Currently si_usecount is effectively a sum of usecounts from all associated
vnodes. This is maintained by special-casing for VCHR every time usecount is
modified. Apart from complicating the code a little bit, it has a scalability
impact since it forces a read from a cacheline shared with said count.
There are no consumers of the feature in the ports tree. In head there are only
2: revoke and devfs_close. Both can get away with a weaker requirement than the
exact usecount, namely just the count of active vnodes. Changing the meaning to
the latter means we only need to modify it on 0<->1 transitions, avoiding the
check plenty of times (and entirely in something like vrefact).
Reviewed by: kib, jeff
Tested by: pho
Differential Revision: https://reviews.freebsd.org/D22202
Typical reasons for doublefault faults are either kernel stack
overflow or bugs in the code that manipulates protection CPU state.
The later code is the code which often has to set up gsbase for
kernel. Switching to explicit load of GSBASE MSR in the fault handler
makes it more probable to output a useful information.
Now all IST handlers have nmi_pcpu structure on top of their stacks.
It would be even more useful to save gsbase value at the moment of the
fault. I did not this because I do not want to modify PCB layout now.
Tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Somehow I managed to commit the manual page with unresolved conflicts in
it.
While here, I also replaced .sp with .Pp.
MFC after: 3 weeks
X-MFC with: r354804
It's unclear how this didn't get caught in my last iteration, but the fix is
easy- the interface is still compatible, it was just gratuituously renamed
to match my arbitrary definition of consistency... VCBUS, the BCM2835 name,
represents an address on the VideoCore CPU Bus.
In a similar fashion, while it is a physical address, the ARMC portion
represents that these are addresses as seen by the ARM CPU.
To make things even more fun, the BCM2711 peripheral documentation describes
not virtual address space vs. physical address space, but instead the 32-bit
address map vs. the address map in "Low Peripheral" mode. The latter of
these is what the *ARMC* macros translate to/from.
We could maintain the static conversions for the !AArch64 Raspberry Pis, but
I'm not sure it's worth it -- we'll traverse the platform list exactly once
(of which there are only two for armv7), then every conversion there-after
traverses the memory map listing of which there are at-most two entries for
these boards: sdram and peripheral space.
Detecting this at runtime is necessary for the AArch64 SOC, though, because
of the distinct IO windows being otherwise not discernible just from support
compiled into the kernel. We currently select the correct window based on
/compatible in the FDT.
We also use a similar mechanism to describe the DMA restrictions- the RPi 4
can have up to 4GB of RAM while the DMA controller and mailbox mechanism can
technically, kind of, only access the lowest 1GB. See the comment in
bcm2835_vcbus.h for a fun description/clarification of this.
Differential Revision: https://reviews.freebsd.org/D22301
so that we avoid the hashtables. The hashtable is now only required if
a zone is created with OFFPAGE specified initially, not internally. This
flag signals to UMA that it can't touch the allocated memory and so
can't store a slab pointer in the containing page.
Reviewed by: markj
Differential Revision: https://reviews.freebsd.org/D22453
reudundant complicated checks and additional locking required only for
anonymous memory. Introduce vm_object_allocate_anon() to create these
objects. DEFAULT and SWAP objects now have the correct settings for
non-anonymous consumers and so individual consumers need not modify the
default flags to create super-pages and avoid ONEMAPPING/NOSPLIT.
Reviewed by: alc, dougm, kib, markj
Tested by: pho
Differential Revision: https://reviews.freebsd.org/D22119
This round of refactoring is mostly about streamlining the interrupt handler
to make it easier to verify and reason about operations taking place while
trying to bring FreeBSD up on the RPi4.
Move the include for sysctl.h out of the middle of the file to the
includes at the beginning. This is will make it easier to add new
sysctls.
No functional changes.
MFC after: 3 weeks
Sponsored by: Netflix
This patch fixes a race condition where the receive callback is called
while the device is being reset. Since the rx_merge variable may change
during reset, the receive callback may operate inconsistently with what
the guest expects.
Also, get rid of the unused rx_vhdrlen variable.
PR: 242023
Reported by: aleksandr.fedorov@itglobal.com
Reviewed by: markj, jhb
MFC with: r354552
Differential Revision: https://reviews.freebsd.org/D22440
Move the SYSCTL_DECL to the top of the file. Move the sysctl function
before SYSCTL_PROC so that we don't need an extra function declaration in
the middle of the file.
No functional changes.
MFC after: 3 weeks
Sponsored by: Netflix
Resort functions within file in a way that they depend on each other as
that makes it easier to rework various things.
Also allows us to remove file local function declarations.
No functional changes.
MFC after: 3 weeks
Sponsored by: Netflix
invalidation in reclaim_pv_chunk(). (2) Prevent an "invalid ASID" assertion
failure in reclaim_pv_chunk(). The detailed explanation for this change is
provided by r354792.
X-MFC with: r354792
This is due to LLD_REVISION_STRING being renamed to LLD_REVISION in
r351442 and the value being moved to another location in r351965.
`make test-system-linker` can be used to see the values being used here.
Reported by: ler
With VIMAGE kernels modules get special treatment as they need
to also keep the original values and make copies for each instance.
For that a few pages of vnet modspace are provided and the
kernel-linker and the VNET framework know how to deal with things.
When the modspace is (almost) full, other modules which would
overflow the modspace cannot be loaded and kldload will fail.
ip_mroute uses a lot of variable space, mostly be four big arrays:
set_vnet 0000000000000510 vnet_entry_multicast_register_if
set_vnet 0000000000000700 vnet_entry_viftable
set_vnet 0000000000002000 vnet_entry_bw_meter_timers
set_vnet 0000000000002800 vnet_entry_bw_upcalls
Dynamically malloc the three big ones for each instance we need
and free them again on vnet teardown (the 4th is an ifnet).
That way they only need module space for a single pointer and
allow a lot more modules using virtualized variables to be loaded
on a VNET kernel.
PR: 206583
Reviewed by: hselasky, kp
MFC after: 3 weeks
Differential Revision: https://reviews.freebsd.org/D22443
in6ifa_ifpforlinklocal() asserts the net epoch. The test case from r354832
revealed code paths where we call into the function without having
acquired the net epoch first and consequently we hit the assert.
This happens in certain MLD states during VNET shutdown and most people
normaly not notice this.
For correctness acquire the net epoch around calls to
mld_v1_transmit_report() in all cases to avoid the assertion firing.
MFC after: 2 weeks
Sponsored by: Netflix
The assumptions of linker_set don't play nicely with
AddressSanitizer. AddressSanitizer adds a 'redzone' of zeros around
globals (including those in named sections), whereas linker_set
assumes they are all packed consecutively like a pointer array. So:
let's annotate linker_set so that AddressSanitizer ignores it.
Submitted by: Matthew Bryan <matthew.bryan@isilon.com>
Reviewed by: kib, rang_acm.org
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D22239
only. Rename it swp_pager_meta_lookup. Stop checking for obj->type
== swap there and assert it instead. Make the caller responsible for
the obj->type check.
Move the meta_ctl 'pop' functionality to swap_pager_unswapped, the
only place that uses it, and assume obj->type == swap there too.
Assisted by: ota_j.email.ne.jp
Reviewed by: kib
Tested by: pho
Differential Revision: https://reviews.freebsd.org/D22437
A SIM-private field is used for that.
The pointer can be useful when examining a state of a queued ccb.
E.g., a ccb on a da_softc.pending_ccbs.
MFC after: 2 weeks
Use RB_TREE to make some algorithms O(lg N) and O(N lg N) instead of O(N)
and O(N^2). Because N is typically small and the former linear array also has
great constant factors (as a property of CPU caching), this doesn't provide
material benefit most or all of the time.
While here, remove arbitrarily limit on number of macros understood.
This allows easy and care-free scaling of NUM_DMA_SEGS with proper-ish
calculations to make sure we can actually handle the number of segments we'd
like to handle on average so that performance comparisons can be easily made
at different values if/once we can actually handle it. It also makes it
helps the untrained reader understand more quickly the reasoning behind the
choice of maxsize/maxsegs/maxsegsize.
Summary: A follow-on to r276634, which added the VMX note to userland cores,
and r334538 for VSX notes.
Copied from note_type_linux_core().
Reviewed by: emaste
Differential Revision: https://reviews.freebsd.org/D22404
PLX NTB sends translated DMA requests not only from itsels, but from all
slots and functions of its bus. By default DMAR blocks those additional.
MFC after: 1 week
Sponsored by: iXsystems, Inc.
ENOENT is leftover from mmu_oea.c's moea_pvo_enter(), where it's used to
syncicache() on the first new mapping of a page. This sync is done
differently in OEA64.
'tlbilxpid' is 'tlbilx 1, 0', while the existing form is 'tlbilx 0, 0',
which translates to 'tlbilxlpid', invalidating a LDPID. This effectively
invalidates the entire TLB, causing unnecessary reloads.
The lua-based makesyscalls produces slightly different output than its
makesyscalls.sh predecessor, all whitespace differences more closely
matching the source syscalls.master.
flua is bootstrapped as part of the build for those on older
versions/revisions that don't yet have flua installed. Once upgraded past
r354833, "make sysent" will again naturally work as expected.
Reviewed by: brooks
Differential Revision: https://reviews.freebsd.org/D21894
Summary:
The ?: operator has a lower precedence than == and &&, so the result will always
be recorded as true. Found by gcc8.
Reviewed by: ngie, ae
Differential Revision: https://reviews.freebsd.org/D22427
FreeBSDlua ("flua") is a FreeBSD-private lua, flavored with whatever
extensions we need for base system operations. We currently support a subset
of lfs and lposix that are used in the rewrite of makesyscall.sh into lua,
added in r354786.
flua is intentionally written such that one can install standard lua and
some set of lua modules from ports and achieve the same effect.
linit_flua is a copy of linit.c from contrib/lua with lfs and lposix added
in. This is similar to what we do in stand/. linit.c has been renamed to
make it clear that this has flua-specific bits.
luaconf has been slightly obfuscated to make extensions more difficult. Part
of the problem is that flua is already hard enough to use as a bootstrap
tool because it's not in PATH- attempting to do extension loading would
require a special bootstrap version of flua with paths changed to protect
the innocent.
src.lua.mk has been added to make it easy for in-tree stuff to find flua,
whether it's bootstrap-flua or relying on PATH frobbing by Makefile.inc1.
Reviewed by: brooks, emaste (both earlier version), imp
Differential Revision: https://reviews.freebsd.org/D21893
After r354748 mld_input() can change the mbuf. The new pointer
is never returned to icmp6_input() and when passed to
icmp6_rip6_input() the mbuf may no longer valid leading to
a panic.
Pass a pointer to the mbuf to mld_input() so we can return an
updated version in the non-error case.
Add a test sending an MLD packet case which will trigger this bug.
Pointyhat to: bz
Reported by: gallatin, thj
MFC After: 2 weeks
X-MFC with: r354748
Sponsored by: Netflix