The TLS segment is a little wierd in that it grows backwards from percpu_base
instead of forwards.  This causes the alignment code to calculate wrong
offsets when the segment size is 8 (mod 16).  A failure was seen where
::percpu_base was set at offset 0xfffffffffffffa08 in code that was in the
same translation unit as ::percpu_base, and 0xfffffffffffffa10 elsewhere.
This caused all dynamic_percpu instances to crash.

Fix by aligning the segment size.  For good measure, align also the segment
base, both to a cacheline boundary.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

When system classloader is used as a parent some of its protected
methods are called to lookup resources.

This also adds delegation for all remaining protected and public
methods.

Signed-off-by: Tomasz Grabiec <tgrabiec@cloudius-systems.com>

Signed-off-by: Tomasz Grabiec <tgrabiec@cloudius-systems.com>

Signed-off-by: Tomasz Grabiec <tgrabiec@cloudius-systems.com>

Signed-off-by: Tomasz Grabiec <tgrabiec@cloudius-systems.com>

Fix 1. Renaming existing directory to existing file when the
destination path has trailing slash should fail:

  rename("/tmp/dir", "/tmp/file/") == -1 && errno == ENOTDIR

Fix 2. Renaming exisitng directory to nonexisting path when the path
has trailing slash should succeed:

  rename("/tmp/dir", "/tmp/im_missing/") == 0

Fix 3. Renaming exisitng file to nonexisting path when the path
has trailing slash should fail:

  rename("/tmp/file", "/tmp/im_missing/") == -1 && errno == ENOTDIR

Reference:
http://pubs.opengroup.org/onlinepubs/9699919799/functions/rename.html



This change also adds a bunch of test cases for various conditions
mentioned in the spec. Note that for some cases where the raname()
should fail there are discrepancies between Linux and OSv in error
codes set by the call. To my understanding, the spec is ambiguous.
There are conditions which match more than one error code
description. The tests are therefore allowing the call to set error
code from a set of matching error codes (eg. both EEXIST and ENOTEMPTY
are allowed when applicable). If we decide that OSv should return
exactly the same error codes as Linux for the same conditions, this
can be changed later.

vfs: rename() should fail if paths differ only with trailing slash

This should fail:

  rename("/tmp/file", "/tmp/file/")

Signed-off-by: Tomasz Grabiec <tgrabiec@cloudius-systems.com>

The call to namei("/dir/file/") currently fails with ENOENT when
"/dir/file" exists. A more standard way is to return ENOTDIR
instead. This way calls to stat, open, rename, etc. will be in
line with the POSIX spec.

It is also useful to rename() implementation which needs to
differentiate behaviour between the case in which target does not
exist and the case in which it does but the path has trailing slash
and the last component is not a directory.

In addition to that the check was performed in an inconsistent matter
- only when dentry lookup failed. This change makes the check
performed always.

Signed-off-by: Tomasz Grabiec <tgrabiec@cloudius-systems.com>

Renaming to destination which is a substring of the source path
but not its parent should succeed:

  rename("/tmp/one_two", "/tmp/one") == 0

Renaming to a child should fail:

  rename("/tmp/", "/tmp/child") == -1 && errno = EINVAL

Signed-off-by: Tomasz Grabiec <tgrabiec@cloudius-systems.com>

When converting a source file from .c to .cc, make will complain that the
old source file was missing.  This is annoying, especially when bisecting
or compile-testing a patch set.

Fix by removing the source file dependency.  Since the dependency is alread
specified explicitly by the makefile, no information is lost.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

The sched::timer set() function expects an absolute time value. Fix an
incorret use of it AcpiOsWaitSemaphore().

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Signed-off-by: Tomasz Grabiec <tgrabiec@cloudius-systems.com>

The expression *p >> 24, where p is an unsigned*, is optimized by
the compiler to *((u8*)p + 3) - reading the most significant byte only and
dropping the shift.

When this optimization is applied to reading the APIC ID, QEMU
returns zero for all processors, since the manual requires reading
entire words.

Fix by using a volatile pointer, disabling the optimization.

Note that QEMU is technically correct here though it violates all known
real x86 implementations.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

xapic::init_ipi() shifts apic_id by 24, unaware that xapic::ipi will do it
again.  The result is that the boot processor is reset instead of the
auxiliary processor.

Remove the extraneous shift.

Found by booting with QEMU without kvm.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

Spotted by GCC with '-Wformat=1':

../../drivers/xenfront-xenbus.cc: In function ‘XenbusState xenbus_get_state(device_t)’:
../../drivers/xenfront-xenbus.cc:219:80: warning: format ‘%d’ expects argument of type ‘int*’, but argument 6 has type ‘XenbusState* {aka xenbus_state*}’ [-Wformat=]
     xs_scanf(XST_NIL, dev->get_node_path().c_str(), "state", NULL, "%d", &state);

Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Spotted by GCC with '-Wformat=1':

../../bsd/sys/netinet/tcp_subr.c: In function ‘tcp_log_addr’:
../../bsd/sys/netinet/tcp_subr.c:2286:3: warning: unknown conversion type character ‘b’ in format [-Wformat=]
   sprintf(sp, " tcpflags 0x%b", th->th_flags, PRINT_TH_FLAGS);
   ^
../../bsd/sys/netinet/tcp_subr.c:2286:3: warning: too many arguments for format [-Wformat-extra-args]

BSD has some nice printf extension for printing out bitfields. We don't
so switch to hexadecimal output.

Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Add 'werror' option to Makefile that can be used to disable the -Werror
GCC option temporarily as follows:

  make werror=0

This is useful when fixing up build breakage caused by addition of new
default compiler warnings.

Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Spotted by Clang:

In file included from ../../loader.cc:27:
In file included from ../../drivers/virtio-net.hh:12:
In file included from ../../bsd/sys/net/if_var.h:80:
In file included from ../../bsd/sys/sys/mbuf.h:40:
In file included from ../../bsd/porting/uma_stub.h:161:
../../include/osv/percpu.hh:35:42: error: arithmetic on a pointer to void
        return reinterpret_cast<T*>(base + offset);
                                    ~~~~ ^

Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

smp doesn't currently work on non-accelerated qemu, and doesn't help anyway
during mkfs.  Restrict ourselves to 1 vcpu to avoid problems.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

This patch fixes regression in d2397b7e.
dlopen() must return NULL in case of error, pointer to shared_ptr holding NULL
returned instead.

Signed-off-by: Dmitry <dmitry@daynix.com>

Follow section 3.1.5 of RFC2131.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

The new behavior wait for the DHCP_ACK to be received by OSv to configure the
network parameters.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

Commit 7fc023e8 fixed issue #64, and
made unhandled exceptions work correctly.

This patch adds a test to tst-except.cc to test this - that an unhandled
exception really does call the custom termination handler that the C++
standard allows to set (before the above commit, this test would hang).

Unfortunately, gcc's libsupc++ tries very hard to ensure that the
terminination handler really does abort - if the handler returns, or
throws an exception, the library nevertheless aborts. So we need to
trick it with an ugly longjmp to make this test be able to complete
without aborting the system.

Signed-off-by: Nadav Har'El <nyh@cloudius-systems.com>

Some distributions have enabled gcc -Wformat-security by default.
This causes the build to raise format warnings even though -Wformat=0
is used.

Setting no-format-security for build consistency between distros.

Signed-off-by: Or Cohen <orc@fewbytes.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Run 'dos2unix' on the sucker...

Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

C++ does not allow reinterpret_cast in constant expressions:

http://stackoverflow.com/questions/10369606/constexpr-pointer-value



GCC doesn't enforce it but Clang does.

Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

The current implementation was copying the whole file contents into an
array and then into stringstream which under the hood copied it again
into even bigger array. These allocations were causing memory
fragmentation which some times led to malloc failure.

The new approach is using fixed amount of memory. For each file it
wraps the cpio input stream in a restricted stream which limits the
amount of data that can be read from it so that it represents a view
for only that file's data. After the consumer has read it we simply
ignore the remaining padding from the cpio stream.

This fixes issue #66

Signed-off-by: Tomasz Grabiec <tgrabiec@cloudius-systems.com>

Sometimes nc picks ipv6, which isn't supported; force it to use ipv4.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

The initialization is done a this time so everything needed is ready.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

These are the two missing functions for ACPICA be able to initialize fully.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

Start to fill the missing ACPICA OSv primitives in order to be able to fully
initialize ACPICA for use in drivers probe functions.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

OpenSolaris version of strftime accepts (char *)0 as format, and translates it
to the locale default, which is the same as "%c" for strftime().

http://www.unix.com/man-page/opensolaris/3/strftime/



Signed-off-by: Or Cohen <orc@fewbytes.com>

Signed-off-by: Or Cohen <orc@fewbytes.com>

Signed-off-by: Or Cohen <orc@fewbytes.com>

For example, assuming the following file exists:

  /tmp/dir/x

..the following was sondiered absent:

  /tmp/dir/

..this was considered present:

  /tmp/dir

Signed-off-by: Tomasz Grabiec <tgrabiec@cloudius-systems.com>

This happens when you move a file to a different directory.

Thread 2 (Thread 2):
 1  0x000000000032ed3b in halt_no_interrupts ()
   at ../../arch/x64/processor.hh:239
   osv::halt () at ../../core/power.cc:28
 2  0x000000000020e7fe in abort (
    msg=msg@entry=0x580160 "page fault outside application")
    at ../../runtime.cc:109
 3  0x00000000002ecfdc in page_fault (ef=0xffffc0003e362008)
    at ../../core/mmu.cc:981
 4  <signal handler called>
 5  merge (b=0xffffc0003b4e8000, a=0xfffffffffffffff8)
    at ../../core/mempool.cc:421
 6  memory::free_page_range_locked (range=range@entry=0xffffc0003b4e8000)
    at ../../core/mempool.cc:436
 7  0x00000000003186d7 in memory::free_page_range (
    range=range@entry=0xffffc0003b4e8000) at ../../core/mempool.cc:448
 8  0x0000000000319280 in free_large (obj=0xffffc0003b4e9000)
    at ../../core/mempool.cc:460
 9  std_free (object=0xffffc0003b4e9000) at ../../core/mempool.cc:790
 10 free (obj=0xffffc0003b4e9000) at ../../core/mempool.cc:892
 11 0x0000000000340e52 in ramfs_free_node (np=np@entry=0xffffc0003d250d40)
    at ../../fs/ramfs/ramfs_vnops.c:78
 12 0x0000000000340ed2 in ramfs_remove_node (dnp=0xffffc0003d250a00,
    np=0xffffc0003d250d40) at ../../fs/ramfs/ramfs_vnops.c:131
 13 0x0000000000340ffd in ramfs_rename (dvp1=0xffffc0003fc9a180,
    vp1=0xffffc0003fc9a100, name1=<optimized out>, dvp2=0xffffc0003fc9a280,
    vp2=<optimized out>, name2=0x2000001fedf1 "prefix_suffix")
    at ../../fs/ramfs/ramfs_vnops.c:380
 14 0x0000000000340069 in sys_rename (src=<optimized out>,
    dest=<optimized out>) at ../../fs/vfs/vfs_syscalls.c:671
 15 0x000000000033a3b2 in rename (oldpath=<optimized out>,
    newpath=0xffffc0003e858cc0 "/tmp/t3-26e1b47c/prefix_suffix")
    at ../../fs/vfs/main.cc:634

Signed-off-by: Tomasz Grabiec <tgrabiec@cloudius-systems.com>

Spotted by Clang:

../../drivers/vga.hh:15:18: error: 'VGAConsole::write' hides overloaded virtual function
      [-Werror,-Woverloaded-virtual]
    virtual void write(const char *str);
                 ^
../../drivers/console.hh:16:18: note: hidden overloaded virtual function 'Console::write'
      declared here: different number of parameters (2 vs 1)
    virtual void write(const char *str, size_t len) = 0;
                 ^

Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>