Add a method to iicbus to request IIC_M_NOSTOP behaviour for multibyte
transfers to be default. It simplifies porting code which assumes
such settings.
Discussed with: avg, llos, nwhitehorn
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Neel Natu [Sun, 26 Oct 2014 19:03:06 +0000 (19:03 +0000)]
Change the type of the first argument to the I/O emulation handlers to
'struct vm *'. Previously it used to be a 'void *' but there is no reason
to hide the actual type from the handler.
Ian Lepore [Sun, 26 Oct 2014 18:30:35 +0000 (18:30 +0000)]
Remove the ARM_DEVICE_MULTIPASS option and make its effect be the default.
Multipass device attachment was tested on many arm platforms by users and
only success was reported on the arm@ mailing list. This is just the
long-delayed followup of making it the default.
Multipass attachment is necessary when using vendor-supplied FDT data,
because our devices may need to be attached in a different order than they
are described in the FDT data.
Alan Cox [Sun, 26 Oct 2014 17:56:47 +0000 (17:56 +0000)]
By the time that pmap_init() runs, vm_phys_segs[] has been initialized. Obtaining
the end of memory address from vm_phys_segs[] is a little easier than obtaining it
from phys_avail[].
Jilles Tjoelker [Sun, 26 Oct 2014 17:50:33 +0000 (17:50 +0000)]
sh: Make getopts memory-safe if with changing arguments.
POSIX does not permit to continuing a getopts loop with different
arguments. For parsing the positional parameters, we handle this case by
resetting the getopts state when the positional parameters are changed in
any way (and the getopts state is local to a function). However, in the
syntax getopts <optstring> <var> <arg...>, changes could lead to invalid
memory access.
In the syntax getopts <optstring> <var> <arg...>, store a copy of the
arguments and continue to use them until getopts is reset.
Mateusz Guzik [Sun, 26 Oct 2014 06:04:09 +0000 (06:04 +0000)]
Tidy up sys_setgroups and kern_setgroups.
- 'groups' initialization to NULL is always ovewrwriten before use, so plug it
- get rid of 'goto out'
- kern_setgroups's callers already validate ngrp, so only assert the condition
- ngrp is an u_int, so 'ngrp < 1' is more readable as 'ngrp == 0'
Neel Natu [Sun, 26 Oct 2014 04:44:28 +0000 (04:44 +0000)]
Move the ACPI PM timer emulation into vmm.ko.
This reduces variability during timer calibration by keeping the emulation
"close" to the guest. Additionally having all timer emulations in the kernel
will ease the transition to a per-VM clock source (as opposed to using the
host's uptime keep track of time).
Neel Natu [Sun, 26 Oct 2014 03:03:41 +0000 (03:03 +0000)]
Don't pass the 'error' return from an I/O port handler directly to vm_run().
Most I/O port handlers return -1 to signal an error. If this value is returned
without modification to vm_run() then it leads to incorrect behavior because
'-1' is interpreted as ERESTART at the system call level.
Fix this by always returning EIO to signal an error from an I/O port handler.
Glen Barber [Sun, 26 Oct 2014 01:41:54 +0000 (01:41 +0000)]
Fix a few issues with creating VOLUME_LABEL for the
installation ISOs:
- TYPE, BRANCH, and REVISION are only defined if
OSRELEASE is not defined, so in situations where
one might set OSRELEASE for an in-house ISO build,
VOLUME_LABEL would be empty.
- makefs(8) limits the volume label to 32 characters,
which for the powerpc64 case, OSRELEASE expands to
FreeBSD-11.0-CURRENT-powerpc-powerpc64. Even with
removing the prefixing 'FreeBSD-', the string is 30
characters long, leaving zero room for suffixing the
type of ISO media (BO for bootonly, CD for cdrom, and
DVD for dvdrom).
Resolve these by defining VOLUME_LABEL when defining
OSRELEASE if unset. If OSRELEASE is defined by the
builder, use the OSRELEASE from that definition as the
VOLUME_LABEL.
In addition, for cases where both TARGET and TARGET_ARCH
are used for the VOLUME_LABEL, use TARGET_ARCH if it
differs from TARGET.
There are probably a few sharp edges here yet, but these
problems are going to affect the powerpc/powerpc64 builds
for 10.1-RELEASE, so the immediate concern is fixing the
underlying problem at hand quickly, and less so about the
elegance of the fix.
MFC after: 3 days
X-MFC-10.1: yes, asap
Sponsored by: The FreeBSD Foundation
John Baldwin [Sat, 25 Oct 2014 19:31:34 +0000 (19:31 +0000)]
Clarify that pthread_cleanup_push()/pop() are implemented as macros that
create a new code block and thus must be balanced at the same lexical
scope. (This is also a requirement in POSIX.)
Josh Paetzel [Sat, 25 Oct 2014 17:42:44 +0000 (17:42 +0000)]
This change addresses 4 bugs in ZFS exposed by Richard Kojedzinszky's
crash.sh script attached to FreeNAS bug 4109:
https://bugs.freenas.org/issues/4109
Three are in the snapshot layer:
a) AVG explains in his notes: https://wiki.freebsd.org/AvgVfsSolarisVsFreeBSD
"VOP_INACTIVE must not do any destructive actions to a vnode
and its filesystem node, nor invalidate them in any way."
gfs_vop_inactive and zfsctl_snapshot_inactive did just that. In
OpenSolaris VOP_INACTIVE is much closer to FreeBSD's VOP_RECLAIM.
Rename & move them to gfs_vop_reclaim and zfsctl_snapshot_reclaim
and merge in the requisite vnode_destroy from zfsctl_common_reclaim.
b) gfs_lookup_dot and various zfsctl functions do not honor the
FreeBSD VFS convention of only locking from the root downward. When
looking up ".." the convention is to drop the current leaf vnode lock before
acquiring the directory vnode and then subsequently re-acquiring the lock on the
leaf vnode. This fixes that in all the places that our exercised by crash.sh.
c) The snapshot may already be unmounted when the directory vnode is reclaimed.
Check for this case and return.
One in the common layer:
d) Callers of traverse expect the reference to the vnode passed in to be
maintained. Don't release it.
This last one may be an unclear contract. There may in fact be some callers that
do expect the reference to be dropped on success in addition to callers that
expect it to be released. In this case a further audit of the callers is needed
and a consensus on the correct behavior.
Alexander Motin [Sat, 25 Oct 2014 15:16:19 +0000 (15:16 +0000)]
Revert somewhat hackish geom_disk optimization, committed as part of r256880,
and the following r273143 commit, supposed to workaround introduced issue by
quite innocent-looking change.
While there is no clear understanding why, but r273143 is accused in data
corruption in some environments with high I/O load. I personally don't see
any problem in that commit, and possibly it is just a trigger to some other
bug somewhere, but better safe then sorry for now.
Michael Tuexen [Sat, 25 Oct 2014 09:25:29 +0000 (09:25 +0000)]
Fix a use of an uninitialized variable by makeing sure
that sctp_med_chunk_output() always initialized the reason_code
instead of relying on the caller.
The variable is only used for debugging purpose.
This issue was reported by Peter Bostroem from Google.
Enji Cooper [Sat, 25 Oct 2014 04:52:48 +0000 (04:52 +0000)]
- Print out "Bail out!" in die(..) so prove terminates immediately
- Handle the output from newer versions of openssl md5, similar to what
pjd@ did in r248304
Fix cpsw_detach() to not panic when called from cpsw_attach().
For an unkown reason (at moment), sometimes if_cpsw cannot read from PHY
and fails to attach calling cpsw_detach() which end up in a panic.
Fix it by doing the proper check before detach the miibus and also fix the
leak of few variables.
And to actually make it work, ether_ifattach() has to be moved to the end
of cpsw_attach() to avoid a race where calling ether_ifdetach() before
domain_init() (which will only run later on) would make it crash at
INP_INFO_RLOCK() on in_pcbpurgeif0().
Mark Johnston [Fri, 24 Oct 2014 20:29:14 +0000 (20:29 +0000)]
Eliminate conf_lock and instead rely on the NSS write lock to protect
NSS configuration state.
As a side effect, this fixes a race condition which can occur if multiple
threads call nsdispatch(3) concurrently before nsswitch.conf has been
parsed. Previously, the thread holding conf_lock could cause other threads
to return from nss_configure() before nsswitch.conf had been parsed, forcing
them to fall back to the default sources for their NSS methods.
John Baldwin [Fri, 24 Oct 2014 20:02:44 +0000 (20:02 +0000)]
The current POSIX semaphore implementation stores the _has_waiters flag
in a separate word from the _count. This does not permit both items to
be updated atomically in a portable manner. As a result, sem_post()
must always perform a system call to safely clear _has_waiters.
This change removes the _has_waiters field and instead uses the high bit
of _count as the _has_waiters flag. A new umtx object type (_usem2) and
two new umtx operations are added (SEM_WAIT2 and SEM_WAKE2) to implement
these semantics. The older operations are still supported under the
COMPAT_FREEBSD9/10 options. The POSIX semaphore API in libc has
been updated to use the new implementation. Note that the new
implementation is not compatible with the previous implementation.
However, this only affects static binaries (which cannot be helped by
symbol versioning). Binaries using a dynamic libc will continue to work
fine. SEM_MAGIC has been bumped so that mismatched binaries will error
rather than corrupting a shared semaphore. In addition, a padding field
has been added to sem_t so that it remains the same size.
John Baldwin [Fri, 24 Oct 2014 19:58:24 +0000 (19:58 +0000)]
Add COMPAT_FREEBSD9 and COMPAT_FREEBSD10 options to wrap code that
provides compatability for FreeBSD 9.x and 10.x binaries. Enable
these options in kernel configs that enable other COMPAT_FREEBSD<n>
options.
Fix a bug where DMA maps created with bus_dmamap_create() won't increment
the map count and without being able to keep track of the current map
allocation, bus_dma_tag_destroy() will fail to proceed and will return
EBUSY even after all the maps have been correctly destroyed with
bus_dmamap_destroy().
Found while testing the detach method of a NIC.
Tested on: BBB (am335x)
Reviewed by: cognet, ian
MFC after: 1 week
Rui Paulo [Fri, 24 Oct 2014 18:39:15 +0000 (18:39 +0000)]
HPET: create /dev/hpetN as a way to access HPET from userland.
In some cases, TSC is broken and special applications might benefit
from memory mapping HPET and reading the registers to count time.
Most often the main HPET counter is 32-bit only[1], so this only gives
the application a 300 second window based on the default HPET
interval.
Other applications, such as Intel's DPDK, expect /dev/hpet to be
present and use it to count time as well.
Although we have an almost userland version of gettimeofday() which
uses rdtsc in userland, it's not always possible to use it, depending
on how broken the multi-socket hardware is.
Install the acpi_hpet.h so that applications can use the HPET register
definitions.
[1] I haven't found a system where HPET's main counter uses more than
32 bit. There seems to be a discrepancy in the Intel documentation
(claiming it's a 64-bit counter) and the actual implementation (a
32-bit counter in a 64-bit memory area).
Enji Cooper [Fri, 24 Oct 2014 16:18:16 +0000 (16:18 +0000)]
- Add sys/types.h for the APIs in sys/sysctl.h
- Poke at VM_MIN_ADDRESS in machine/vmparam.h because FreeBSD doesn't have a
vm.minaddress sysctl analog
- Expect ENOMEM instead of EAGAIN in mlock_limits
- Provide mlock an mmap'ed page twice to simulate MAP_WIRED on NetBSD
Enji Cooper [Fri, 24 Oct 2014 16:07:31 +0000 (16:07 +0000)]
Correct my previous commit:
- getrusage_utime_back succeeds reliably on FreeBSD
- getrusage_utime_zero passes/fails in a seemingly non-deterministic manner.
Skip it for now (and fix it later)
In the initial port of this testcase to FreeBSD, the results failed reliably
in the same manner as it does on NetBSD
Roger Pau Monné [Fri, 24 Oct 2014 09:48:58 +0000 (09:48 +0000)]
amd64: make uiomove_fromphys functional for pages not mapped by the DMAP
Place the code introduced in r268660 into a separate function that can be
called from uiomove_fromphys. Instead of pre-allocating two KVA pages use
vmem_alloc to allocate them on demand when needed. This prevents blocking if
a page fault is taken while physical addresses from outside the DMAP are
used, since the lock is now removed.
Also introduce a safety catch in PHYS_TO_DMAP and DMAP_TO_PHYS.
amd64/amd64/pmap.c:
- Factor out the code to deal with non DMAP addresses from pmap_copy_pages
and place it in pmap_map_io_transient.
- Change the code to use vmem_alloc instead of a set of pre-allocated
pages.
- Use pmap_qenter and don't pin the thread if there can be page faults.
amd64/amd64/uio_machdep.c:
- Use pmap_map_io_transient in order to correctly deal with physical
addresses not covered by the DMAP.
amd64/include/pmap.h:
- Add the prototypes for the new functions.
amd64/include/vmparam.h:
- Add safety catches to make sure PHYS_TO_DMAP and DMAP_TO_PHYS are only
used with addresses covered by the DMAP.
Enji Cooper [Fri, 24 Oct 2014 03:53:49 +0000 (03:53 +0000)]
- Test for EINVAL requirement when passing an invalid flag in to msync(2)
- Expect ENOMEM instead of EFAULT when msync'ing a previously munmap'ed region
on FreeBSD
Enji Cooper [Fri, 24 Oct 2014 03:42:37 +0000 (03:42 +0000)]
- Ignore EINVAL check with mknod(path, S_IFCHR, -1) as the testcase is always
executed on a non-devfs filesystem
- Expect mknod(path, S_IFREG, 0) to fail on FreeBSD
Justin Hibbits [Fri, 24 Oct 2014 03:33:01 +0000 (03:33 +0000)]
Three updates to PowerPC FBT:
* Use a constant to define the number of stack frames in a probe exception.
* Only allow function symbols in powerpc64 ('.' prefixed)
* Set the fbtp_roffset for return probes, so the correct dtrace_probe call is
made.
In all cases except CTLTYPE_STRING, penv is NULL here, so passing it
indiscriminately to printf() and freeenv() is incorrect. Add a NULL
check before freeenv(); as for printf(), we could use req.newptr
instead, but we'd have to select the correct format string based on
the type, and that's too much work for an error message, so just
remove it.
Marcel Moolenaar [Thu, 23 Oct 2014 22:33:27 +0000 (22:33 +0000)]
Add generated headers xoversion.h and xoconfig.h. These are the result
of configuring the source tree:
% automake --add-missing
% autoconf
% ./configure
Ian Lepore [Thu, 23 Oct 2014 22:21:22 +0000 (22:21 +0000)]
Install a temporary workaround to avoid problems in fdt data with linux's
workaround for an imx6 chip erratum. Linux works around the bug with
changes in fdt data that we can't currently handle, so to enable running
with standard vendor-supplied fdt data, this watches for an attempt to map
the gpio1_6 interrupt and remaps it back to the standard ethernet interrupt.
This can be undone when the intrng project is completed and our gpio drivers
can also be interrupt controllers.
Alan Cox [Thu, 23 Oct 2014 21:02:40 +0000 (21:02 +0000)]
Under PAE ULONG is insufficient for representing the physical memory size. Use QUAD
for "hw.physmem" so that a physical memory size greater than 4 GB can be specified
from the boot loader.