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.
Move if_get_counter initialization from if_attach into if_alloc.
Also, initialize all counters before ifnet will become available in the system.
This fixes possible access to uninitialized ifned fields.
Fix a bug where some DTS layouts could cause the premature ending of the
search (i.e. without returning any result) and you would end up with a
random MAC address.
Change the search algorithm to a recursive one to ensure that all the nodes
on DTS will be verified.
The previous algorithm could not keep up if the DTS has too many sub-nodes.
Enji Cooper [Thu, 23 Oct 2014 08:05:03 +0000 (08:05 +0000)]
- Omit setrlimit_nthr testcase on FreeBSD (requires lwp.h, et al)
- Expect overflow with rlim_max at INT64_MAX, not UINT64_MAX (rlim_t is int64_t
on FreeBSD)
Enji Cooper [Thu, 23 Oct 2014 06:18:28 +0000 (06:18 +0000)]
Port t_pipe2.c to FreeBSD
- Omit the pipe2_nosigpipe testcase on FreeBSD (FreeBSD doesn't have
O_NOSIGPIPE).
- Convert "fcntl(n, F_CLOSEM)" to "closefrom(n)".
- Save and restore the resource limit on the number of files (RLIMIT_NOFILE).
Ian Lepore [Thu, 23 Oct 2014 03:13:14 +0000 (03:13 +0000)]
Unconditionally enable the clocks for all imx6 devices that we have drivers
for, or that are required to run the chip (such as busses). Turn off all
the devices we don't yet have drivers for.
Some day we will have a fully functional imx6 clock driver so that we can
manage clocks based on fdt data. This will have to do until then.
Colin Percival [Wed, 22 Oct 2014 23:41:15 +0000 (23:41 +0000)]
Populate the GELI passphrase cache with the kern.geom.eli.passphrase
variable (if any) provided in the boot environment. Unset it from
the kernel environment after doing this, so that the passphrase is
no longer present in kernel memory once we enter userland.
This will make it possible to provide a GELI passphrase via the boot
loader; FreeBSD's loader does not yet do this, but GRUB (and PCBSD)
will have support for this soon.
Mark Johnston [Wed, 22 Oct 2014 23:35:56 +0000 (23:35 +0000)]
Fix some buglets in the error-handling of getdevice(). In particular, report
an error if the argument to pciconf -a doesn't have a unit number, rather
than triggering an assertion failure.
Colin Percival [Wed, 22 Oct 2014 23:35:32 +0000 (23:35 +0000)]
Avoid leaking data from the kernel environment: When we convert the
initial static environment to a dynamic one, zero the static environment
buffer, and zero individual values when kern_unsetenv and freeenv are
called.
Rick Macklem [Wed, 22 Oct 2014 22:27:51 +0000 (22:27 +0000)]
Clip the settings for the NFS rsize, wsize mount options
to a power of 2. For non-power of 2 settings, intermittent
page faults have been reported. Although the bug that causes
these page faults/crashes has not been identified, it does
not appear to occur when rsize, wsize is a power of 2.
Enji Cooper [Wed, 22 Oct 2014 21:04:54 +0000 (21:04 +0000)]
The NetBSD libc tests use several definitions/macros that aren't available in
FreeBSD
Add the missing compat definitions/macros to lib/libnetbsd so the testcases
can be compiled with libnetbsd without having to invent ad hoc #define's, or
having to convert things over to FreeBSD idioms
Rick Macklem [Wed, 22 Oct 2014 20:47:11 +0000 (20:47 +0000)]
Clip the settings for the NFS rsize, wsize mount options
to a power of 2. For non-power of 2 settings, intermittent
page faults have been reported. Although the bug that causes
these page faults/crashes has not been identified, it does
not appear to occur when rsize, wsize is a power of 2.
Roger Pau Monné [Wed, 22 Oct 2014 17:09:12 +0000 (17:09 +0000)]
netback: change xnb naming convention
Current FreeBSD netback names the interface with xnb<device unit>, but
this is not suitable for usage with the Xen toolstack, which expects
something similar to <prefix><domid><handle>. In order to solve this,
change the netback naming convention to use xnb<domid>.<handle>.
Sponsored by: Citrix Systems R&D
dev/xen/netback/netback.c:
- Change netback to use the nomenclature stated above.
Roger Pau Monné [Wed, 22 Oct 2014 17:07:20 +0000 (17:07 +0000)]
xen: implement the privcmd user-space device
This device is only attached to priviledged domains, and allows the
toolstack to interact with Xen. The two functions of the privcmd
interface is to allow the execution of hypercalls from user-space, and
the mapping of foreign domain memory.
Sponsored by: Citrix Systems R&D
i386/include/xen/hypercall.h:
amd64/include/xen/hypercall.h:
- Introduce a function to make generic hypercalls into Xen.
xen/interface/xen.h:
xen/interface/memory.h:
- Import the new hypercall XENMEM_add_to_physmap_range used by
auto-translated guests to map memory from foreign domains.
dev/xen/privcmd/privcmd.c:
- This device has the following functions:
- Allow user-space applications to make hypercalls into Xen.
- Allow user-space applications to map memory from foreign domains,
this is accomplished using the newly introduced hypercall
(XENMEM_add_to_physmap_range).
xen/privcmd.h:
- Public ioctl interface for the privcmd device.
x86/xen/hvm.c:
- Remove declaration of hypercall_page, now it's declared in
hypercall.h.
conf/files:
- Add the privcmd device to the build process.
Roger Pau Monné [Wed, 22 Oct 2014 16:58:52 +0000 (16:58 +0000)]
xen: add a Xen to BSD error translation
Since Xen and FreeBSD error codes are completely different add a
translation layer in order to convert Xen error codes into native
FreeBSD error codes. This will be used by the privcmd device, which
needs to return the hypercall errors into user-space.
Sponsored by: Citrix Systems R&D
xen/error.h:
- Import Xen error codes.
- Create a table to map Xen error codes into FreeBSD native error
codes.
- Create an inline function that performs the translation.
Roger Pau Monné [Wed, 22 Oct 2014 16:57:11 +0000 (16:57 +0000)]
xen: import a proper event channel user-space device
The user-space event channel device is used by applications to receive
and send event channel interrupts. This device is based on the Linux
evtchn device.
Sponsored by: Citrix Systems R&D
xen/evtchn/evtchn_dev.c:
- Remove the old event channel device, which was already disabled in
the build system.
dev/xen/evtchn/evtchn_dev.c:
- Import a new event channel device based on the one present in
Linux.
- This device allows the following operations:
- Bind VIRQ event channels (ioctl).
- Bind regular event channels (ioctl).
- Create and bind new event channels (ioctl).
- Unbind event channels (ioctl).
- Send notifications to event channels (ioctl).
- Reset the device shared memory ring (ioctl).
- Unmask event channels (write).
- Receive event channel upcalls (read).
- The new code is MP safe, and can be used concurrently.
conf/files:
- Add the new device to the build system.
Roger Pau Monné [Wed, 22 Oct 2014 16:51:52 +0000 (16:51 +0000)]
xen: allow to register event channels without handlers
This is needed by the event channel user-space device, that requires
registering event channels without unmasking them. intr_add_handler
will unconditionally unmask the event channel, so we avoid calling it
if no filter/handler is provided, and then the user will be in charge
of calling it when ready.
In order to do this, we need to change the opaque type
xen_intr_handle_t to contain the event channel port instead of the
opaque cookie returned by intr_add_handler, since now registration of
event channels without handlers are allowed. The cookie will now be
stored inside of the private xenisrc struct. Also, introduce a new
function called xen_intr_add_handler that allows adding a
filter/handler after the event channel has been registered.
Sponsored by: Citrix Systems R&D
x86/xen/xen_intr.c:
- Leave the event channel without a handler if no filter/handler is
provided to xen_intr_bind_isrc.
- Don't perform an evtchn_mask_port, intr_add_handler will already do
it.
- Change the opaque type xen_intr_handle_t to contain a pointer to
the event channel port number, and make the necessary changes to
related functions.
- Introduce a new function called xen_intr_add_handler that can be
used to add filter/handlers to an event channel after registration.
xen/xen_intr.h:
- Add prototype of xen_intr_add_handler.
Actually change the USB_PORT_ROOT_RESET_DELAY definition as stated in
the r241987 commit message, instead of having users locally overriding
the value using tunables in /boot/loader.conf .
Warner Losh [Wed, 22 Oct 2014 03:39:11 +0000 (03:39 +0000)]
My previous commit exposed an issue as it fixed a different
issue. lib/atf isn't a prereq_lib, since it isn't required for other
libraries to build. Remove it. The old kludge of always building it
had effectively been retired. Since we don't want to build the
libraries with the tests when we're bootstrapping, invent
MK_TESTS_SUPPORT which normally defaults to the current MK_TESTS
value, except when explicitly defined. Make lib/atf depend on it being
yes. When building the libraries set MK_TESTS to no, and
MK_TESTS_SUPPORT to the current value of MK_TESTS so that later stages
of the build work correctly. This should fix (and does for me)
people's issues with parallel builds racing between lib/atf and
libexec/atf. Since lib/atf is built during the libraries phase, the
race disappears.
Marcel Moolenaar [Wed, 22 Oct 2014 01:37:32 +0000 (01:37 +0000)]
Virtual machines can easily have more than 16 option ROMs and
when that happens, we happily access our resource array out of
bounds. Make sure we stay within the MAX_ROMS limit.
While here, bump MAX_ROMS from 16 to 32 to minimize the chance
of leaving option ROMs unaccounted for.