Age | Commit message (Collapse) | Author |
|
* vm/vm_resident.c (vm_page_grab): Only block unprivileged
allocations over vm_page_external_limit if they are actually
external allocations. Reported by Mark Kettenis
(kettenis@wins.uva.nl>.
|
|
* i386/i386at/model_dep.c (init_alloc_aligned): Revert previous
change except for the one to AVAIL_NEXT. All the rest of the
previous code was correct. Correction from Momchil Velikov
(velco@fadata.bg).
|
|
|
|
|
|
* Drivers.macros (AC_DRIVER): Use `test' instead of `[ ... ]'.
Reported by "Momchil 'Velco' Velikov" <velco@fadata.bg>.
|
|
|
|
Force subdirectory targets so they get recompiled properly.
* i386/Makefile.in (linux/linux.o): Depend on FORCE instead of linux.
(FORCE): New target.
* i386/Makefrag (i386/sysdep.o, FORCE): Likewise.
|
|
* linux/dev/drivers/block/genhd.c (check_partition): Change message
to say these are only the DOS partitions.
* i386/linux/dev/include/linux/autoconf.h (CONFIG_BSD_DISKLABEL,
CONFIG_SMD_DISKLABEL): #undef these. The dev/glue/block.c code
deals with BSD disklabels and expects the Linux genhd.c code to
handle only the DOS partitions; having genhd.c do BSD disklabels
too just confuses things.
|
|
* i386/i386at/model_dep.c (init_alloc_aligned): Make the checks
correctly for addresses falling into skip regions. Reported by
Momchil Velikov (velco@fadata.bg).
|
|
* i386/linux/Makefile.in: For each .o file in $(linux-objs)
include a corresponding .d file, unless `no_deps' is `t'.
(make-deps): New function.
(%.d): New pattern rules for building %.d from %.c and %.S.
|
|
* i386/linux/Makefile.in: For each .o file in $(linux-objs)
include a corresponding .d file, unless `no_deps' is `t'.
(make-deps): New function.
(%.d): New pattern rules for building %.d from %.c and %.S.
|
|
* vm/vm_resident.c (vm_page_grab): Don't bounce requests when
vm_page_external_count == vm_page_external_limit, so that before
initialization is complete we can still allocate memory without
relying on current_thread.
(vm_page_wait): Also block if we are over the external page
limit.
* vm/vm_pageout.c (vm_pageout_scan): Rewrite the
scan-for-pages-to-pageout loop to not crash when we hit the end of
the inactive queue; instead jump back to the
pause-for-a-little-while code.
|
|
|
|
* vm/vm_resident.c (vm_page_grab): Don't bounce requests when
vm_page_external_count == vm_page_external_limit, so that before
initialization is complete we can still allocate memory without
relying on current_thread.
|
|
|
|
|
|
* vm/vm_pageout.c (vm_pageout): Initialize vm_page_external_limit
and vm_page_external_target.
|
|
* vm/vm_resident.c (vm_page_external_limit): Define new variable.
|
|
|
|
|
|
|
|
|
|
* vm/vm_resident.c (vm_page_external_count): Define variable.
(vm_page_grab): New argument `external'. All callers changed.
Keep track of number of externally managed pages. Don't let
non-privileged threads exceed the externally-managed page limit.
(vm_page_grab_contiguous_pages): New argument `external'. All
callers changed. Keep track of number of externally managed
pages. Don't let non-privileged threads exceed the
externally-managed page limit.
(vm_page_convert): New argument `external'. All callers changed.
(vm_page_release): New argument `external'. All callers
changed. Keep track of number of externally managed pages.
(vm_page_bootstrap): Initialize M->external.
* vm/vm_page.h (vm_page_external_limit, vm_page_external_count):
New variables.
(struct vm_page): New members `external' and `extcounted'.
* vm/vm_pageout.c (vm_pageout): Initialize vm_page_external_limit
and vm_page_external_target.
(VM_PAGE_EXTERNAL_LIMIT, VM_PAGE_EXTERNAL_TARGET): New macro.
(vm_pageout_external_target): New variable.
(vm_pageout_scan): Regard "too many externally managed pages" as a
reason to keep doing work, but if that's the only reason we're
doing work, then the only thing we do is schedule cleaning of
pages. Help keep track of the number of externally managed pages
that we care about.
* vm/vm_pageout.c (VM_PAGEOUT_BURST_WAIT): Reduce to 10ms/page.
(VM_PAGEOUT_EMPTY_WAIT): Reduce to 75 ms.
(VM_PAGE_FREE_RESERVED): Increase to 50 pages.
(VM_PAGEOUT_RESERVED_INTERNAL): Adjust to `(reserve) - 25'.
(VM_PAGEOUT_RESERVED_REALLY): Adjust to `(reserve) - 40'.
|
|
* vm/vm_resident.c (vm_page_external_count): Define variable.
(vm_page_grab): New argument `external'. All callers changed.
Keep track of number of externally managed pages. Don't let
non-privileged threads exceed the externally-managed page limit.
(vm_page_grab_contiguous_pages): New argument `external'. All
callers changed. Keep track of number of externally managed
pages. Don't let non-privileged threads exceed the
externally-managed page limit.
(vm_page_convert): New argument `external'. All callers changed.
(vm_page_release): New argument `external'. All callers
changed. Keep track of number of externally managed pages.
(vm_page_bootstrap): Initialize M->external.
* vm/vm_page.h (vm_page_external_limit, vm_page_external_count):
New variables.
(struct vm_page): New members `external' and `extcounted'.
* vm/vm_pageout.c (vm_pageout): Initialize vm_page_external_limit
and vm_page_external_target.
(VM_PAGE_EXTERNAL_LIMIT, VM_PAGE_EXTERNAL_TARGET): New macro.
(vm_pageout_external_target): New variable.
(vm_pageout_scan): Regard "too many externally managed pages" as a
reason to keep doing work, but if that's the only reason we're
doing work, then the only thing we do is schedule cleaning of
pages. Help keep track of the number of externally managed pages
that we care about.
|
|
|
|
|
|
|
|
|
|
* i386/Files: Don't mention i386/pc/Makerules,
i386/include/Makerules, or i386/util/Makerules.
|
|
|
|
* Makefile.in (all-real-archs): Delete variable.
(dist): No longer make `machine' symlink.
|
|
* Makefile.in (dist): Sanify file permissions before building tar
file.
|
|
* linux/Files: New file.
* linux/Subdirs: New file.
* Makefile.in (all-archs): Add `linux'.
(all-real-archs): New variable.
(dist): Create `machine' symlink only on
$(all-real-archs).
|
|
* i386/Subdirs: Remove i386/i386at/boot.
Sun Jun 20 18:33:59 1999 Thomas Bushnell, BSG <tb@mit.edu>
* i386/Files: Regenerated.
* i386/Subdirs: Regenerated.
|
|
* version.c (version): Updated to version 1.2.
* i386/Files: Regenerated.
* i386/Subdirs: Regenerated.
* linux/Files: New file.
* linux/Subdirs: New file.
* Makefile.in (all-archs): Add `linux'.
(all-real-archs): New variable.
(dist): Create `machine' symlink only on
$(all-real-archs).
* Makefile.in (dist): Use `cp' instead of `ln'.
* config.guess: Updated from autoconf 2.13.
* config.sub: Likewise.
|
|
* linux/dev/drivers/net/Space.c (ethif_probe): Probe tc59x_probe
(CONFIG_VORTEX) after el3_probe (CONFIG_EL3), because the latter
card matches the former probe, but the driver doesn't work with
it. Reported by Marcus Brinkmann <brinkmd@debian.org>.
|
|
* ipc/ipc_init.c (ipc_bootstrap): Don't make the IOT_PORT zone
exhaustible, because allocation of ports for internal objects
panics if zalloc fails.
|
|
* Drivers.macros (AC_DRIVER): Only turn on driver if $enableval is
not `no'. Reported by Kalle Olavi Niemitalo <tosi@ees2.oulu.fi>.
|
|
|
|
* linux/dev/drivers/block/genhd.c [MACH] (mach_minor): New variable.
(add_bsd_partition): New function.
(bsd_disklabel_partition) [MACH]: Call add_bsd_partition.
(msdos_partition) [MACH]: Set mach_minor to MINOR before
calling bsd_disklabel_partition.
* linux/dev/drivers/block/ide.c (do_request) [MACH]: Fix "bad access"
message.
|
|
* linux/dev/drivers/block/genhd.c [MACH] (mach_minor): New variable.
(add_bsd_partition): New function.
(bsd_disklabel_partition) [MACH]: Call add_bsd_partition.
(msdos_partition) [MACH]: Set mach_minor to MINOR before
calling bsd_disklabel_partition.
|
|
|
|
* kern/bootstrap.c (build_args_and_stack): If kernel_cmdline is
nonempty, give the bootstrap task an environment variable of
"MULTIBOOT_CMDLINE=kernel command line" on its stack.
|
|
|
|
* kern/startup.c (start_kernel_threads): Call record_time_stamp to
reset the kernel task's creation_time field, so it's not zero.
|
|
|
|
* i386/i386/fpu.c (fp_free): Clear task switch flag before calling
fwait.
(fpu_get_state): Only save FPU state info if the live FPU state
belongs to our target, i.e. if THREAD is the current thread.
(fp_load): Print warning if we try to load an invalid FPU state,
and reset the FPU to some sane state instead.
|
|
|
|
* kern/task.c (task_info): Allow count for TASK_BASIC_INFO to be less
two words, not just one. The new member creation_time is two words.
|
|
|