jhb [Tue, 16 May 2017 18:42:44 +0000 (18:42 +0000)]
Skip tests depending on coredumps if coredumps are disabled via kern.coredump.
The kern.coredump sysctl can be set to 0 to disable coredumps. Skip the
'status_coredump' and 'wait6_coredumped' tests if this sysctl is set to 0
rather than reporting a failure.
jhb [Tue, 16 May 2017 18:39:23 +0000 (18:39 +0000)]
Fix p_endcopy.
For p_endcopy to work correctly, it must be the name of the next element
in struct proc after the end of the copy region, not the name of the
last element in the copy region. Currently, the last element
(p_elf_flags) is not being copied. In addition, the p_xexit and
p_xsig fields should not be copied on fork, so move them out of the
copy region.
Reviewed by: kib
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D10677
tsoome [Tue, 16 May 2017 17:35:05 +0000 (17:35 +0000)]
libstand: increase nfs max read size to 16k
With ip fragment reassembly implemented, it makes sense to allow
larger nfs reads. Note due to loader heap size limit, we do not want
to set too large maximum read size. Also we do not change default read size.
cem [Tue, 16 May 2017 16:20:22 +0000 (16:20 +0000)]
Correct page frame mask constant used in pmap_change_attr_locked
This was introduced in r290156. It's present in 11.0, but not any 10.x
release unless someone decided to MFC it.
It affects ordinary pages right above the DMAP limit, which is effectively
system memory rounded up to a 1 GB (3rd level superpage) boundary (or up to
a minimum of 4 GB, on small systems).
hselasky [Tue, 16 May 2017 14:21:37 +0000 (14:21 +0000)]
Avoid use of contiguous memory allocations in busdma when possible.
This patch improves the boundary checks in busdma to allow more cases
using the regular page based kernel memory allocator. Especially in
the case of having a non-zero boundary in the parent DMA tag. For
example AMD64 based platforms set the PCI DMA tag boundary to
PCI_DMA_BOUNDARY, 4GB, which before this patch caused contiguous
memory allocations to be preferred when allocating more than PAGE_SIZE
bytes. Even if the required alignment was less than PAGE_SIZE bytes.
This patch also fixes the nsegments check for using kmem_alloc_attr()
when the maximum segment size is less than PAGE_SIZE bytes.
Updated some comments describing the code in question.
jonathan [Tue, 16 May 2017 13:27:44 +0000 (13:27 +0000)]
Rename rtld's parse_libdir to parse_integer.
This is a more accurate name, as the integer doesn't have to be a library
directory descriptor. It is also a prerequisite for more argument parsing
coming in the near future (e.g., parsing explicit binary descriptors).
loos [Tue, 16 May 2017 05:10:15 +0000 (05:10 +0000)]
Add the Marvell SDHCI controller to the list of supported devices in
sdhci_fdt.
Enable the SDHCI controller, bus and devices on ARMADA38X kernel.
Tested on: ClearFog Pro
Reviewed by: Marcin Wojtas <mw at semihalf.com>
Sponsored by: Rubicon Communications, LLC (Netgate)
Differential Revision: https://reviews.freebsd.org/D10606
loos [Tue, 16 May 2017 04:47:50 +0000 (04:47 +0000)]
Add the SDHCI Address Decoder registers and routines for ARMADA 38X.
Tested on: ClearFog Pro
Reviewed by: Marcin Wojtas <mw at semihalf.com>
Sponsored by: Rubicon Communications, LLC (Netgate)
Differential Revision: https://reviews.freebsd.org/D10601
ngie [Mon, 15 May 2017 22:52:25 +0000 (22:52 +0000)]
Start writing up some basic feature tests for procstat
These tests query a running process for information related to the -b,
-c, -e, and -f flags; the -f testcase is largely stubbed out, pending
additional work to determine a good, deterministic descriptor.
Core file test support is coming soon--it requires a bit more effort
due to the fact that:
- coredumps can be disabled (kern.coredump=0).
- corefiles can be put in different directories than the current
directory, or be named something other than `<prog>.core`
(`kern.corefile`).
glebius [Mon, 15 May 2017 21:58:36 +0000 (21:58 +0000)]
Reduce in_pcbinfo_init() by two params. No users supply any flags to this
function (they used to say UMA_ZONE_NOFREE), so flag parameter goes away.
The zone_fini parameter also goes away. Previously no protocols (except
divert) supplied zone_fini function, so inpcb locks were leaked with slabs.
This was okay while zones were allocated with UMA_ZONE_NOFREE flag, but now
this is a leak. Fix that by suppling inpcb_fini() function as fini method
for all inpcb zones.
markj [Mon, 15 May 2017 21:44:40 +0000 (21:44 +0000)]
Ensure that profile and tick probes provide a non-zero PC value.
The idle thread may process callouts while reloading the timer in
cpu_activeclock(). In this case, provide a representative value, &cpu_idle,
instead of 0 for args[0] so that the active thread can be more easily
identified from the probe.
This addresses intermittent failures of the profile-n/tst.argtest.d test.
kib [Mon, 15 May 2017 20:52:43 +0000 (20:52 +0000)]
Ensure that resume path on amd64 only accesses page tables for normal
operation after processor is configured to allow all required
features.
In particular, NX must be enabled in EFER, otherwise load of page
table element with nx bit set causes reserved bit page fault. Since
malloc uses direct mapping for small allocations, in particular for
the suspension pcbs, and DMAP is nx after r316767, this commit tripped
fault on resume path.
Restore complete state of EFER while wakeup code is still executing
with custom page table, before calling resumectx, instead of trying to
guess which features might be needed before resumectx restored EFER on
its own.
Bisected and tested by: trasz
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
emaste [Mon, 15 May 2017 20:41:29 +0000 (20:41 +0000)]
bsdgrep: add more tests for different binary flags
The existing 'binary' test in netbsd-tests/ does a basic check of the
default treatment for binary behavior, but not much more than that.
Given some opportunity for breakage recently that did not trigger any
failures, add some tests to cover the three different binary file
behaviors (a, -I, -U) and their --binary-files= equivalent values.
emaste [Mon, 15 May 2017 20:18:14 +0000 (20:18 +0000)]
uniq: allow -c to be used with -d or -u
Bring in some bits from NetBSD and lift the restriction in uniq(1) that
-c cannot be used with the -d and -u options. This restriction seems
unnecessary and is supported at least by GNU, OpenBSD, and NetBSD. Lift
the restriction and simplify the show() logic a little bit to maintain
functionality when -c is provided with -d/-u.
Also with this change, -d and -u are now actually a mutually exclusive,
albeit valid, combination. Given that they both indicate opposite
behavior, uniq(1) will no longer output anything if both -d and -u are
supplied. This is in line with NetBSD as well as GNU.
Adjust the man page and usage() to reflect that -c is its own standalone
option.
kib [Mon, 15 May 2017 18:48:58 +0000 (18:48 +0000)]
Make ld-elf.so.1 directly executable.
Check if passed phdr is actually phdr of the interpreter itself, and
decide that this is the case of direct execution. In this case, the
binary to activate is specified in the argv[1]. After opening it,
shift down on-stack structure with argv, env and aux vectors to
emulate execution of the binary and not of the interpreter.
Reviewed by: emaste
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
Differential revision: https://reviews.freebsd.org/D10701
kib [Mon, 15 May 2017 18:47:25 +0000 (18:47 +0000)]
Fix the AT_EXECFD functionality.
If the mapped object is linked at specific address, we must obey it.
If AT_EXECFD is not used, only in-kernel ELF image activator needed to
keep the mapping address, since only binaries are linked at the fixed
address, and binaries are mapped by kernel in this case.
Reviewed by: emaste
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
X-Differential revision: https://reviews.freebsd.org/D10701
emaste [Mon, 15 May 2017 17:57:09 +0000 (17:57 +0000)]
getusershell: don't write past end of line buffer reading local shells
_local_initshells did not reset cp to the beginning of the line buffer
for every iteration that it called fgets(3), leading to writing past the
end of line with fairly long /etc/shells or excessively long line
lengths. Correct this by properly resetting cp.
emaste [Mon, 15 May 2017 17:51:01 +0000 (17:51 +0000)]
bsdgrep: don't allow negative -A / -B / -C
Previously, when given a negative -A/-B/-C argument bsdgrep would
overflow the respective context flag(s) and exhibited surprising
behavior.
Fix this by removing unsignedness of Aflag/Bflag and erroring out if
we're given a value < 0. Also adjust the type used to track 'tail'
context in procfile() so that it accurately reflects the Aflag value
rather than overflowing and losing trailing context.
This also fixes an inconsistency previously existing between -n and
-C "n" behavior. They are now both limited to LLONG_MAX, to be
consistent.
Add some test cases to make sure grep errors out properly for both
negative context values as well as non-numeric context values rather
than giving bogus matches.
kib [Mon, 15 May 2017 17:34:17 +0000 (17:34 +0000)]
Simplify cleanup on failure in realpath(3).
If realpath() allocated memory for result and failed, the memory is
freed in each place where return is performed. More, the function
needs to track the allocation status, to not free user-supplied
buffer.
Consolidate the memory handling in the wrapper, freeing the buffer if
the actual worker failed.
Reviewed by: emaste (previous version)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
Differential revision: https://reviews.freebsd.org/D10670
kib [Mon, 15 May 2017 17:14:53 +0000 (17:14 +0000)]
Fix several buffer overflows in realpath(3).
- The statement "left_len -= s - left;" does not take the slash into
account if one was found. This results in the invariant
"left[left_len] == '\0'" being violated (and possible buffer
overflows). The patch replaces the variable "s" with a size_t
"next_token_len" for more clarity.
- "slen" from readlink(2) can be 0 when encountering empty
symlinks. Then, further down, "symlink[slen - 1]" underflows the
buffer. When slen == 0, realpath(3) should probably return ENOENT
(http://austingroupbugs.net/view.php?id=825,
https://lwn.net/Articles/551224/).
Some other minor issues:
- The condition "resolved_len >= PATH_MAX" cannot be true.
- Similarly, "s - left >= sizeof(next_token)" cannot be true, as long
as "sizeof(next_token) >= sizeof(left)".
- Return ENAMETOOLONG when a resolved symlink from readlink(2) is too
long for the symlink buffer (instead of just truncating it).
- "resolved_len > 1" below the call to readlink(2) is always true as
"strlcat(resolved, next_token, PATH_MAX);" always results in a
string of length > 1. Also, "resolved[resolved_len - 1] = '\0';" is
not needed; there can never be a trailing slash here.
- The truncation check for "strlcat(symlink, left, sizeof(symlink));"
should be against "sizeof(symlink)" (the third argument to strlcat)
instead of "sizeof(left)".
kib [Mon, 15 May 2017 10:02:45 +0000 (10:02 +0000)]
mnt_vnode_next_active: use conventional lock order when trylock fails.
Previously, when the VI_TRYLOCK failed, we would spin under the mutex
that protects the vnode active list until we either succeeded or
noticed that we had hogged the CPU. Since we were violating the lock
order, this would guarantee that we would become a hog under any
deadlock condition (e.g. a race with vdrop(9) on the same vnode). In
the presence of many concurrent threads in sync(2) or vdrop etc, the
victim could hang for a long time.
Now, avoid spinning by dropping and reacquiring the locks in the
conventional lock order when the trylock fails. This requires a dance
with the vnode hold count.
cy [Mon, 15 May 2017 03:39:35 +0000 (03:39 +0000)]
Just like r318173, which was for outputting IPv6 addresses in tree
pools, implement outputting of IPv6 addresses in the ippool debug list
of hash type pools (ippool -l -d -t hash). Currently IPv6 in ippool tree
type pool handling is mostly implemented.
This continues theseries of commits to remediate ippool.
This will be MFCed with a yet to be committed series of fixes to ippool
after it has been fully remediated.
marius [Sun, 14 May 2017 21:33:01 +0000 (21:33 +0000)]
- Unlike as in the PCI case, when attached to ACPI, Intel Bay Trail
and Braswell eMMC and SDXC controllers share the same IDs. Like in
the PCI case, Braswell eMMC needs the SDHCI_QUIRK_DATA_TIMEOUT_1MHZ
quirk (see r311794 for the corresponding change to the sdhci(4) PCI
PCI front-end), though. However, due to the shared ACPI IDs, this
is trickier to do.
- Intel Apollo Lake eMMC and SDXC controllers are affected by the
APL18 ("Using 32-bit Addressing Mode With SD/eMMC Controller May
Lead to Unpredictable System Behavior") silicon bug [1]. When this
erratum hits, typically both SDHCI and XHCI controllers wedge.
According to Intel, using ADMA2 with 64-bit addressing and 96-bit
descriptors serves as a workaround. Until such times when sdhci(4)
has ADMA2 support, flag DMA as broken for affected interfaces.
This turns out to work around the problem, too, at the cost of
performance.
- In the sdhci(4) ACPI front-end, probe the Intel Apollo Lake eMMC
and SDXC controllers, too.
cy [Sun, 14 May 2017 21:18:01 +0000 (21:18 +0000)]
Separate the ipfilter function/static string from the error with a
colon (:) in error messages to assist the user in parsing out the error
from where or which object the error message refers to.
ngie [Sun, 14 May 2017 18:47:09 +0000 (18:47 +0000)]
Mark errf _Noreturn, and mark errf and warn __printflike
The _Noreturn attribute was added to placate Coverity and other static
analysis tools. The __printflike attribute was added to catch issues
with the calls related to printf(3) abuse.
- Modify the code to facilitate the __printflike attribute addition.
- Convert errf calls in to_mb(..) and to_mb_string(..) to warn(..) so
the calls will return instead of exiting, as the code suggests it
should.
jilles [Sun, 14 May 2017 13:14:19 +0000 (13:14 +0000)]
sh: Fix '-' from quoted arithmetic in case/glob pattern range.
It does not make much sense to generate the '-' in a pattern bracket
expression using arithmetic expansion, but it does not make sense to forbid
it either.
Try to avoid reprocessing the string if it is unnecessary.
rmacklem [Sun, 14 May 2017 00:38:41 +0000 (00:38 +0000)]
Change the default uid/gid values for nobody/nogroup to 65534/65533.
The default values found in /etc/passwd and /etc/group are 65534, 65533.
In mountd.c, the defaults were -2, which was 65534 back when uid_t was 16bits.
Without this patch, a file created by root on an NFS exported volume without
the "-root=" export option will end up owned by uid 4**32 - 2.
When discussed on freebsd-current@, it seemed that users preferred the
values being changed to 65534/65533.
I have not added code to acquire these values from the databases, since
the mountd daemon might get "stuck" during startup waiting for a non-responsive
password database server.
dim [Sat, 13 May 2017 22:36:54 +0000 (22:36 +0000)]
Silence a -Wunused warning about the junk variable being used to raise
an inexact floating point exception. The variable cannot be eliminated,
unfortunately, otherwise the desired addition triggering the exception
will be emitted neither by clang, nor by gcc.
jilles [Sat, 13 May 2017 20:28:32 +0000 (20:28 +0000)]
sh: Add test for arithmetic expansion in [x-y] pattern range.
It does not make much sense to generate the '-' in a pattern bracket
expression using arithmetic expansion, but it does not make sense to forbid
it either.
kib [Sat, 13 May 2017 18:59:27 +0000 (18:59 +0000)]
In _rtld(), reorder local declarations to compact the block and
partially sort them by style(9). Move locals declarations from nested
blocks into the block at function start.
ngie [Sat, 13 May 2017 18:41:24 +0000 (18:41 +0000)]
Add missing braces around MCAST_EXCLUDE check when KTR support is
compiled into the kernel
This ensures that .iss_asm (the number of ASM listeners) isn't incorrectly
decremented for MLD-layer source datagrams when inspecting im*s_st[1]
(the second state in the structure).
jasone [Sat, 13 May 2017 17:49:53 +0000 (17:49 +0000)]
Fix __pthread_mutex_trylock() to call THR_CRITICAL_LEAVE() on failure rather
than on success. This regression was introduced by r300043 (Add implementation
of robust mutexes...).
andrew [Sat, 13 May 2017 13:03:20 +0000 (13:03 +0000)]
Allocate a cacheline when reading or writing to write through memory. The
hardware will still write to memory, however following reads will be from
the cache.
andrew [Sat, 13 May 2017 13:01:15 +0000 (13:01 +0000)]
Add the VM_MEMATTR_WRITE_THROUGH memory type to arm64 and use it to support
VM_MEMATTR_WRITE_COMBINING in the kernel. This fixes a bug where Xorg would
use write back cached memory for its graphics buffers. This would produce
artifacts on the screen as cachelines were written to memory.
kib [Fri, 12 May 2017 15:34:59 +0000 (15:34 +0000)]
Do not wake up sleeping thread in reschedule_signals() if the signal
is blocked. The spurious wakeup might result in spurious EINTR.
The reschedule_signals() function is called when the calling thread
has the signal mask changed. For each newly blocked signal, we try to
find a thread which might have the signal not blocked. If no such
thread exists, sigtd() returns random thread, which must not be waken
up. I decided that re-checking, as suggested by PR submitter, is more
reasonable change than to change sigtd() interface, due to other uses
of sigtd(). signotify() already performs this check.
Submitted by: Duane <parakleta@darkreality.org>
PR: 219228
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
lidl [Fri, 12 May 2017 15:20:12 +0000 (15:20 +0000)]
Refine and update blacklist support in sshd
Adjust notification points slightly to catch all auth failures,
rather than just the ones caused by bad usernames.
Modify notification point for bad usernames to send new type of
BLACKLIST_BAD_USER. (Support in libblacklist will be forthcoming soon.)
Add guards to allow library headers to expose the enum of action values.
Reviewed by: des
Approved by: des
Sponsored by: The FreeBSD Foundation
adrian [Fri, 12 May 2017 06:33:07 +0000 (06:33 +0000)]
[iwm] Adjust if_iwm_sta.h prototypes, don't pass iwm_node to rm_sta().
* Since a RUN -> INIT/SCAN transition seems to immediately destroy the
ieee80211_node for the AP, we can't read the in_assoc value from there.
Instead just directly pass that information via a boolean_t argument.
* Adds iwm_mvm_rm_sta_id() function, which just unconditionally removes
the station from the firmware.
* The iwm_mvm_rm_sta() function shouldn't actually remove the station from
firmware when we are still associated (i.e. during a RUN -> INIT/SCAN
transition).
* So when disassociating we will first call iwm_mvm_rm_sta() to drain the
queues/fifos. Later during disassociation we will then use
iwm_mvm_rm_sta_id() to actually remove the station.
adrian [Fri, 12 May 2017 05:49:24 +0000 (05:49 +0000)]
[iwm] Process multiple frames per RX buffer.
* Factor out iwm_handle_rxb() function from iwm_notif_intr().
* Removing the IWM_FH_RCSR_CHNL0_RX_CONFIG_SINGLE_FRAME_MSK flag allows
the device to put multiple frames (both command responses and 80211
frames) into a single RX buffer.
* Uses m_copym() to split up the receive buffers when multiple 80211
frames are received in one RX buffer. The effect is basically the same
as when using m_split(), but we want to keep the original mbuf around
when calling iwm_mvm_rx_rx_mpdu() to make error handling a bit easier
for now.
* Contains a small optimization to avoid the m_copym() when only a single
80211 frame is received in one RX buffer (i.e. matching the existing
behaviour).
adrian [Fri, 12 May 2017 05:19:08 +0000 (05:19 +0000)]
[if_iwm] Partly sync if_iwm_binding.c to Linux iwlwifi code.
* Store macid and color values in struct iwm_vap, to avoid hardcoded
constants a bit.
* Add iwm_mvm_binding_remove_vif() function (will be used in disconnecting
from an access point without resetting the whole device).
* Not adding code from Linux iwlwifi yet, to handle one PHY context to
be bound to several VAPs/virtual-interfaces, it's definitely not needed
in the near future.
ngie [Fri, 12 May 2017 05:17:49 +0000 (05:17 +0000)]
Fix up previous commit
- Apply the logic to the FreeBSD block
- Fix a typo with the getconf(1) call that I would have caught, were
it not for the fact that I got the blocks wrong.
- Consolidate the hardcoded buffer sizes to the NetBSD block.
This would have been discovered had I run the test on a system where
PATH_MAX != 1024 (I don't have that at my disposal right at this moment).
ngie [Fri, 12 May 2017 05:06:48 +0000 (05:06 +0000)]
ssp_test:read:: query the value of MAXPATHLEN via getconf(1)
In the event the value of PATH_MAX was changed, the assumption that
MAXPATHLEN is 1024 (and hence the buffer length required to trigger
SSP to fail for read(2)) would be invalidated. Query getconf(1) for
the actual value of MAXPATHLEN via _XOPEN_PATH_MAX instead, and
increment the value by 1 to ensure that the SSP support tests the
stack smashing support properly.
markj [Thu, 11 May 2017 17:03:45 +0000 (17:03 +0000)]
Let ptracestop() suspend threads sleeping in an SBDRY section.
When a thread enters ptracestop(), for example because it had received
SIGSTOP from ptrace(PT_ATTACH), it attempts to suspend other threads in
the same process. In the case of a thread sleeping interruptibly in an
SBDRY section, sig_suspend_threads() must wake the thread and allow it to
reach the user-mode boundary. However, sig_suspend_threads() would
erroneously avoid waking up such threads, resulting in an apparent hang.
gjb [Thu, 11 May 2017 16:37:29 +0000 (16:37 +0000)]
Update release/scripts/atlas-upload.sh to account for API changes
made recently by Atlas Hashicorp. The data returned from GET and
POST requests has changed, which caused a number of regex patterns
to fail to be properly identified as 'success' or 'failure', which
ended up in upload/publish failures.
Tested with: 12-CURRENT
MFC after: 3 days
Sponsored by: The FreeBSD Foundation
asomers [Thu, 11 May 2017 16:26:56 +0000 (16:26 +0000)]
vdev_geom may associate multiple vdevs per g_consumer
vdev_geom.c currently uses the g_consumer's private field to point to a
vdev_t. That way, a GEOM event can cause a change to a ZFS vdev. For
example, when you remove a disk, the vdev's status will change to REMOVED.
However, vdev_geom will sometimes attach multiple vdevs to the same GEOM
consumer. If this happens, then geom events will only be propagated to one
of the vdevs.
Fix this by storing a linked list of vdevs in g_consumer's private field.
trasz [Thu, 11 May 2017 08:39:55 +0000 (08:39 +0000)]
Improve build(7): add missing "buildkernel" and "installkernel"
to the example, change the architectures to something more common,
and improve description of defaults for TARGET.
ngie [Thu, 11 May 2017 08:06:46 +0000 (08:06 +0000)]
Mark all md tests as requiring unsafe AIO in order to function
These tests have been flapping (failing<->passing) on Jenkins for months.
It passes reliably for me if unsafe AIO is permitted, but it doesn't
pass on Jenkins reliably if unsafe AIO is disabled (the current default).
Mark the tests as requiring unsafe AIO to mitigate the intermittent
failures when unsafe AIO isn't permitted. If the kernel code is changed
to reliably function with md(4) devices using unsafe AIO, this commit can
be reverted.
ngie [Thu, 11 May 2017 06:35:23 +0000 (06:35 +0000)]
procstat(1): clarify the Signal Disposition section
- Fix a typo (SIGIGN -> SIG_IGN). Use .Dv when referencing SIG_IGN.
- Use semi-colons as soft breaks when separating sentences for
the FLAGS section.
- Tweak wording for C slightly to flow better and to be a bit
more technically correct (signals with handlers installed will
be caught by the target program).
- Reference signal(3) in the SEE ALSO section.