eugen [Mon, 26 Nov 2018 11:22:04 +0000 (11:22 +0000)]
MFC r339816: mount_msdosfs
mount_msdosfs: do not fail mounts requiring locale name conversion table
that is already present in a kernel statically.
For example, the command "mount_msdosfs -L ru_RU.KOI8-R" fails with error
"mount_msdosfs: msdosfs_iconv: File exists" for a kernel having
options LIBICONV and MSDOSFS_ICONV. After this change, it mounts
successfully.
emaste [Sun, 25 Nov 2018 00:34:00 +0000 (00:34 +0000)]
MFC r340771: proto: change device permissions to 0600
C Turt reports that the driver is not thread safe and may have
exploitable races.
Note that the proto device is intended for prototyping and development,
and is not for use on production systems. From the man page:
SECURITY CONSIDERATIONS
Because programs have direct access to the hardware, the proto
driver is inherently insecure. It is not advisable to use this
driver on a production machine.
The proto device is not included in any of FreeBSD's kernel config files
(although the module is built).
The issues in the proto device still need to be fixed, and the device is
inherently (and intentionally) insecure, but it might as well be limited
to root only.
admbugs: 782
Reported by: C Turt <ecturt@gmail.com>
Sponsored by: The FreeBSD Foundation
emaste [Fri, 23 Nov 2018 20:41:54 +0000 (20:41 +0000)]
MFC r340663 (rmacklem):
Improve sanity checking for the dircount hint argument to
NFSv3's ReaddirPlus and NFSv4's Readdir operations. The code
checked for a zero argument, but did not check for a very large value.
This patch clips dircount at the server's maximum data size.
emaste [Fri, 23 Nov 2018 20:39:37 +0000 (20:39 +0000)]
MFC r340662 (rmacklem):
nfsm_advance() would panic() when the offs argument was negative.
The code assumed that this would indicate a corrupted mbuf chain, but
it could simply be caused by bogus RPC message data.
This patch replaces the panic() with a printf() plus error return.
emaste [Fri, 23 Nov 2018 20:38:50 +0000 (20:38 +0000)]
MFC r340661 (rmacklem):
r304026 added code that started statistics gathering for an operation
before the operation number (the variable called "op") was sanity checked.
This patch moves the code down to below the range sanity check for "op".
tijl [Thu, 22 Nov 2018 09:47:51 +0000 (09:47 +0000)]
MFC r340674:
Fix another user address dereference in linux_sendmsg syscall.
This was hidden behind the LINUX_CMSG_NXTHDR macro which dereferences its
second argument. Stop using the macro as well as LINUX_CMSG_FIRSTHDR. Use
the size field of the kernel copy of the control message header to obtain
the next control message.
tijl [Thu, 22 Nov 2018 09:41:54 +0000 (09:41 +0000)]
MFC r340631:
Do proper copyin of control message data in the Linux sendmsg syscall.
Instead of calling m_append with a user address, allocate an mbuf cluster
and copy data into it using copyin. For the SCM_CREDS case, instead of
zeroing a stack variable and appending that to the mbuf, zero part of the
mbuf cluster directly. One mbuf cluster is also the size limit used by
the FreeBSD sendmsg syscall (uipc_syscalls.c:sockargs()).
marius [Wed, 21 Nov 2018 18:53:30 +0000 (18:53 +0000)]
MFC: r340495
- Restore setting the clock for devices which support the default/legacy
transfer mode only (lost with r321385). [1]
- Similarly, don't try to set the power class on MMC devices that comply
to version 4.0 of the system specification but are operated in default/
legacy transfer or 1-bit bus mode as no power class is specified for
these cases. Trying to set a power class nevertheless resulted in an -
albeit harmless - error message.
eugen [Tue, 20 Nov 2018 10:44:49 +0000 (10:44 +0000)]
MFC r339558: New sysctl: net.inet.icmp.error_keeptags
Currently, icmp_error() function copies FIB number from original packet
into generated ICMP response but not mbuf_tags(9) chain.
This prevents us from easily matching ICMP responses corresponding
to tagged original packets by means of packet filter such as ipfw(8).
For example, ICMP "time-exceeded in-transit" packets usually generated
in response to traceroute probes lose tags attached to original packets.
This change adds new sysctl net.inet.icmp.error_keeptags
that defaults to 0 to avoid extra overhead when this feature not needed.
Set net.inet.icmp.error_keeptags=1 to make icmp_error() copy mbuf_tags
from original packet to generated ICMP response.
kevans [Mon, 19 Nov 2018 19:05:07 +0000 (19:05 +0000)]
MFC r340392: Add dynamic_kenv assertion to init_static_kenv
Both to formally document the requirement that this not be called after the
dynamic kenv is setup, and to perhaps help static analyzers figure out
what's going on. While calling init_static_kenv this late isn't fatal, there
are some caveats that the caller should be aware of:
- Late calls are effectively a no-op, as far as default FreeBSD is
concerned, as everything will switch to searching the dynamic kenv once it's
available.
- Each of the kern_getenv calls will leak memory, as it's assumed that
these are searching static environment and allocations will not be made.
As such, this usage is not sensible and should be detected.
eugen [Mon, 19 Nov 2018 06:37:38 +0000 (06:37 +0000)]
MFC r339465: rc.initdiskless: add support for auxiliary NVRAM.
Currently, rc.inidiskless assumes that local system configuration
changes are kept in some mountable file system. For example,
nanobsd uses dedicated partition mounted as /cfg for this.
However, small embedded devices like MIPS routers may have no enough flash
space to keep full-blown file system but have only one or couple
small flash blocks to keep persistent local configuration overrides.
This change extends rc.initdiskless and introduces ability to run auxiliary
command /conf/T/M/extract that is supposed to extract configuration overrides
from such local storage.
For example, the command /conf/default/etc/extract may contain something like:
cd "$1" && bsdcpio --quiet -idu < /dev/map/cfg
bsdcpio command extracts compressed archive from the storage to /etc
assuming the storage is exposed by the kernel as /dev/map/cfg to userland.
rmacklem [Sun, 18 Nov 2018 22:59:54 +0000 (22:59 +0000)]
MFC: r339999
Fix NFS client vnode locking to avoid a crash during forced dismount.
A crash was reported where the crash occurred in nfs_advlock() when the
NFS_ISV4(vp) macro was being executed. This was caused by the vnode
being VI_DOOMED due to a forced dismount in progress.
This patch fixes the problem by locking the vnode before executing the
NFS_ISV4() macro.
emaste [Sun, 18 Nov 2018 14:53:29 +0000 (14:53 +0000)]
MFC r340329: build(7): clarify buildenv target can be used for non-cross builds
make buildenv can be used for building for the same architecture as
the host (perhaps this is a degenerate case of cross-building).
TARGET and TARGET_ARCH do not need to be set in this case.
kp [Sun, 18 Nov 2018 12:09:27 +0000 (12:09 +0000)]
MFC r340067:
pfsync: Ensure uninit is done before pf
pfsync touches pf memory (for pf_state and the pfsync callback
pointers), not the other way around. We need to ensure that pfsync is
torn down before pf.
kp [Sun, 18 Nov 2018 12:04:25 +0000 (12:04 +0000)]
MFC r340066:
Notify that the ifnet will go away, even on vnet shutdown
pf subscribes to ifnet_departure_event events, so it can clean up the
ifg_pf_kif and if_pf_kif pointers in the ifnet.
During vnet shutdown interfaces could go away without sending the event,
so pf ends up cleaning these up as part of its shutdown sequence, which
happens after the ifnet has already been freed.
Send the ifnet_departure_event during vnet shutdown, allowing pf to
clean up correctly.
kp [Sun, 18 Nov 2018 10:57:39 +0000 (10:57 +0000)]
MFC r339676:
pf: Fix copy/paste error in IPv6 address rewriting
We checked the destination address, but replaced the source address. This was
fixed in OpenBSD as part of their NAT rework, which we don't want to import
right now.
kp [Sun, 18 Nov 2018 10:47:50 +0000 (10:47 +0000)]
MFC r339470:
pf synproxy will do the 3WHS on behalf of the target machine, and once
the 3WHS is completed, establish the backend connection. The trigger
for "3WHS completed" is the reception of the first ACK. However, we
should not proceed if that ACK also has RST or FIN set.
jhb [Sun, 18 Nov 2018 01:07:36 +0000 (01:07 +0000)]
MFC 339312,339364: Restore more descriptors during VM exits.
339312:
Fully restore the GDTR, IDTR, and LDTR after VT-x VM exits.
The VT-x VMCS only stores the base address of the GDTR and IDTR. As a
result, VM exits use a fixed limit of 0xffff for the host GDTR and
IDTR losing the smaller limits set in when the initial GDT is loaded
on each CPU during boot. Explicitly save and restore the full GDTR
and IDTR contents around VM entries and exits to restore the correct
limit.
Similarly, explicitly save and restore the LDT selector. VM exits
always clear the host LDTR as if the LDT was loaded with a NULL
selector and a userspace hypervisor is probably using a NULL selector
anyway, but save and restore the LDT explicitly just to be safe.
339364:
Reload the LDT selector after an AMD-v #VMEXIT.
cpu_switch() always reloads the LDT, so this can only affect the
hypervisor process itself. Fix this by explicitly reloading the host
LDT selector after each #VMEXIT. The stock bhyve process on FreeBSD
never uses a custom LDT, so this change is cosmetic.
ae [Sun, 18 Nov 2018 00:34:24 +0000 (00:34 +0000)]
MFC r339542:
Retire IPFIREWALL_NAT64_DIRECT_OUTPUT kernel option. And add ability
to switch the output method in run-time. Also document some sysctl
variables that can by changed for NAT64 module.
NAT64 had compile time option IPFIREWALL_NAT64_DIRECT_OUTPUT to use
if_output directly from nat64 module. By default is used netisr based
output method. Now both methods can be used, but they require different
handling by rules.
ae [Sun, 18 Nov 2018 00:31:09 +0000 (00:31 +0000)]
MFC r339533:
Add sadb_x_sa2 extension to SADB_ACQUIRE requests.
SADB_ACQUIRE requests are send by kernel, when security policy doesn't
have corresponding security association for outbound packet. IKE daemon
usually registers its handler for such messages and when the kernel asks
for SA it can handle this request. Now such requests will contain
additional fields that can help IKE daemon to create SA. And IKE now
can create SAs using only information from SADB_ACQUIRE request, this
is useful when many if_ipsec(4) interfaces are in use and IKE doesn track
security policies that was installed by kernel.
ae [Sun, 18 Nov 2018 00:28:56 +0000 (00:28 +0000)]
MFC r339539:
Add IPFW_RULE_JUSTOPTS flag, that is used by ipfw(8) to mark rule,
that was added using "new rule format". And then, when the kernel
returns rule with this flag, ipfw(8) can correctly show it.
Reported by: lev
Sponsored by: Yandex LLC
Differential Revision: https://reviews.freebsd.org/D17373
ae [Sun, 18 Nov 2018 00:26:09 +0000 (00:26 +0000)]
MFC r339535:
Do not allow use `create` keyword as hostname when ifconfig(8) is invoked
for already existing interface.
It appeared, that ifconfig(8) assumes `create` keyword as hostname and
tries to resolve it, when `ifconfig ifname create` invoked for already
existing interface. This can produce some unexpected results, when hostname
resolving has successfully happened. This patch adds check for such case.
When an interface is already exists, and create is only one argument,
return error message. But when there are some other arguments, just remove
create keyword from the arguments list.
jhb [Sun, 18 Nov 2018 00:11:19 +0000 (00:11 +0000)]
MFC 338511: bhyve: Use MAP_GUARD when mapping guest memory ranges.
Instead of relying on PROT_NONE mappings with MAP_ANON, use MAP_GUARD
to reserve address space around guest memory ranges including the
guard ranges of address space around mappings.
ygy [Thu, 15 Nov 2018 08:43:17 +0000 (08:43 +0000)]
MFC r338977:
Add description, parameters, options, sysctl and examples of using AQMs to ipfw man page. CoDel, PIE, FQ-CoDel and FQ-PIE AQM for Dummynet exist in FreeBSD 11 and 10.3.
scottl [Tue, 13 Nov 2018 18:49:43 +0000 (18:49 +0000)]
Fix a regression from prior to 11.2 that caused MSI (not MSI-X) interrupt
allocation to fail. While here, refactor the code so that it's more clear
and less likely to break in the future. This is not an MFC due to the code
in 12/head being very different, but it follows the latter's structure
more closely than before.
hselasky [Sat, 10 Nov 2018 10:31:35 +0000 (10:31 +0000)]
MFC r340212:
Sometimes the complete split packet may be queued too early and the
transaction translator will return a NAK. Ignore this message and
retry the complete split instead.
emaste [Fri, 9 Nov 2018 21:45:42 +0000 (21:45 +0000)]
Fix objcopy for little-endian MIPS64 objects.
MFC r338478 (jhb): Fix objcopy for little-endian MIPS64 objects.
MIPS64 does not store the 'r_info' field of a relocation table entry as
a 64-bit value consisting of a 32-bit symbol index in the high 32 bits
and a 32-bit type in the low 32 bits as on other architectures. Instead,
the 64-bit 'r_info' field is really a 32-bit symbol index followed by four
individual byte type fields. For big-endian MIPS64, treating this as a
64-bit integer happens to be compatible with the layout expected by other
architectures (symbol index in upper 32-bits of resulting "native" 64-bit
integer). However, for little-endian MIPS64 the parsed 64-bit integer
contains the symbol index in the low 32 bits and the 4 individual byte
type fields in the upper 32-bits (but as if the upper 32-bits were
byte-swapped).
To cope, add two helper routines in gelf_getrel.c to translate between the
correct native 'r_info' value and the value obtained after the normal
byte-swap translation. Use these routines in gelf_getrel(), gelf_getrela(),
gelf_update_rel(), and gelf_update_rela(). This fixes 'readelf -r' on
little-endian MIPS64 objects which was previously decoding incorrect
relocations as well as 'objcopy: invalid symbox index' warnings from
objcopy when extracting debug symbols from kernel modules.
Even with this fixed, objcopy was still crashing when trying to extract
debug symbols from little-endian MIPS64 modules. The workaround in
gelf_*rel*() depends on the current ELF object having a valid ELF header
so that the 'e_machine' field can be compared against EM_MIPS. objcopy
was parsing the relocation entries to possibly rewrite the 'r_info' fields
in the update_relocs() function before writing the initial ELF header to
the destination object file. Move the initial write of the ELF header
earlier before copy_contents() so that update_relocs() uses the correct
symbol index values.
Note that this change should really go upstream. The binutils readelf
source has a similar hack for MIPS64EL though I implemented this version
from scratch using the MIPS64 ABI PDF as a reference.
MFC r339083 (emaste): libelf: correct mips64el test to use ELF header
libelf maintains two views of endianness: e_byteorder, and
e_ident[EI_DATA] in the ELF header itself. e_byteorder is not always
kept in sync, so use the ELF header endianness to test for mips64el.
MFC r339473 (emaste): libelf: also test for 64-bit ELF in _libelf_is_mips64el
Although _libelf_is_mips64el is only called in contexts where we've
already checked that e_class is ELFCLASS64 but this may change in the
future. Add a safety belt so that we don't access an invalid e_ehdr64
union member if it does.
emaste [Fri, 9 Nov 2018 21:38:53 +0000 (21:38 +0000)]
MFC r331078 (cem): nm: Initialize allocated memory before use
In out of memory scenarios (where one of these allocations failed but
other(s) did not), nm(1) could reference the uninitialized value of these
allocations (undefined behavior).
Always initialize any successful allocations as the most expedient
resolution of the issue. However, I would encourage upstream elftoolchain
contributors to clean up the error path to just abort immediately, rather
than proceeding sloppily when one allocation fails.
if present to enable some devices like WaveShare touchscreens. Unlike
Windows we discard content of the blob. We try mimic Windows driver
behaviour from the USB device point of view.
Submitted by: glebius (initial version)
MFC r337289:
wmt(4): Use internal function to calculate input report size
Usbhid's hid_report_size() calculates integral size of all reports of given
kind found in the HID descriptor rather then exact size of report with given
ID as its userland counterpart does. As all input data processed by the
driver is located within the same report, calculate required driver's buffer
size with userland version, imported in one of the previous commits.
This allows us to skip zeroing of buffer on processing of each report.
While here do some minor refactoring.
MFC r338458:
wmt(4): Fix regression introduced in r337289
r337289 has a side effect of reducing usb frame 0 buffer size down to
touch report size. That broke some devices e.g. "Raydium Touch System"
which are capable of generating non-touch frames of bigger length.
Fix it with enlarging frame 0 buffer up to internal wmt(4) buffer size.
MFC r340100:
Do not use bzero() for the O_ICMP6TYPE opcode.
The buffer is already zeroed in compile_rule() function, and also it
may contain configured F_NOT flag in o.len field. This fixes the
filling for "not icmp6types" opcode.
MFC r340175:
Do not print "ip6" keyword in print_icmp6types() for O_ICMP6TYPE opcode.
It produces incompatibility when rules listing is used again to
restore saved ruleset, because "ip6" keyword produces separate opcode.
The kernel already has the check and only IPv6 packets will be checked
for matching.
jhb [Thu, 8 Nov 2018 22:42:55 +0000 (22:42 +0000)]
MFC 340164,340168,340170: Add custom cpu_lock_delay() for x86.
340164:
Add a KPI for the delay while spinning on a spin lock.
Replace a call to DELAY(1) with a new cpu_lock_delay() KPI. Currently
cpu_lock_delay() is defined to DELAY(1) on all platforms. However,
platforms with a DELAY() implementation that uses spin locks should
implement a custom cpu_lock_delay() doesn't use locks.
340168:
Add a delay_tsc() static function for when DELAY() uses the TSC.
This uses slightly simpler logic than the existing code by using the
full 64-bit counter and thus not having to worry about counter
overflow.
340170:
Add a custom implementation of cpu_lock_delay() for x86.
Avoid using DELAY() since it can try to use spin locks on CPUs without
a P-state invariant TSC. For cpu_lock_delay(), always use the TSC if
it exists (even if it is not P-state invariant) to delay for a
microsecond. If the TSC does not exist, read from I/O port 0x84 to
delay instead.
gjb [Thu, 8 Nov 2018 21:58:51 +0000 (21:58 +0000)]
MFC r340260 (emaste):
Avoid buffer underwrite in icmp_error
icmp_error allocates either an mbuf (with pkthdr) or a cluster depending
on the size of data to be quoted in the ICMP reply, but the calculation
failed to account for the additional padding that m_align may apply.
Include the ip header in the size passed to m_align. On 64-bit archs
this will have the net effect of moving everything 4 bytes later in the
mbuf or cluster. This will result in slightly pessimal alignment for
the ICMP data copy.
Also add an assertion that we do not move m_data before the beginning of
the mbuf or cluster.
Approved by: re (kib, insta-MFC)
Security: CVE-2018-17156
Sponsored by: The FreeBSD Foundation
tijl [Thu, 8 Nov 2018 19:56:45 +0000 (19:56 +0000)]
MFC r340181, r340185:
On amd64 both Linux compat modules, linux.ko and linux64.ko, provide
linux_ioctl_(un)register_handler that allows other driver modules to
register ioctl handlers. The ioctl syscall implementation in each Linux
compat module iterates over the list of handlers and forwards the call to
the appropriate driver. Because the registration functions have the same
name in each module it is not possible for a driver to support both 32 and
64 bit linux compatibility.
Move the list of ioctl handlers to linux_common.ko so it is shared by
both Linux modules and all drivers receive both 32 and 64 bit ioctl calls
with one registration. These ioctl handlers normally forward the call
to the FreeBSD ioctl handler which can handle both 32 and 64 bit.
Keep the special COMPAT_LINUX32 ioctl handlers in linux.ko in a separate
list for now and let the ioctl syscall iterate over that list first.
Later, COMPAT_LINUX32 support can be added to the 64 bit ioctl handlers
via a runtime check for ILP32 like is done for COMPAT_FREEBSD32 and then
this separate list would disappear again. That is a much bigger effort
however and this commit is meant to be MFCable.
This enables linux64 support in x11/nvidia-driver*.
oshogbo [Wed, 7 Nov 2018 20:31:36 +0000 (20:31 +0000)]
MFC r335844:
core(5): overwrite the oldest core dump
The '%I' format in the kern.corefile sysctl limits the number of
core files that a process can generate to the number stored in the
debug.ncores sysctl. The '%I' format is replaced by the single digit
index. Previously, if all indexes were taken the kernel would overwrite
only a core file with the highest index in a filename.
Currently the system will create a new core file if there is a free
index or if all slots are taken it will overwrite the oldest one.
emaste [Wed, 7 Nov 2018 18:12:18 +0000 (18:12 +0000)]
MFC r340137: rtld: move relro enforcement after ifunc processing
Previously the combination of relro (implicit), -z now and ifunc use
resulted in a segfault when applying ifuncs after relro (test binary
here just calls amd64_get_fsbase()):
As a follow-up to r339930 and various reports implement logging in case
we fail during module load because the pcpu or vnet module sections are
full. We did return a proper error but not leaving any indication to
the user as to what the actual problem was.
In r78161 the lookup_set linker method was introduced which optionally
returns the section start and stop locations as well as a count if the
caller asks for them.
There was only one out-of-file consumer of count which did not actually
use it and hence was eliminated in r339407.
In r194784 parse_dpcpu(), and in r195699 parse_vnet() (a copy of the
former) started to use the link_elf_lookup_set() interface internally
also asking for the count.
count is computed as the difference of the void **stop - void **start
locations and as such, if the absoulte numbers
(stop - start) % sizeof(void *) != 0
a round-down happens, e.g., **stop 0x1003 - **start 0x1000 => count 0.
To get the section size instead of "count is the number of pointer
elements in the section", the parse_*() functions do a
count *= sizeof(void *).
They use the result to allocate memory and copy the section data
into the "master" and per-instance memory regions with a size of
count.
As a result of count possibly round-down this can miss the last
bytes of the section. The good news is that we do not touch
out of bounds memory during these operations (we may at a later stage
if the last bytes would overflow the master sections).
Given relocation in elf_relocaddr() works based on the absolute
numbers of start and stop, this means that we can possibly try to
access relocated data which was never copied and hence we get
random garbage or at best zeroed memory.
Stop the two (last) consumers of count (the parse_*() functions)
from using count as well, and calculate the section size based on
the absolute numbers of stop and start and use the proper size for
the memory allocation and data copies. This will make the symbols
in the last bytes of the pcpu or vnet sections be presented as
expected.
The countp argument passed to linker_file_lookup_set() in
linker_load_dependencies() is unused, so no need to ask for the
value in first place. Remove the unused "count" variable.
With more excessive use of modules, more kernel parts working with
VIMAGE, and feature richness and global state increasing the 8k of
vnet module space are no longer sufficient for people and loading
multiple modules, e.g., pf(4) and ipl(4) or ipsec(4) will fail on
the second module.
Increase the module space to 8 * PAGE_SIZE which should be enough
to hold multiple firewalls, ipsec, multicast (as in the old days was
a problem), epair, carp, and any kind of other vnet enabled modules.
Sadly this is a global byte array part of the vnet_set, so we cannot
dynamically change its size; otherwise a TUNABLE would have been
a better solution.
jhb [Thu, 1 Nov 2018 20:42:54 +0000 (20:42 +0000)]
MFC 338813: Clear all of the VFP state in fill_fpregs().
Zero the entire FP register set structure returned for ptrace() if a
thread hasn't used FP registers rather than leaking garbage in the
fp_sr and fp_cr fields.
jhb [Thu, 1 Nov 2018 18:34:26 +0000 (18:34 +0000)]
MFC 338360,338415,338624,338630,338631,338725: Dynamic x86 IRQ layout.
338360:
Dynamically allocate IRQ ranges on x86.
Previously, x86 used static ranges of IRQ values for different types
of I/O interrupts. Interrupt pins on I/O APICs and 8259A PICs used
IRQ values from 0 to 254. MSI interrupts used a compile-time-defined
range starting at 256, and Xen event channels used a
compile-time-defined range after MSI. Some recent systems have more
than 255 I/O APIC interrupt pins which resulted in those IRQ values
overflowing into the MSI range triggering an assertion failure.
Replace statically assigned ranges with dynamic ranges. Do a single
pass computing the sizes of the IRQ ranges (PICs, MSI, Xen) to
determine the total number of IRQs required. Allocate the interrupt
source and interrupt count arrays dynamically once this pass has
completed. To minimize runtime complexity these arrays are only sized
once during bootup. The PIC range is determined by the PICs present
in the system. The MSI and Xen ranges continue to use a fixed size,
though this does make it possible to turn the MSI range size into a
tunable in the future.
As a result, various places are updated to use dynamic limits instead
of constants. In addition, the vmstat(8) utility has been taught to
understand that some kernels may treat 'intrcnt' and 'intrnames' as
pointers rather than arrays when extracting interrupt stats from a
crashdump. This is determined by the presence (vs absence) of a
global 'nintrcnt' symbol.
This change reverts r189404 which worked around a buggy BIOS which
enumerated an I/O APIC twice (using the same memory mapped address for
both entries but using an IRQ base of 256 for one entry and a valid
IRQ base for the second entry). Making the "base" of MSI IRQ values
dynamic avoids the panic that r189404 worked around, and there may now
be valid I/O APICs with an IRQ base above 256 which this workaround
would incorrectly skip.
If in the future the issue reported in PR 130483 reoccurs, we will
have to add a pass over the I/O APIC entries in the MADT to detect
duplicates using the memory mapped address and use some strategy to
choose the "correct" one.
While here, reserve room in intrcnts for the Hyper-V counters.
338415:
Fix build of x86 UP kernels after dynamic IRQ changes in r338360.
338624:
msi: remove the check that interrupt sources have been added
When running as a specific type of Xen guest the hypervisor won't
provide any emulated IO-APICs or legacy PICs at all, thus hitting the
following assert in the MSI code:
panic: Assertion num_io_irqs > 0 failed at /usr/src/sys/x86/x86/msi.c:334
cpuid = 0
time = 1
KDB: stack backtrace:
db_trace_self_wrapper() at db_trace_self_wrapper+0x2b/frame 0xffffffff826ffa70
vpanic() at vpanic+0x1a3/frame 0xffffffff826ffad0
panic() at panic+0x43/frame 0xffffffff826ffb30
msi_init() at msi_init+0xed/frame 0xffffffff826ffb40
apic_setup_io() at apic_setup_io+0x72/frame 0xffffffff826ffb50
mi_startup() at mi_startup+0x118/frame 0xffffffff826ffb70
start_kernel() at start_kernel+0x10
Fix this by removing the assert in the MSI code, since it's possible
to get to the MSI initialization without having registered any other
interrupt sources.
338630:
lapic: skip setting intrcnt if lapic is not present
Instead of panicking. Legacy PVH mode doesn't provide a lapic, and
since native_lapic_intrcnt is called unconditionally this would cause
the assert to trigger. Change the assert into a continue in order to
take into account the possibility of systems without a lapic.
338631:
xen: legacy PVH fixes for the new interrupt count
Register interrupts using the PIC pic_register_sources method instead
of doing it in apic_setup_io. This is now required, since the internal
interrupt structures are not yet setup when calling apic_setup_io.
338725:
Fix a regression in r338360 when booting an x86 machine without APIC.
The atpic_register_sources callback tries to avoid registering interrupt
sources that would collide with an I/O APIC. However, the previous
implementation was failing to register IRQs 8-15 since the slave PIC
saw valid IRQs from the master and assumed an I/O APIC was present. To
fix, go back to registering all 8259A interrupt sources in one loop when
the master's register_sources method is invoked.