]> Zhao Yanbai Git Server - minix.git/log
minix.git
15 years agomerging in memory freeer.
Ben Gras [Mon, 8 Jun 2009 15:31:45 +0000 (15:31 +0000)]
merging in memory freeer.

15 years agoquick hack (sorry) for making *sdevio* work to other
Ben Gras [Mon, 8 Jun 2009 14:33:15 +0000 (14:33 +0000)]
quick hack (sorry) for making *sdevio* work to other
processes than the caller..

also disable kernel sanity checks

15 years agorebase from trunk
Ben Gras [Mon, 8 Jun 2009 12:25:24 +0000 (12:25 +0000)]
rebase from trunk

15 years ago..
Ben Gras [Mon, 8 Jun 2009 07:33:42 +0000 (07:33 +0000)]
..

15 years ago(no commit message)
Ben Gras [Mon, 8 Jun 2009 06:35:18 +0000 (06:35 +0000)]

15 years agocleanup, sanity checking
Ben Gras [Mon, 8 Jun 2009 06:08:11 +0000 (06:08 +0000)]
cleanup, sanity checking

15 years agoallow empty senda
Ben Gras [Mon, 8 Jun 2009 04:39:26 +0000 (04:39 +0000)]
allow empty senda

15 years agominor cleanup
Ben Gras [Mon, 8 Jun 2009 04:30:16 +0000 (04:30 +0000)]
minor cleanup

15 years agokeep some processes mapped in always; direct message copying
Ben Gras [Mon, 8 Jun 2009 04:02:22 +0000 (04:02 +0000)]
keep some processes mapped in always; direct message copying
where possible (no buffering); no more explicit vm checkranges
in kernel; new allocator for vm using avl tree without needing
remapping

15 years agominor cleanup
Ben Gras [Sun, 7 Jun 2009 16:25:37 +0000 (16:25 +0000)]
minor cleanup

15 years agopde cache check works
Ben Gras [Sun, 7 Jun 2009 15:55:44 +0000 (15:55 +0000)]
pde cache check works

no more silly vm checkranges

15 years agomy state.
Ben Gras [Sat, 6 Jun 2009 23:27:10 +0000 (23:27 +0000)]
my state.

trying to get some memory optimisation (less pagetable reloading,
less tlb purging) features working smoothly.

to be documented when committing to trunk :)

15 years agoonly switch pagetable if necessary and it's different for copying messages
Ben Gras [Wed, 3 Jun 2009 15:28:13 +0000 (15:28 +0000)]
only switch pagetable if necessary and it's different for copying messages

15 years agomove field offset from vir region from phys_block to phys_region, so
Ben Gras [Wed, 3 Jun 2009 12:18:13 +0000 (12:18 +0000)]
move field offset from vir region from phys_block to phys_region, so
the same blocks of physical memory can be mapped in in different offsets
within regions.

15 years agostate
Ben Gras [Wed, 3 Jun 2009 11:22:49 +0000 (11:22 +0000)]
state

15 years agofurther messing with page fault handling
Ben Gras [Fri, 29 May 2009 18:47:31 +0000 (18:47 +0000)]
further messing with page fault handling

15 years agono vm_setbuf any more (kernel doesn't create its page table any more),
Ben Gras [Thu, 28 May 2009 14:23:38 +0000 (14:23 +0000)]
no vm_setbuf any more (kernel doesn't create its page table any more),
no relocking field (locks not checked)

15 years agobeng work in progress, to be explained in a future commit message :).
Ben Gras [Thu, 28 May 2009 13:47:20 +0000 (13:47 +0000)]
beng work in progress, to be explained in a future commit message :).

15 years agounused confusing malloc/free functions in sysutil
Ben Gras [Thu, 28 May 2009 12:14:37 +0000 (12:14 +0000)]
unused confusing malloc/free functions in sysutil

15 years agocopy of beng's working copy
Ben Gras [Wed, 20 May 2009 16:54:58 +0000 (16:54 +0000)]
copy of beng's working copy

15 years agoupdate readlink(2) manpage accordingly
David van Moolenbroek [Wed, 20 May 2009 11:49:10 +0000 (11:49 +0000)]
update readlink(2) manpage accordingly

15 years agomove symlink type check for readlink() into VFS, and return the right POSIX error
David van Moolenbroek [Wed, 20 May 2009 09:46:06 +0000 (09:46 +0000)]
move symlink type check for readlink() into VFS, and return the right POSIX error

15 years agoVFS consistency: use I_PIPE/NO_PIPE when checking v_pipe
David van Moolenbroek [Tue, 19 May 2009 14:34:44 +0000 (14:34 +0000)]
VFS consistency: use I_PIPE/NO_PIPE when checking v_pipe

15 years agodon't include /usr/src changes in packages.
Ben Gras [Tue, 19 May 2009 10:55:15 +0000 (10:55 +0000)]
don't include /usr/src changes in packages.

15 years ago4MB mask
Ben Gras [Mon, 18 May 2009 15:56:13 +0000 (15:56 +0000)]
4MB mask

15 years agomap in kernel in 4MB, global-bit-set 'pages' if hardware
Ben Gras [Mon, 18 May 2009 15:34:42 +0000 (15:34 +0000)]
map in kernel in 4MB, global-bit-set 'pages' if hardware
supports it. helps performance.

broken use of region data structure for kernel mapping to be fixed.

15 years agoVarious VFS and MFS fixes to improve correctness, consistency and
David van Moolenbroek [Mon, 18 May 2009 11:27:12 +0000 (11:27 +0000)]
Various VFS and MFS fixes to improve correctness, consistency and
POSIX compliance.

VFS changes:
* truncate() on a file system mounted read-only no longer panics MFS.
* ftruncate() and fcntl(F_FREESP) now check for write permission on
  the file descriptor instead of the file, write().
* utime(), chown() and fchown() now check for file system read-only
  status.

MFS changes:
* link() and rename() no longer return the internal EENTERMOUNT and
  ELEAVEMOUNT errors to the application as part of a check on the
  source path.
* rename() now treats EENTERMOUNT from the destination path check as
  an error, preventing file system corruption from renaming a normal
  directory to an existing mountpoint directory.
* mountpoints (mounted-on dirs) are hidden better during lookups:
  - if a lookup starts from a mountpoint, the first component has to
    be ".." (anything else being a VFS-FS protocol violation).
  - in that case, the permissions of the mountpoint are not checked.
  - in all other cases, visiting a mountpoint always results in
    EENTERMOUNT.
* a lookup on ".." from a mount root or chroot(2) root no longer
  succeeds if the caller does not have search permission on that
  directory.
* POSIX: getdents() now updates directory access times.
* POSIX: readlink() now returns partial results instead of ERANGE.

Miscellaneous changes:
* semaphore file handling bug (leading to hangs) fixed in test 32.

The VFS changes should now put the burden of checking for read-only
status of file systems entirely on VFS, and limit the access
permission checks that file systems have to perform, to checking
search permission on directories during lookups. From this point on,
any deviation from that spceification should be considered a bug.
Note that for legacy reasons, the root partition is assumed to be
mounted read-write.

15 years agoLibrary call for cpu features; make kernel and vm use this to query cpu
Ben Gras [Fri, 15 May 2009 17:07:36 +0000 (17:07 +0000)]
Library call for cpu features; make kernel and vm use this to query cpu
features (specifically: 4MB pages and TLB global bit).  Only enable
these features in CR4 if available. 4MB pages to be used in the near
future.

15 years agocorrect the revision variable used in /etc/version
Ben Gras [Fri, 15 May 2009 13:02:23 +0000 (13:02 +0000)]
correct the revision variable used in /etc/version

15 years ago-Remove qemu_pci boot variable. Useres no longer need to set qemu_pci when
Arun Thomas [Thu, 14 May 2009 19:07:37 +0000 (19:07 +0000)]
-Remove qemu_pci boot variable. Useres no longer need to set qemu_pci when
booting MINIX under QEMU/KVM.
-Kept the diagnostic message, however.

15 years agoC CPUID interface.
Ben Gras [Thu, 14 May 2009 15:55:28 +0000 (15:55 +0000)]
C CPUID interface.

15 years ago-Installation info is on the wiki now, so remove setup guides and update
Arun Thomas [Thu, 14 May 2009 15:54:02 +0000 (15:54 +0000)]
-Installation info is on the wiki now, so remove setup guides and update
README.
-Remove obsolete FAT partitioning utility.
-Update startup banner.

15 years agoClear trace bit for child on fork.
David van Moolenbroek [Wed, 13 May 2009 21:58:10 +0000 (21:58 +0000)]
Clear trace bit for child on fork.

Without this, a forking single-stepped process will have its child
die from a TRAP signal right away.

15 years agosupport in 'mount' for specifying file system type and options
David van Moolenbroek [Wed, 13 May 2009 15:39:44 +0000 (15:39 +0000)]
support in 'mount' for specifying file system type and options

15 years ago'service' null pointer dereference fix
David van Moolenbroek [Tue, 12 May 2009 16:52:00 +0000 (16:52 +0000)]
'service' null pointer dereference fix

15 years agoto let tty run with its own page table (instead of with the kernel identity
Ben Gras [Tue, 12 May 2009 12:43:18 +0000 (12:43 +0000)]
to let tty run with its own page table (instead of with the kernel identity
map table), make it map in video memory.

sadly, this breaks tty in non-paged mode.

happily, this simplifies the code by throwing out the messing
around with segments, and throws out vidcopy.s.

15 years agodon't bother with unmapping physical memory in non-sanitycheck mode.
Ben Gras [Tue, 12 May 2009 11:51:53 +0000 (11:51 +0000)]
don't bother with unmapping physical memory in non-sanitycheck mode.

15 years agobasic sparepage optimisation
Ben Gras [Tue, 12 May 2009 11:38:29 +0000 (11:38 +0000)]
basic sparepage optimisation

some simplification of linear/virtual address translation
(less hardcoding and more use of arch_*2* functions)

15 years agoremove bogus arch_map2vir calls
Ben Gras [Tue, 12 May 2009 11:36:15 +0000 (11:36 +0000)]
remove bogus arch_map2vir calls

15 years agodon't call this function with a bogus addr (done by region.c print code)
Ben Gras [Tue, 12 May 2009 11:35:49 +0000 (11:35 +0000)]
don't call this function with a bogus addr (done by region.c print code)

15 years agomore sanity checking. sanity checking disabled by default.
Ben Gras [Tue, 12 May 2009 11:35:01 +0000 (11:35 +0000)]
more sanity checking. sanity checking disabled by default.

give every process a full pagetable by default now.

first step to disabling kernel page table code (processes
might not have page tables -> no address translation).

15 years agolarge page size constant.
Ben Gras [Tue, 12 May 2009 09:23:27 +0000 (09:23 +0000)]
large page size constant.

15 years agostart all processes, including VM, in VM_PROCSTART in linear address space,
Ben Gras [Mon, 11 May 2009 19:11:37 +0000 (19:11 +0000)]
start all processes, including VM, in VM_PROCSTART in linear address space,
to make space for kernel to be able to map in things below there.

15 years agoDon't build arch objs more than once.
Ben Gras [Mon, 11 May 2009 19:05:45 +0000 (19:05 +0000)]
Don't build arch objs more than once.

15 years agoremove debug msg
Ben Gras [Mon, 11 May 2009 11:57:20 +0000 (11:57 +0000)]
remove debug msg

15 years agoVFS-FS error codes: better safe than sorry
David van Moolenbroek [Mon, 11 May 2009 11:30:04 +0000 (11:30 +0000)]
VFS-FS error codes: better safe than sorry

15 years agoLimited support for nested FS->VFS requests during VFS->FS call.
David van Moolenbroek [Mon, 11 May 2009 10:02:28 +0000 (10:02 +0000)]
Limited support for nested FS->VFS requests during VFS->FS call.
- Changed VFS-FS protocol to only store OK or negative error code in
  m_type field of reply messages.
- Changed VFS to treat nonzero positive replies from FS as requests.
- Added backwards compatibility to VFS and MFS.
No protection of global data structures is provided in VFS, so many
VFS calls cannot be made safely by FS servers during many FS calls.
Use with caution (or, preferably, not at all).

15 years agowrong field index in at_wini env_parse call
David van Moolenbroek [Sun, 10 May 2009 16:58:23 +0000 (16:58 +0000)]
wrong field index in at_wini env_parse call

15 years agopotential buffer overruns in env_* routines
David van Moolenbroek [Sun, 10 May 2009 16:54:37 +0000 (16:54 +0000)]
potential buffer overruns in env_* routines

15 years agoregression fix: vfs lookup passes incorrect chroot information after crossing mountpoints
David van Moolenbroek [Sat, 9 May 2009 17:53:22 +0000 (17:53 +0000)]
regression fix: vfs lookup passes incorrect chroot information after crossing mountpoints

15 years agofix for commands that try to include mfs constants from vfs
David van Moolenbroek [Fri, 8 May 2009 20:37:06 +0000 (20:37 +0000)]
fix for commands that try to include mfs constants from vfs

15 years agoquick cleanup of old mfs cruft from vfs
David van Moolenbroek [Fri, 8 May 2009 14:12:41 +0000 (14:12 +0000)]
quick cleanup of old mfs cruft from vfs

15 years ago make unpause() decrease susp_count, as it shouldn't be decreased
Ben Gras [Fri, 8 May 2009 13:56:41 +0000 (13:56 +0000)]
 make unpause() decrease susp_count, as it shouldn't be decreased
 if the process was REVIVING. (susp_count doesn't count those
 processes.) this together with dev_io SELECT suspend side effect
 for asynch. character devices solves the hanging pipe bug. or
 at last vastly improves it.

 added sanity checks, turned off by default.

 made the {NOT_,}{SUSPENDING,REVIVING} constants weirder to
 help sanity checking.

15 years agoremove unused global variable from vfs
David van Moolenbroek [Fri, 8 May 2009 13:54:01 +0000 (13:54 +0000)]
remove unused global variable from vfs

15 years agodon't suspend the process as a side-effect if
Ben Gras [Fri, 8 May 2009 13:50:29 +0000 (13:50 +0000)]
don't suspend the process as a side-effect if
device returns SUSPEND if it's select; select already
does this.

15 years agoReenable RS table dump from IS (Shift+F6)
David van Moolenbroek [Fri, 8 May 2009 12:38:14 +0000 (12:38 +0000)]
Reenable RS table dump from IS (Shift+F6)

15 years agoFix for large transfer operations not advancing buffer address
David van Moolenbroek [Fri, 8 May 2009 12:29:57 +0000 (12:29 +0000)]
Fix for large transfer operations not advancing buffer address
offset when DMA transfer unit is smaller than given buffer size.
Bug tracker item #82.

15 years agoKernel interrupt hook management fixes:
David van Moolenbroek [Thu, 7 May 2009 14:52:07 +0000 (14:52 +0000)]
Kernel interrupt hook management fixes:
- properly assign unique hook IDs
- after hook removal, remove hook-specific interrupt disable flag

15 years agodisable scary looking debug messages.
Ben Gras [Thu, 7 May 2009 09:58:16 +0000 (09:58 +0000)]
disable scary looking debug messages.

15 years agoturn off scary looking debug messages.
Ben Gras [Thu, 7 May 2009 09:57:43 +0000 (09:57 +0000)]
turn off scary looking debug messages.

15 years agoignore errors of pipe read (can happen with shutdown now,
Ben Gras [Wed, 6 May 2009 15:38:32 +0000 (15:38 +0000)]
ignore errors of pipe read (can happen with shutdown now,
now that all fd's are closed neatly in vfs), change messaging
in unexpected restarts

15 years agocleanup of vfs shutdown logic; makes clean unmounts easier (but
Ben Gras [Wed, 29 Apr 2009 16:59:18 +0000 (16:59 +0000)]
cleanup of vfs shutdown logic; makes clean unmounts easier (but
needs checking if fp_wd or fp_rd is NULL before use)

15 years ago - delete unused .h files
Ben Gras [Mon, 27 Apr 2009 16:13:51 +0000 (16:13 +0000)]
 - delete unused .h files
 - slight code cleanup
 - neater exit procedure: exit when unmount
   message received and kill signal (from RS 'down' or
   reboot/shutdown) received (speed up unmount, but don't
   confuse VFS by exiting before/during unmount msg)

15 years agospurious debug
Ben Gras [Mon, 27 Apr 2009 16:11:38 +0000 (16:11 +0000)]
spurious debug

15 years agolose -s flag for umount.
Ben Gras [Mon, 27 Apr 2009 14:23:57 +0000 (14:23 +0000)]
lose -s flag for umount.

15 years agoprocess restarts are pretty rare/serious.
Ben Gras [Mon, 27 Apr 2009 14:07:47 +0000 (14:07 +0000)]
process restarts are pretty rare/serious.

15 years agominor fixes to install script - allow 8kB blocksize (necessary for
Ben Gras [Mon, 27 Apr 2009 12:59:49 +0000 (12:59 +0000)]
minor fixes to install script - allow 8kB blocksize (necessary for
large filesystems)

15 years agoWhen we receive a PROC_EVENT message from PM, exit nicely; avoids
Ben Gras [Mon, 27 Apr 2009 12:02:31 +0000 (12:02 +0000)]
When we receive a PROC_EVENT message from PM, exit nicely; avoids
annoying graceful RS timeout on unmount.

15 years agobad dma fallback to pio mode
Ben Gras [Mon, 27 Apr 2009 11:53:11 +0000 (11:53 +0000)]
bad dma fallback to pio mode

15 years agosome more vm bits.
Ben Gras [Thu, 23 Apr 2009 15:21:03 +0000 (15:21 +0000)]
some more vm bits.

15 years agoset global flag for kernel pages, so tlb entries for kernel aren't thrown
Ben Gras [Thu, 23 Apr 2009 15:11:16 +0000 (15:11 +0000)]
set global flag for kernel pages, so tlb entries for kernel aren't thrown
away on cr3 reload. minor optimization.

15 years agoMove queue.h to include/sys so that it can be used elsewhere. Pull in
Arun Thomas [Wed, 22 Apr 2009 20:02:39 +0000 (20:02 +0000)]
Move queue.h to include/sys so that it can be used elsewhere. Pull in
FreeBSD's r179210 queue.h.

15 years agoDisable IOMMU warnings.
Arun Thomas [Wed, 22 Apr 2009 16:53:46 +0000 (16:53 +0000)]
Disable IOMMU warnings.

15 years agoMake the rtl8139 and orinoco drivers handle the system shutdown case
Arun Thomas [Wed, 22 Apr 2009 12:42:37 +0000 (12:42 +0000)]
Make the rtl8139 and orinoco drivers handle the system shutdown case
like other drivers. Also, some minor cleanups.

15 years agomark pages whose refcount were >1 and drop to 1 and are
Ben Gras [Wed, 22 Apr 2009 12:39:29 +0000 (12:39 +0000)]
mark pages whose refcount were >1 and drop to 1 and are
read/write writable in the pagetable right away instead of waiting for
a pagefault. minor optimization.

some a sanity check of SLAB-allocated pointers.

vm gets its own _exit and __exit like PM, so the stock (library) panic works.

15 years ago - enable ipc warnings by default
Ben Gras [Fri, 17 Apr 2009 13:46:37 +0000 (13:46 +0000)]
 - enable ipc warnings by default
 - ipc checking code in kernel didn't properly catch the
   sendrec() to self case; added special case check
 - triggered by PM using stock panic() - needs its own _exit()

reported by Joren l'Ami.

15 years agoclarify not found error message a bit.
Ben Gras [Tue, 14 Apr 2009 14:16:24 +0000 (14:16 +0000)]
clarify not found error message a bit.

15 years agodon't flush output for SIGWINCH. found by Joren l'Ami.
Ben Gras [Mon, 6 Apr 2009 09:39:42 +0000 (09:39 +0000)]
don't flush output for SIGWINCH. found by Joren l'Ami.

15 years agounnecessary debugging message
Ben Gras [Thu, 2 Apr 2009 16:43:35 +0000 (16:43 +0000)]
unnecessary debugging message

15 years agomoved type and constants for random data to include file;
Ben Gras [Thu, 2 Apr 2009 15:24:44 +0000 (15:24 +0000)]
moved type and constants for random data to include file;
added consistency check in random; added source of randomness
internal to random using timing; only retrieve random bins that are full.

15 years agoprint who the message is from.
Ben Gras [Thu, 2 Apr 2009 11:56:50 +0000 (11:56 +0000)]
print who the message is from.

15 years agodon't make susp_count negative.
Ben Gras [Thu, 2 Apr 2009 11:44:26 +0000 (11:44 +0000)]
don't make susp_count negative.

15 years agotrace fix contributed by Joren l'Ami
Ben Gras [Thu, 2 Apr 2009 11:38:23 +0000 (11:38 +0000)]
trace fix contributed by Joren l'Ami

15 years agoincrement nph when printing physical regions; suggested by Guanqun Lu
Ben Gras [Tue, 31 Mar 2009 14:26:24 +0000 (14:26 +0000)]
increment nph when printing physical regions; suggested by Guanqun Lu

15 years agochange DmaMode checks from DEV_WRITE to DEV_WRITE_S
Ben Gras [Tue, 31 Mar 2009 14:23:33 +0000 (14:23 +0000)]
change DmaMode checks from DEV_WRITE to DEV_WRITE_S

15 years agoTweak 64-bit integer type declarations; Fixes GCC sysutil build
Arun Thomas [Mon, 30 Mar 2009 17:07:39 +0000 (17:07 +0000)]
Tweak 64-bit integer type declarations; Fixes GCC sysutil build
breakage.

15 years agosuppress more mostly-harmless messages.
Ben Gras [Thu, 26 Mar 2009 16:11:27 +0000 (16:11 +0000)]
suppress more mostly-harmless messages.

15 years agosuppress these noisy, alarming messages.
Ben Gras [Thu, 26 Mar 2009 15:56:08 +0000 (15:56 +0000)]
suppress these noisy, alarming messages.

15 years agono 'small minix' option
Ben Gras [Thu, 26 Mar 2009 15:54:09 +0000 (15:54 +0000)]
no 'small minix' option

15 years agodon't need to explicitly enable lance any more.
Ben Gras [Thu, 26 Mar 2009 15:23:07 +0000 (15:23 +0000)]
don't need to explicitly enable lance any more.

15 years agoignore linmem.
Ben Gras [Thu, 26 Mar 2009 15:22:08 +0000 (15:22 +0000)]
ignore linmem.
don't disable the driver by default.

15 years agoSizeMB isn't used any more.
Ben Gras [Thu, 26 Mar 2009 13:19:14 +0000 (13:19 +0000)]
SizeMB isn't used any more.

15 years agodon't execute hlt in real mode.
Ben Gras [Tue, 24 Mar 2009 16:08:10 +0000 (16:08 +0000)]
don't execute hlt in real mode.
workaround for getting minix under qemu kvm.
Ameya, ape800 at few.vu.nl.

15 years agomore inodes than default for /usr on cd.
Ben Gras [Tue, 24 Mar 2009 15:41:51 +0000 (15:41 +0000)]
more inodes than default for /usr on cd.

15 years agoworkaround for qemu writing the configuration byte on the AUX port pre-0.10.
Ben Gras [Tue, 24 Mar 2009 15:41:18 +0000 (15:41 +0000)]
workaround for qemu writing the configuration byte on the AUX port pre-0.10.

15 years agomake USRMB settable
Ben Gras [Thu, 19 Mar 2009 14:14:57 +0000 (14:14 +0000)]
make USRMB settable

15 years agomultiple ram disks; also make release process use this
Ben Gras [Thu, 19 Mar 2009 13:48:19 +0000 (13:48 +0000)]
multiple ram disks; also make release process use this
so temporary partitions aren't needed any more.

15 years agoIncrease ARG_MAX
Arun Thomas [Tue, 17 Mar 2009 07:38:35 +0000 (07:38 +0000)]
Increase ARG_MAX

15 years agoUpdate setup's NIC selection dialog
Arun Thomas [Fri, 6 Mar 2009 17:45:22 +0000 (17:45 +0000)]
Update setup's NIC selection dialog

15 years agoDefine 64 bit integer types when compiling with GCC
Arun Thomas [Fri, 6 Mar 2009 16:56:46 +0000 (16:56 +0000)]
Define 64 bit integer types when compiling with GCC