine DESC_SIZE, fix kernel makefile variable name
(crosscompiling), some null pointer checks that trap now because of a
sparser pagetable, acpi sanity checking

12 years agoNo more intel/minix segments.
Ben Gras [Mon, 7 May 2012 14:03:35 +0000 (16:03 +0200)]
No more intel/minix segments.

This commit removes all traces of Minix segments (the text/data/stack
memory map abstraction in the kernel) and significance of Intel segments
(hardware segments like CS, DS that add offsets to all addressing before
page table translation). This ultimately simplifies the memory layout
and addressing and makes the same layout possible on non-Intel
architectures.

There are only two types of addresses in the world now: virtual
and physical; even the kernel and processes have the same virtual
address space. Kernel and user processes can be distinguished at a
glance as processes won't use 0xF0000000 and above.

No static pre-allocated memory sizes exist any more.

Changes to booting:
        . The pre_init.c leaves the kernel and modules exactly as
          they were left by the bootloader in physical memory
        . The kernel starts running using physical addressing,
          loaded at a fixed location given in its linker script by the
          bootloader.  All code and data in this phase are linked to
          this fixed low location.
        . It makes a bootstrap pagetable to map itself to a
          fixed high location (also in linker script) and jumps to
          the high address. All code and data then use this high addressing.
        . All code/data symbols linked at the low addresses is prefixed by
          an objcopy step with __k_unpaged_*, so that that code cannot
          reference highly-linked symbols (which aren't valid yet) or vice
          versa (symbols that aren't valid any more).
        . The two addressing modes are separated in the linker script by
          collecting the unpaged_*.o objects and linking them with low
          addresses, and linking the rest high. Some objects are linked
          twice, once low and once high.
        . The bootstrap phase passes a lot of information (e.g. free memory
          list, physical location of the modules, etc.) using the kinfo
          struct.
        . After this bootstrap the low-linked part is freed.
        . The kernel maps in VM into the bootstrap page table so that VM can
          begin executing. Its first job is to make page tables for all other
          boot processes. So VM runs before RS, and RS gets a fully dynamic,
          VM-managed address space. VM gets its privilege info from RS as usual
          but that happens after RS starts running.
        . Both the kernel loading VM and VM organizing boot processes happen
  using the libexec logic. This removes the last reason for VM to
  still know much about exec() and vm/exec.c is gone.

Further Implementation:
        . All segments are based at 0 and have a 4 GB limit.
        . The kernel is mapped in at the top of the virtual address
          space so as not to constrain the user processes.
        . Processes do not use segments from the LDT at all; there are
          no segments in the LDT any more, so no LLDT is needed.
        . The Minix segments T/D/S are gone and so none of the
          user-space or in-kernel copy functions use them. The copy
          functions use a process endpoint of NONE to realize it's
          a physical address, virtual otherwise.
        . The umap call only makes sense to translate a virtual address
          to a physical address now.
        . Segments-related calls like newmap and alloc_segments are gone.
        . All segments-related translation in VM is gone (vir2map etc).
        . Initialization in VM is simpler as no moving around is necessary.
        . VM and all other boot processes can be linked wherever they wish
          and will be mapped in at the right location by the kernel and VM
          respectively.

Other changes:
        . The multiboot code is less special: it does not use mb_print
          for its diagnostics any more but uses printf() as normal, saving
          the output into the diagnostics buffer, only printing to the
          screen using the direct print functions if a panic() occurs.
        . The multiboot code uses the flexible 'free memory map list'
          style to receive the list of free memory if available.
        . The kernel determines the memory layout of the processes to
          a degree: it tells VM where the kernel starts and ends and
          where the kernel wants the top of the process to be. VM then
          uses this entire range, i.e. the stack is right at the top,
          and mmap()ped bits of memory are placed below that downwards,
          and the break grows upwards.

Other Consequences:
        . Every process gets its own page table as address spaces
          can't be separated any more by segments.
        . As all segments are 0-based, there is no distinction between
          virtual and linear addresses, nor between userspace and
          kernel addresses.
        . Less work is done when context switching, leading to a net
          performance increase. (8% faster on my machine for 'make servers'.)
. The layout and configuration of the GDT makes sysenter and syscall
  possible.

12 years agoprofiling related cleanup
Ben Gras [Sat, 16 Jun 2012 16:38:31 +0000 (16:38 +0000)]
profiling related cleanup

. do not declare any data in <minix/profile.h>
. addr check no longer necessary

12 years agoboot programs gzip/strip changes
Ben Gras [Thu, 5 Jul 2012 15:32:15 +0000 (17:32 +0200)]
boot programs gzip/strip changes

as disk space typically isn't a concern when crosscompiling, but
convenience and ability to debug is, change the strip and gzip defaults

. do not strip or gzip the binaries when crosscompiling

this makes it faster to rebuild and restart a compiled system, with
debugging info if so desired.

12 years agoAdd entry in fstab and let the mount system call start devman.
Kees Jongenburger [Tue, 3 Jul 2012 12:57:13 +0000 (12:57 +0000)]
Add entry in fstab and let the mount system call start devman.

After this commit you will need an additional entry in fstab for
the /sys file system (see docs/UPDATING).

12 years agoAllow to specify the rslabel to use for a file system.
Kees Jongenburger [Wed, 4 Jul 2012 12:00:28 +0000 (12:00 +0000)]
Allow to specify the rslabel to use for a file system.

When a file system is mounted some heuristics are used to define
a RS label for that system. This commit allows to specify the
label to use in an optional mount argument using either
mount -o rslabel=fs_myfs or as a mount option in fstab.

This can be used to start services that later also need to be
accessed directly.

12 years agoFixes for mount -a.
Kees Jongenburger [Thu, 5 Jul 2012 09:04:15 +0000 (11:04 +0200)]
Fixes for mount -a.

* Display an error message upon failure to mount a device.
* Handle a special case when the source device is "none"
* pass the mount options stored in fourth field of fstab
  to mount(3).

12 years agoAdd a /sys entry to minix.mtree
Kees Jongenburger [Tue, 3 Jul 2012 11:43:42 +0000 (11:43 +0000)]
Add a /sys entry to minix.mtree

12 years agoCleanup minix.mtree.
Kees Jongenburger [Mon, 2 Jul 2012 16:32:30 +0000 (16:32 +0000)]
Cleanup minix.mtree.

Use consistent indenting and sort entries alphabetically.

12 years agodevmand call the cleanup method before doing an exit.
Kees Jongenburger [Tue, 3 Jul 2012 14:42:08 +0000 (14:42 +0000)]
devmand call the cleanup method before doing an exit.

Call the cleanup mehtod before doing an exit to prevent stale pid
files.

12 years agodon't leave _VCS_REVISION in
Ben Gras [Fri, 29 Jun 2012 14:34:41 +0000 (16:34 +0200)]
don't leave _VCS_REVISION in

12 years agoalign messages
Ben Gras [Fri, 29 Jun 2012 11:39:40 +0000 (13:39 +0200)]
align messages

Has a small but measurable performance improvement.

12 years agovarious symbol referencing fixes
Ben Gras [Thu, 28 Jun 2012 12:56:50 +0000 (14:56 +0200)]
various symbol referencing fixes

make weak symbol references and namespace renames references
the renamed versions.

function renaming, weak symbol references and libc namespace.h
protection interact in hairy ways and causes weak symbol references
for renamed functions to be unresolved; e.g. vfork should be an
alias for _vfork but _vfork doesn't exist because __vfork14()
exists.

this is a problem for dynamically linked executables as all symbols
have to be resolved, used or not, at link time. it was masked by
clang-compiled base system libraries but is a problem when gcc does
it.

12 years agolibvassert: don't build shared lib
Thomas Veerman [Tue, 26 Jun 2012 09:32:02 +0000 (09:32 +0000)]
libvassert: don't build shared lib

12 years agoreleasetools: use build system macros
Thomas Veerman [Tue, 26 Jun 2012 09:31:30 +0000 (09:31 +0000)]
releasetools: use build system macros

12 years agoxz: add libc dependency
Ben Gras [Mon, 25 Jun 2012 15:21:01 +0000 (17:21 +0200)]
xz: add libc dependency

12 years agold.so: be more mmap()-behaviour-independent
Ben Gras [Fri, 22 Jun 2012 09:55:11 +0000 (11:55 +0200)]
ld.so: be more mmap()-behaviour-independent

. if the layout of virtual address regions as returned
  by mmap() without a location hint changes, ld.so could
  trip itself up (under minix). this change allocates
  the full size it needs for every object that's loaded
  so that if that succeeds, it's sure there's virtual address
  space for the whole thing no matter what other bits happen
  to be there already.

. this fix exposed a bug in the test; at atexit() execution
  time the loaded object is unmapped, so that part of the
  test is removed.

12 years agofetch.sh: serve tarballs ourselves
Thomas Veerman [Mon, 25 Jun 2012 10:46:31 +0000 (10:46 +0000)]
fetch.sh: serve tarballs ourselves

And use the original sources as backup.

12 years agoImport lib{mpc,gmp,mpfr}
Thomas Veerman [Mon, 25 Jun 2012 09:15:27 +0000 (09:15 +0000)]
Import lib{mpc,gmp,mpfr}

This decreases external dependencies for crosscompilation. Note that
these libraries are not built nor used by Minix itself.

Furthermore, the shell scripts that download the tarballs for these
libraries, gcc, binutils, and gmake now also support curl in addition
to wget.

12 years agosed: unbreak crosscompilation
Thomas Veerman [Thu, 21 Jun 2012 14:07:59 +0000 (14:07 +0000)]
sed: unbreak crosscompilation