call to __main(), and therefore saves the size of the call and the
size of a stub implementation of __main().
in the primary boot block, don't bother saving/restoring the argument
passed in from the caller. There is no such argument (that we care
about, at least) to the primary. (for secondary, it's the firmware
FD being used.)
Clean up the "Region 1" related definitions, and define load addresses,
max load size, and max total size for as many boot block types as we can.
(types = unified, primary, secondary). We can't always define all
values for all boot blocks, though.
Make CPP flags selection less gross.
Use objcopy rather than headersize (yay, evil gets a stake to the heart!).
Use a little shell script to verify that the sizes of the boot blocks are OK.
Do not compile too much more of libsa than we actually have to.
size to be reduced substantially. (backward compatibility verified
by compiling one of the alpha boot blocks which uses all of the code
before and after, diffing the object files, and manually verifying that
the differences were 'correct'. some differences were "unavoidable,"
it wanting to avoid a double-commit, because e.g. local variables which
were previously used were no longer used.) a README which describes
supported options (or at least the ones mentioned below) is forthcoming.
add support for the preprocessor macro LIBSA_NO_TWIDDLE, which
causes calls to twiddle() to be omitted if it's defined.
add support for the preprocessor macros:
LIBSA_NO_FS_CLOSE
LIBSA_NO_FS_WRITE
LIBSA_NO_FS_SEEK
which, if defined, cause the corresponding file system operations
in the individual file system implementations to be omitted. (note
that all of those macros are not supported by all file systems at
this point. comments were added to individual file system files
to indicate lack of support, and should be cleaned up later. Backward
compatibility options e.g. UFS_NOCLOSE, etc., are supported.)
add support for the preprocessor macro LIBSA_NO_FS_SYMLINK, which
removes support for symbolic links from the file system support
functions. (same notes as for the macros above apply.)
add support for the preprocessor macro LIBSA_FS_SINGLECOMPONENT which
removes all subdirectory and symlink support from the file system
support functions. (same notes as for the macros above apply.)
add support for the preprocessor macro LIBSA_NO_FD_CHECKING, which
causes code relating to libsa file descriptor checks (e.g. range
checking and checking that a file descriptor is valid) to be
omitted if it's defined.
add support for the preprocessor macro LIBSA_NO_RAW_ACCESS, which
causes code relating to raw device access to be omitted if it's
defined.
change some structure copies to use bcopy() instead. that way
use of bcopy vs. memcpy() can easily be selected by
LIBSA_USE_MEMCPY. (without changes like these, you could end up
having both bcopy() and memcpy() included. eventually, all
calls to bcopy should be changed to calls to memcpy() or memmove()
as appropriate -- hopefully never the latter -- with an option to
use bcopy instead.)
add support for the preprocessor macro LIBSA_NO_DISKLABEL_MSGS, which
causes disklabel() to return '1' as msg rather than a string. Can
be used if the boot blocks don't care about the string, and need to
save the space.
add support for the preprocessor macro LIBSA_SINGLE_FILESYSTEM, which
if defined causes all of the file system switch code to be removed.
Its value should be the name of the file system supported by the
boot block, e.g. "ufs" for the FFS file system. calls to the
file system functions open, close, etc., which were previously
done through a function switch are then done via direct invocation
of <fs>_open, <fs>_close, etc. (e.g. ufs_open, ...).
add support for the preprocessor macro LIBSA_SINGLE_DEVICE, which
does the equivalent of LIBSA_SINGLE_FILESYSTEM but for the device
switch table. Device entry pointes are expected to be named
<dev>foo, e.g. the 'strategy' routine used when LIBSA_SINGLE_DEVICE
is set to 'disk' is diskstrategy.
make ufs.c f_nindir array be unsigned ints. the fact that it was signed
caused ufs.c to require signed division routines (which were otherwise
unnecessary for a small boot block).
a small implementation of memcpy(). libsa memcpy() wouldn't
do the right thing if LIBSA_USE_MEMCPY was defined, and the whole
point of that define is to get rid of either bcopy() or memcpy().
(cloned from the bcopy() code.)
NMBCLUSTERS for the mbuf cluster pool. On platforms which use direct-mapped
segments for pool pages (MIPS and Alpha), this makes NMBCLUSTERS actually
meaningful (such ports don't even allocate mb_map, as it is not used to
map mbuf cluster pages).
Improve the message logged at a maximum rate of once per second. The
new message: "WARNING: mclpool limit reached; increase NMBCLUSTERS".
In the back-end pool page allocator, remove the message about mb_map
being full. The message was not necessarily correct as the allocator
may have been starved for pages, rather than for space in the map. Also,
the hard limit on the mbuf cluster pool will be reached before the map
fills (the last cluster will always fit into the map), so the message
is redundant.
Add a comment in mbinit() about considering setting low water marks on
the mbuf and mbuf cluster pools.
- Add support for hard limits, with optional rate-limited logging of
a warning message when the pool limit is reached. (This will be used
to fix a bug in mbuf cluster allocation on the MIPS and Alpha ports.)
- Fix some locking protocol errors. This required splitting pr_flags
into pr_flags (which is protected by the spin lock) and pr_roflags (which
are `read only' flags, set when the pool is initialized, and never changed
again; these do not need to be protected by a mutex).
- Make the low water support actually mean something. When a low water
mark is set, add free items to the pool until the low water mark is
reached. When an item allocation causes the number of free items to
drop below the low water mark, make the pool catch up to it. This can
make the pool allocator more useful for several applications (e.g.
pmap `pv entry' management) and more robust for others (for e.g. mbuf
and mbuf cluster allocation, so that the pagedaemon can use NFS to clean
pages on diskless systems without completely running dry on buffers to
receive packets in during extreme memory shoratages).
- Add a comment where we sleep waiting for more pages for the back-end
page allocator. Specifically, instead of sleeping potentially forever,
perhaps we should just wake up once a second to try allocating a page
again. XXX Revisit this soon.
-read retries were botched, use the right sector count
-read-ahead buffer was effectively unused
-concentrate the handling of the weird BIOS geometry report at one place
-fallback for old floppies left cylinder count uninitialized
- ALIAS() is not needed, use XLEAF() or XNESTED() instead
- use AENT() instead of .aent
- _END_LABEL() is not needed (and was wrong)
- define ALEAF(), NLEAF(), NON_LEAF(), NNON_LEAF() by
XLEAF(), LEAF_NONPROFILE(), NESTED(), NESTED_NONPROFILE()
* We shouldn't truncate the file.
* We were leaving the vnode locked (unless the truncate happened to fail).
Solaris clients may cause this under some conditions.
Problem reported by chopps, analysis and fix by me.
in turn forces a flush of the vnode, whether or not it is involved in a dirop.
(This can happen during a remove or rmdir, when the directory is shrunk.)
Because of the nature of dirops, however, flushing a vnode involved in a dirop
is disallowed (and was marked with a panic). This patch has lfs_truncate
call a specialized vinvalbuf that only invalidates buffers following the new
end-of-file, and thus does not require a flush. Also the panic is demoted,
in case I missed any other path to lfs_vflush.
DELAY(1)'s. This should fix interrupt driven lpt driver hang and
reboot problems for the group of users who have experienced them, and
shouldn't hurt anyone else.
tick and the hardware mysteriously responds fast enough that the delay ends
up being 1 tick short. An unlikely event, but just in case anything actually
relies on this...
* The fact that IIR_NOPEND was not set on entry does *not* mean that no
transmission was in progress. Besides, we don't want to throw away receive
interrupts either.
* In the !clearirq case, we didn't splx().
1. don't clear the irq unless it was clear before transmitting
2. also do various bus_space_barrier() ops
Stops console from freezing when kprintf interrupts tty driver output.
IDENTIFY said so: it doesn't help for the drive this was supposed helping,
and seems to break another device.
In interrupt routine, don't return 0 if we are polling: this should fix the
"panic: wdc_exec_command: polled command not done" some people reported
(kern/7269).
* Count page table pages in the RSS.
* Rather than patching it up, panic if access_type has bits not in prot, as
this should now be impossible.
* Add page table reference counting, but currently disabled as it still has
some issues.
the bit is set that declares the geometry valid. The spec itself says
that this field isn't covered by the "geometry valid" bit, but at
least one BIOS implements it that way.
PPP_DEFLATE, PPP_FILTER and PFIL_HOOKS
Partial fix for kern/7264 from Greg Woods. Too much variance in all
ports GENERIC config files to do across the board right now.
EXEC_ELF64 is used for NetBSD/sparc64, and EXEC_ELF32 is used for NetBSD/sparc_elf.
We really need a way to turn these on and off depending on whether we're building
a 32-bit or 64-bit kernel...
When we put a page on the collection list, we must subtract NPVPPG from the
total free count: one for each pv_entry that's free in that page, and one for
each free pv_entry in other pages that we're going to eat by moving the ones
in the page being collected.
pmap_find_pv(), pmap_clean_page() and pmap_remove_all() are only called on
managed pages, after VM initialization. Panic if this invariant is violated.
Also, panic if we try to enter a PT page through pmap_enter(), rather than
silently patching it up.
pmap_initialized is now #ifdef DIAGNOSTIC.
and dec_3maxplus.c. The ERRSYN/CHKSYN register contains data, not an address.
Pass the address of the register rather than the contents to dec_mtasic_err()
instead of the register contents so it can read/write the register.
Correctable memory errors won't trap in dec_mtasic_err() anymore.
and netinet, currently only tested under netinet.
Disabled by default, enabled by compiling the kernel with option
IFA_STATS. Enabling this feature seems to make the ip_output function
take 13% longer than before, which should be OK for people that need
this feature.
* Map the message buffer with access_type = VM_PROT_READ|VM_PROT_WRITE `just
because'.
* Map the file system buffers with access_type = VM_PROT_READ|VM_PROT_WRITE to
avoid possible problems with pagemove().
* Do not use VM_PROT_EXEC with either of the above.
* Map pages for /dev/mem with access_type = prot. Also, DO NOT use
pmap_kenter() for this, as we DO NOT want to lose modification information.
* Map pages in dumpsys() with VM_PROT_READ.
* Map pages in m68k mappedcopyin()/mappedcopyout() and writeback() with
access_type = prot.
* For now, bus_dma*(), pmap_map(), vmapbuf(), and similar functions still use
access_type = 0. This should probably be revisited.
siop2.c. Add wide negotiation and Ultra support. Modify siop.c to match
the siop2.c sync negotiation changes. The CyberStorm MKIII driver now
supports 15 targets. Remove some old table-driven sync rate stuff from
the original Zeus driver.
emulation of managed pages. This required the following `interesting' changes:
* File system buffers must be entered with an access type of
VM_PROT_READ|VM_PROT_WRITE, so that the pages will be accessible immediately.
Otherwise we would have to teach pagemove() to update the R/M information.
Since they're never eligible for paging, the latter is overkill.
* We must insure that pages allocated before the pmap is completely set up
(that is, pages allocated early by the VM system) are not eligible for R/M
emulation, since the memory needed for this isn't available. We do this by
allocating the pmap's internal memory with uvm_pageboot_alloc(). This also
fixes an absolutely horrible hack where the pmap only worked because page 0
happened to be mapped.
to be mapped.
Also:
* Push the wired page counting into the p->v list maintenance functions. This
avoids code duplication, and fixes some cases where we were confused about
which pages to do it with.
* Fix lots of problems associated with pmap_nightmare() (and rename it to
pmap_vac_me_harder()).
* Since the early pages are no longer considered `managed', just make
pmap_*_pv() panic if !pmap_initialized.
memory access a mapping was caused by. This is passed through from uvm_fault()
and udv_fault(), and in most other cases is 0.
The pmap module may use this to preset R/M information. On MMUs which require
R/M emulation, the implementation may preset the bits and avoid taking another
fault. On MMUs which keep R/M information in hardware, the implementation may
preset its cached bits to speed up the next call to pmap_is_modified() or
pmap_is_referenced().
numerous pagedaemon improvements were needed to make this useful:
- don't bother waking up procs waiting for memory if there's none to be had.
- start 4 times as many pageouts as we need free pages.
this should reduce latency in low-memory situations.
- in inactive scanning, if we find dirty swap-backed pages when swap space
is full of non-resident pages, reactivate some number of these to flush
less active pages to the inactive queue so we can consider paging them out.
this replaces the previous scheme of inactivating pages beyond the
inactive target when we failed to free anything during inactive scanning.
- during both active and inactive scanning, free any swap resources from
dirty swap-backed pages if swap space is full. this allows other pages
be paged out into that swap space.
