Fault in the buffer prior to writing to workaround poor performance due
to interaction with kernel fs deadlock avoidance code. See the comment
prior to vn_io_fault_doio() in sys/kern/vfs_vnops.c for details of the
issue.
On my stable/10 desktop with a 16MB obj.o and "ar r out.a obj.o" I see
the following run times (seconds):
x ar.r284891
+ ar.patched
+----------------------------------------------------------------------+
|+ |
|+ x|
|+ xx|
|A |A|
+----------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 3 1.307 1.321 1.315 1.3143333 0.0070237692
+ 3 0.020 0.023 0.022 0.021666667 0.0015275252
Difference at 95.0% confidence
-1.29267 +/- 0.0115203
-98.3515% +/- 0.876513%
(Student's t, pooled s = 0.00508265)
Thanks to kib for diagnosing and explaining the issue and suggesting
the workaround.
Reviewed by: eadler, kib
MFC after: 1 week
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D2933
restore the FPU state from the format of machine FSAVE area. The
intended use is for ABI emulators to provide FSAVE-formatted FPU state
to usermode requiring it, while kernel could use FXSAVE due to
XMM/XSAVE.
The core functionality to convert from/to FXSAVE format is shared with
the fill_fpregs_xmm() and set_fpregs_xmm(). Move the later functions
to npx.c and rename them to npx_fill_fpregs_xmm() and
npx_set_fpregs_xmm(). They differ from nptx_get/set_fsave(9) since
our mcontext contains padding to be zeroed or ignored.
fill_fpregs() and set_fpregs() could be converted to use the new
interface, but there are small differences to handle.
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
obtain the thread %fs and %gs bases. Add x86 PT_SETFSBASE and
PT_SETGSBASE requests to set the bases from debuggers. The set
requests, similarly to the sysarch({I386,AMD64}_SET_FSBASE),
override the corresponding segment registers.
The main purpose of the operations is to retrieve and modify the tcb
address for debuggee.
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
The sentinel attribute was originally implemented in OpenBSD's gcc and
later adopted by upstream GCC 4.0 (and clang). From the OpenBSD's
gcc-local manpage:
- gcc recognizes the extra attribute __sentinel__, which can be used to
mark varargs function that need a NULL pointer to mark argument
termination, like execl(3). This exposes latent bugs for 64-bit
architectures, where a terminating 0 will expand to a 32-bit int, and
not a full-fledged 64-bits pointer.
While here sort the visibility attributes.
Hinted-by: OpenBSD
The message text is from cp, which has had a nicer message for this since
2007 (PR bin/50656).
As with cp, the exit status changes from 64 to 1.
PR: 201083
MFC after: 1 week
adding macros to define class lists.
This change is backwards compatible for all use within C and C++
programs. Only C++ programs will have added support to use the queue
macros within classes. Previously the queue macros could only be used
within structures.
The queue.3 manual page has been updated to describe the new
functionality and some alphabetic sorting has been done while
at it.
Differential Revision: https://reviews.freebsd.org/D2745
PR: 200827 (exp-run)
MFC after: 2 weeks
The code path to support units conversions from the command line
need not initialize neither libedit nor the history. Therefore, only do
that when in interactive mode.
This hides the issue reported in PR bin/201167 whereby running commands
of the form 'echo "$(units ft in)"' would corrupt the terminal. The real
issue causing the corruption most likely still remains somewhere.
PR: bin/201167
Differential Revision: D2935
Reviewed by: eadler
atomic_load_acq(9), on it source, for x86.
Right now, atomic_load_acq() on x86 is sequentially consistent with
other atomics, code ensures this by doing store/load barrier by
performing locked nop on the source. Provide separate primitive
__storeload_barrier(), which is implemented as the locked nop done on
a cpu-private variable, and put __storeload_barrier() before load, to
keep seq_cst semantic but avoid introducing false dependency on the
no-modification of the source for its later use.
Note that seq_cst property of x86 atomic_load_acq() is not documented
and not carried by atomics implementations on other architectures,
although some kernel code relies on the behaviour. This commit does
not intend to change this.
Reviewed by: alc
Discussed with: bde
Tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
This makes sysroot usable for cross building, it also removes the need for
_SHLIBDIRPREFIX (keeps its definition since picobsd uses it and I have no time
to test it)
Differential Revision: https://reviews.freebsd.org/D2920
Submitted by: imp, adrian
Tested by: adrian
VAGRANT_PROVIDERS variable. Right now, it defaults to only
vmware_desktop, virtualbox support is to follow at some point.
While here, fix the hashicorp URL: s/vagrant/atlas/, which
was result of a sed(1) replace (and my fault).
Sponsored by: The FreeBSD Foundation
VAGRANT_${VAR} variables extracted from VAGRANT_UPLOAD_CONF.
Set ATLAS_${VAR} to VAGRANT_${VAR} if VAGRANT_UPLOAD_CONF
is set. There is intent to intentionally have separate
variants of configuration entries, but the defaults do not
yet have any reason to be different.
Sponsored by: The FreeBSD Foundation
According to the GCC documentation:
"This attribute is available in GCC 4.1.3 and later. It is available
if either of the preprocessor macros __GNUC_GNU_INLINE__ or
__GNUC_STDC_INLINE__ are defined."
We don't keep the gcc granularity up to the minor number so it's
better to use the documented way. Current clang defines both
macros.
Reference:
https://gcc.gnu.org/onlinedocs/gcc/Common-Function-Attributes.html#Common-Function-Attributes
offload support was introduced in r284746.
While here also fix the ioctl() handler for IPv4 added in r279819,
which was never compiled in given opt_inet.h was not included.
First, on the write error, bufdone() call from ffs_backgroundwrite()
panics because pbrelvp() cleared bp->b_bufobj, while brelse() would
try to re-dirty the copy of the cg buffer. Handle this by setting
B_INVAL for the case of BIO_ERROR.
Second, we must re-dirty the real buffer containing the cylinder group
block data when background write failed. Real cg buffer was already
marked clean in ffs_bufwrite(). After the BV_BKGRDINPROG flag is
cleared on the real cg buffer in ffs_backgroundwrite(), buffer scan
may reuse the buffer at any moment. The result is lost write, and if
the write error was only transient, we get corrupted bitmaps.
We cannot re-dirty the original cg buffer in the
ffs_backgroundwritedone(), since the context is not sleepable,
preventing us from sleeping for origbp' lock. Add BV_BKGDERR flag
(protected by the buffer object lock), which is converted into delayed
write by brelse(), bqrelse() and buffer scan.
In collaboration with: Conrad Meyer <cse.cem@gmail.com>
Reviewed by: mckusick
Sponsored by: The FreeBSD Foundation (kib),
EMC/Isilon storage division (Conrad)
MFC after: 2 weeks
machine images to the Google Compute Engine platform.
By default, gcutil/gsutil requires an Oauth2 login generated
from a URL that must be opened in a browser, a verification
code copied back to the terminal from which it was invoked,
etc., etc., making it near impossible for automation.
I've hacked together an evil solution to work around this,
so unless GCE_LOGIN_SKIP is set to a non-empty value, this
Makefile will not do anything useful.
As a result of this commit, remove the gce-package.sh script
that was never, nor will ever be, used.
MFC after: 3 days
X-MFC-Note: (hopefully)
Sponsored by: The FreeBSD Foundation
When requesting install(1) to only make relative symlinks, by pass all the
done to actually compute the relative symlink if the path given in argument is
already a relative path
- Add prototype for boot2 main()
- Don't make assignment within if statement, split it into two.
No functional or binary change -- verified with sha256(1).
MFC after: 2 weeks
Switch from make_dev_alias to make_dev_alias_p since make_dev_alias_p can
gracefully fail if the /dev/megaraid_sas_ioctl_node symlink already exists.
This can happen if mfi(4) and mrsas(4) are both attached to cards and
providing Linux emulation support. Let the first one win. An equivalent
change has been done in mfi(4). Extra credit would be to pass the
Linux emulation call to the other driver when appropriate. This will
probably be a rare case and the user can manually change where the symlink
points to.
MFC after: 3 days
and there are some pending IOs at the time of OCR. This is mainly because of
recursive mutext in OCR and IO completion function call. Generic IO completion (from ISR) needs
sim_lock to be held before it calls completion to CAM (xpt_done), but in case of OCR path mrsas_ocr thread
itself take sim_lock, so this condition is now handled in this patch.
MFC after: 3 days
The current linker script generates program headers with VMA == LMA:
Entry point 0xffffffff802e7000
There are 6 program headers, starting at offset 64
Program Headers:
Type Offset VirtAddr PhysAddr
FileSiz MemSiz Flags Align
PHDR 0x0000000000000040 0xffffffff80200040 0xffffffff80200040
0x0000000000000150 0x0000000000000150 R E 8
INTERP 0x0000000000000190 0xffffffff80200190 0xffffffff80200190
0x000000000000000d 0x000000000000000d R 1
[Requesting program interpreter: /red/herring]
LOAD 0x0000000000000000 0xffffffff80200000 0xffffffff80200000
0x00000000010559b0 0x00000000010559b0 R E 200000
LOAD 0x0000000001056000 0xffffffff81456000 0xffffffff81456000
0x0000000000132638 0x000000000052ecf8 RW 200000
DYNAMIC 0x0000000001056000 0xffffffff81456000 0xffffffff81456000
0x00000000000000d0 0x00000000000000d0 RW 8
GNU_STACK 0x0000000000000000 0x0000000000000000 0x0000000000000000
0x0000000000000000 0x0000000000000000 RWE 8
This is fine for the FreeBSD loader, because it completely ignores p_paddr
and instead uses p_vaddr with a hardcoded offset. Other loaders however
acknowledge p_paddr (like the Xen ELF loader), in which case they will try
to load the kernel at the wrong place. Fix this by adding an AT keyword to
the first section specifying the physical address, other sections will
follow suit, so it ends up looking like:
Entry point 0xffffffff802e7000
There are 6 program headers, starting at offset 64
Program Headers:
Type Offset VirtAddr PhysAddr
FileSiz MemSiz Flags Align
PHDR 0x0000000000000040 0xffffffff80200040 0x0000000000200040
0x0000000000000150 0x0000000000000150 R E 8
INTERP 0x0000000000000190 0xffffffff80200190 0x0000000000200190
0x000000000000000d 0x000000000000000d R 1
[Requesting program interpreter: /red/herring]
LOAD 0x0000000000000000 0xffffffff80200000 0x0000000000200000
0x00000000010559b0 0x00000000010559b0 R E 200000
LOAD 0x0000000001056000 0xffffffff81456000 0x0000000001456000
0x0000000000132638 0x000000000052ecf8 RW 200000
DYNAMIC 0x0000000001056000 0xffffffff81456000 0x0000000001456000
0x00000000000000d0 0x00000000000000d0 RW 8
GNU_STACK 0x0000000000000000 0x0000000000000000 0x0000000000000000
0x0000000000000000 0x0000000000000000 RWE 8
Tested on bare metal using the native FreeBSD loader and grub2 from TRUEOS.
Sponsored by: Citrix Systems R&D
Reviewed by: kib
Differential Revision: https://reviews.freebsd.org/D2783
and related data structures. Contexts attach requests initiators to
domains. There is still 1:1 correspondence between contexts and
domains on the running system, since only busdma currently allocates
them, using dmar_get_ctx_for_dev().
Large part of the change is formal rename of the ctx to domain, but
patch also reworks the context allocation and free to allow for
independent domain creation.
The helper dmar_move_ctx_to_domain() is introduced for future use, to
reassign request initiator from one domain to another. The hard issue
which is not yet resolved with the context move is proper handling (or
reserving) RMRR entries in the destination domain as required by ACPI
DMAR table for moved context.
Tested by: pho
Sponsored by: The FreeBSD Foundation
CPU set operations in my upcoming NUMA work.
Tested/compiled:
* i386 (run)
* amd64 (run)
* mips (run)
* mips64 (run)
* armv6 (built)
Sponsored by: Norse Corp, Inc.
