- Enable UFS_DIRHASH if the architecture or kernel model specific config
file can use 128MB of RAM or more.
- Remove experimental tag from UFS_DIRHASH; it's been with RUMP kernel
and by a number of NetBSD developers for years.
- Add LFS_DIRHASH if LFS was enabled.
- Be somewhat consistent with FS options order.
size dependant on memory size. If less than 128MB of memory, default
to no cache. With 128MB of memory or more, use a maximum cache size of
1/64th of memory; cap maximum default cache size to 32MB (for systems
with 2GB of memory or more).
The dirhash cache sizes are still explicityly setable by sysctl(8) or
by adding relevant entry(s) to sysctl.conf(5).
gcc generates calls to this symbol in programs that use
__sync_*_compare_and_swap, which require full sequential consistency
barriers, including store-before-load ordering on both sides of the
atomic; none of the release/acquire operations guarantee that, so we
have to insert explicit DMB instructions.
Note: gcc's own definition omits some of the DMB instructions, but I
can't prove that it's correct that way -- stores preceding the CAS
must complete before the load part of the CAS, and the store part of
the CAS must complete before loads following the CAS. Maybe there's
some way to prove that one of these orderings is guaranteed some
other way than a DMB but I'm not seeing it, and store-before-load
ordering is hard to understand.
Patch by skrll@ based on a patch by mrg@, soliloquy in commit message
by me.
In the modifier ':S,from,to,', parsing the two parts 'from' and 'to' of
the modifier differs depending on whether the expression is actually
evaluated or merely parsed. This not only applies to the ':S' modifier,
but also to ':C', ':@var@body@', ':!cmd!', ':[...]', ':?:', '::=' and
':from=to'.
- No membar_producer in sip_init_rxdesc -- use bus_dmamap_sync with
BUS_DMASYNC_PREWRITE to order updates to the DMA descriptors.
- Omit needless membar_producer in sip_init_txdesc -- the hardware
will not look at any of these descriptors until we set CMDSTS_OWN
on the first one in the sequence, which is done later in the
caller, sipcom_start.
- In gsip_rxintr, make sure to read cmdsts _before_ extsts, by
separating them with BUS_DMASYNC_PREREAD. Otherwise, the CPU might
reorder the loads and read a stale extsts first before witnessing
an updated cmdsts with the CMDSTS_OWN bit that transfers ownership
of the rx packet to us.
The table was created by manually inspecting the code of the various
ApplyModifier functions in var.c.
The modifiers are listed in alphabetical order, except for the SysV
modifier, which is listed at the end since it is used as a fallback
modifier for many other modifiers and because it does not have a fixed
prefix.
NKMEMPAGES_MAX_UNLIMITED. we used to limit kva to 1/8 of physmem
but limiting to 1/4 should be enough, and 1/4 still gives the kernel
enough kva to map all of the RAM that KMSAN has not stolen.
Reported-by: syzbot+ca3710b4c40cdd61aa72@syzkaller.appspotmail.com
kern.entropy.consolidate and kern.entropy.gather are supposed to be
write-only -- it doesn't make any sense to read from them, but I
suppose it's better to read-as-zero than read-as-stack-secrets!
(such as AppleTalk).
Instead, remove/preserve the final four bytes in the packet ourselves on a per-
protocol basis, as we do in arch/arm/xscale/ixp425_if_npe.c (and dev/ic/gem.c,
and so forth).
When parsing the expression ${:D any ${uninterpreted} text}, parsing of
the ':D' modifier only needs to skip to the next delimiter, there is no
need to keep record of the text that has been skipped. In this
particular example, there had been an unnecessary memory allocation
since the nested expression was not copied to the result buffer. Due to
that, the resulting buffer contained " any text", which was not a
substring of the text passed to the parser, thereby forcing the memory
allocation.
No functional change.
vnode state to VS_RECLAIMING, before we actually call VOP_RECLAIM(),
which will release the reference on the lower node of a stacked FS
vnode, which is likely to free the upper node's v_klist backing store.
Acquire the vnode interlock when checking for kevent interest now,
because the vp->v_klist pointer is now volatile.
PR kern/56950
The type-limits warning is actively harmful because it discourages
writing safe portable overflow detection logic which happens, on some
architectures, to be dead code.
Read the interface up/down status from sc_if_flags (under WM_CORE_LOCK)
when deciding if the multicast filter needs to be updated.
Discussed with msaitoh@, knakahara@ and riastradh@
high enough to allow the kernel to map all of RAM into kmem,
so that free physical pages rather than kernel virtual space is
the limiting factor in allocating kernel memory. this gives ZFS
more flexibility in tuning how much memory to use for its ARC cache.
This can happen due to janky MD kludgerosity like x86
x86_genfb_ddb_trap_callback, which should really be cleaned up, but
at least this might help with the recursive traps we've been seeing
in syzbot.
skrll