mgmt: update to latest

Signed-off-by: Pekka Enberg <[email protected]>

Fix empty-path case of futimesat()

The Linux futimesat(2) specifies that:

  Glibc notes: If pathname is  NULL, then the glibc futimesat() wrapper
  function updates the times for the file referred to by dirfd.

In other words, when pathname==NULL, futimesat() behaves like futimes(),
at this works whether or not the given fd is a directory.

We did not support this corner case, and it turns out that OpenJDK
uses exactly this case: futimesat(fd, NULL, times), so not supporting
this resulted in weird "Not a directory" message when trying to change
the times of a non-directory file.

Since futimes(fd, times) is equivalent to the newly fixed
futimesat(fd, NULL, times), we also implement the former function (missing
until now) by calling the latter.

Fixes cloudius-systems#425 (Lyor, please confirm).

This patch also adds tests for this case.

Signed-off-by: Nadav Har'El <[email protected]>
Signed-off-by: Pekka Enberg <[email protected]>

release-ec2: introduce AWS region/zone/placement group parameters

These parameters define created AMIs/instances placement,
useful for performance tests where instances locality matter.

Signed-off-by: Dmitry Fleytman <[email protected]>
Signed-off-by: Pekka Enberg <[email protected]>

libosv: add padding after the content of the note

Both name and desc fields should include a padding so that their size
is a multiple of 4 (fields namesz and descsz contain size of the respective
fields without that padding, though).

With this patch readelf -n no longer complains about corrupted notes.

Signed-off-by: Paweł Dziepak <[email protected]>
Signed-off-by: Pekka Enberg <[email protected]>

Revert "build.mk: move modules/tests/usr.manifest to build/"

This reverts commit 7555bca.

Jaspal writes:

  This commit breaks 'make check' because the location of
  modules/tests/usr.manifest as changed and module.py needs to be
  updated.

Signed-off-by: Pekka Enberg <[email protected]>

tests: fix unsigned/signed comparison in misc-free-perf.cc

thread_allocator::next_core is compared against sched::cpus.size().
GCC 4.9 doesn't like the fact that the former is signed and the latter
is unsigned.

Reviewed-by: Tomasz Grabiec <[email protected]>
Signed-off-by: Paweł Dziepak <[email protected]>
Signed-off-by: Pekka Enberg <[email protected]>

mgmt: update to latest

Signed-off-by: Pekka Enberg <[email protected]>

tests: add test for closing TCP connection with pending packets

The test is supposed to trigger the problem from issue cloudius-systems#259. I was not
able to trigger the problem using guest-local communication, hence the
client is external to the guest.

Reviewed-by: Nadav Har'El <[email protected]>
Signed-off-by: Tomasz Grabiec <[email protected]>
Signed-off-by: Pekka Enberg <[email protected]>

Revert "build: Skip rebuild of usr.img on osv.vmdk and osv.vdi target"

This reverts commit df30ec8. It breaks
building "osv.vdi" altogether.

Signed-off-by: Pekka Enberg <[email protected]>

README.md: the run.py default for OSv Guest requested RAM size is 2GB…

… RAM, not 1GB

Reviewed-by: Tomasz Grabiec <[email protected]>
Signed-off-by: Vlad Zolotarov <[email protected]>
Signed-off-by: Pekka Enberg <[email protected]>