Looks reasonable to me, but Amnon, can you please have a look and confirm
you didn't have a different way in mind to configure the http server to
different paths?
--
Nadav Har'El
n...@scylladb.com
On Thu, Jun 29, 2017 at 10:01 PM, Waldemar Kozaczuk
wrote:
> Here is the lis
On Thu, Jun 29, 2017 at 10:13 PM, Waldemar Kozaczuk
wrote:
> Signed-off-by: Waldemar Kozaczuk
> ---
> modules/httpserver-html5-cli/Makefile| 28
>
> modules/httpserver-html5-cli/httpserver.conf | 20
> modules/httpserver-html5-cli/module
On Thu, Jul 6, 2017 at 1:30 AM, Lonnie
wrote:
> Greetings All,
>
> I have just recently come across OSv and although I do realize that it is
> mainly designed as a minimalist container OS
>
Hi,
"container" is probably the wrong word to use here, because it has other
established meanings in this
redis even assume this when testing for IPv6 support.
So in this patch socket() differentiates between an unknown address family
(returning EAFNOSUPPORT) and a known address family but unknown socket type
or protocol (returning EPROTONOSUPPORT, as before).
Fixes #865.
Signed-off-by: Nadav Har&#x
On Sat, Jul 8, 2017 at 2:34 AM, Waldek Kozaczuk
wrote:
> I have been trying to use existing git repo -
> https://github.com/tari/osv-lighttpd - to deploy lighttpd on OSv. It
> builds but I am getting this error which seems to indicate that symbol
> pcre_free is missing or corresponding so could
On Sun, Jul 9, 2017 at 1:57 AM, Waldek Kozaczuk
wrote:
> Running with -V did not yield any interesting behavior. But when I built
> an image with httpserver hoping to be able to list content of the /usr/lib
> ditrectory (is there any easier way to verify content of built image apart
> from mount
at too includes libpcre.so.1, but not the
one in the build machine but rather the one from external/.
Signed-off-by: Nadav Har'El
---
lighttpd/module.py | 3 +++
lighttpd/Makefile| 48
lighttpd/lighttpd.conf | 13 ++
ok libpcre.so.1 from the build machine instead of external/
But still, this will not work together with libhttpserver (or
module/openssl in general) in the same image, because both try to bring a
different version of libpcre.so.
--
Nadav Har'El
n...@scylladb.com
On Sun, Jul 9, 2017 at 1
On Sat, Jul 8, 2017 at 2:25 AM, Waldek Kozaczuk
wrote:
> Please see my answers below.
>
>
>
>
>> Also, we already have a "node" module in apps, can we just "require" it
>> instead of adding more code to get node?
>> There's a difference - the existing node module fetches node's source and
>> comp
ined below.
>
> On Tue, Jul 4, 2017 at 4:09 PM, Nadav Har'El wrote:
>
>> Hi, sorry about the long delay in my review.
>>
>> In general I think your separation makes sense, although I'm not sure who
>> would actually want to use httpserver-core, our ad-hoc
Thanks. Looks good - I'll commit, none of the comments I have below are
important.
On Thu, Jun 15, 2017 at 1:31 PM, Justin Cinkelj
wrote:
> The iso-read from libcdio will be used to read ISO image
> (contextualization CD, cloud metadata image). This will allow use of
> standard ISO images in OSv
Unfortunately, your relink-binary.sh did not work for me:
./relink-binary.sh: line 14: libtool:: command not found
I'll prepare a patch to fix that.
--
Nadav Har'El
n...@scylladb.com
On Tue, Jul 11, 2017 at 3:28 PM, Nadav Har'El wrote:
> Thanks. Looks good - I'
On Thu, Jun 15, 2017 at 1:31 PM, Justin Cinkelj
wrote:
> In 56c6195b252b127c7fe90cb9fa46c370b384b51a was added option for
> reading configuration file from a secondary disk. Due to missing ISO
> filesystem support, a custom disk format was used. This patch replaces
> custom disk format with stand
that to 0x200.
Signed-off-by: Nadav Har'El
---
libc/mman.cc | 7 +++
1 file changed, 7 insertions(+)
diff --git a/libc/mman.cc b/libc/mman.cc
index 0441265..73a3f41 100644
--- a/libc/mman.cc
+++ b/libc/mman.cc
@@ -130,6 +130,13 @@ void *mmap(void *addr, size_t length, int prot, int
Add functions openat64() and creat64() as aliases to the usual functions.
Refs #747.
Signed-off-by: Nadav Har'El
---
fs/vfs/main.cc | 2 ++
include/api/fcntl.h | 3 ---
2 files changed, 2 insertions(+), 3 deletions(-)
diff --git a/fs/vfs/main.cc b/fs/vfs/main.cc
index 1e2408c..c2
hrow()).
Signed-off-by: Nadav Har'El
---
include/osv/sched.hh | 8
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/include/osv/sched.hh b/include/osv/sched.hh
index 3a192477..e9a1af96 100644
--- a/include/osv/sched.hh
+++ b/include/osv/sched.hh
@@ -1009,9 +1009,9 @@ void
Gcc 7 complains about defining memset when we don't include a header
file with its prototype. So let's do.
Signed-off-by: Nadav Har'El
---
fastlz/lzloader.cc | 1 +
1 file changed, 1 insertion(+)
diff --git a/fastlz/lzloader.cc b/fastlz/lzloader.cc
index 5f95e413..e05387e7 1006
Gcc 7 rightly complains when we try to allocate an array with a signed
int length. Changing the length variable to unsigned int fixes it.
Signed-off-by: Nadav Har'El
---
core/app.cc | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/core/app.cc b/core/app.cc
index d62
In bsd/sys/netinet/in_mcast we get a "maybe uninitialized" warning which
as far as I can tell, is a false alarm. I also checked the freebsd tree,
and they haven't fixed anything in this area, so it's probably correct.
So let's just ignore this warning.
Signed-off-by: Na
Gcc 7 rightly complains that in bsd_init() we create an uninitialized
timeval structure, and then use it to see the random number generator.
The intention was probably to fetch the current time here :-) So let's
do it.
Signed-off-by: Nadav Har'El
---
bsd/init.cc | 1 +
1 file
n't cause this
crash, even on gcc 7.
The patch trivially fixes JMP to CALL, which is what the callee's (main())
expects we would call to preserve the stack's alignment.
Signed-off-by: Nadav Har'El
---
arch/x64/boot.S | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff -
ertainly not a good reason to leave
this bug behind. Especially not when we know of bugs which one guess is
that they are caused by overwrite of the per-cpu area (see issue #382).
Signed-off-by: Nadav Har'El
---
arch/x64/boot.S | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --
In boot.S we reserved 40960 bytes for an "interrupt stack" that is not
actually used anywhere since almost the start of OSv (see commit
da6ba16dbd182517e4c889f806cd9ce57151). So get rid of it.
Signed-off-by: Nadav Har'El
---
arch/x64/boot.S | 4
1 file changed, 4 del
the future, its "new" fully supports any alignment
(and to do that, it has a version of "operator new" taking an alignment -
see http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3396.htm)
so we will be able to revert this patch.
Signed-off-by: Nadav Har'El
--
In one test we assigned a constant-expression-but-not-marked-constexpr
pointer into a constexpr variable, which gcc 7 didn't like. There is no
need to bother with constexpr qualifications in tests so let's just
remove it.
Signed-off-by: Nadav Har'El
---
tests/tst-mmap-file.cc |
The internal structure of std::unique_ptr changed, and we need it to support
the "osv info threads" gdb command. With this patch both old and new layouts
will be supported.
Signed-off-by: Nadav Har'El
---
scripts/loader.py | 6 +-
1 file changed, 5 insertions(+), 1 deletion(-
On Fri, Jul 14, 2017 at 12:02 PM, Justin Cinkelj
wrote:
> Based on @myechuri work,
> https://github.com/myechuri/osv-apps/commits/nginx-osv
Thanks (to both of you)!
I want to commit this patch (we can always improve it later), but have one
question first - does this patch depend on any of your
Oops, this is bad, and unfortunately I committed too many patches yesterday
to understand which one brought this bug...
On Sun, Jul 16, 2017 at 7:57 PM, wrote:
> TEST tst-elf-permissions.so OSv v0.24-399-ge6fadac
> eth0: 192.168.122.15
> Assertion failed: (phdr.p_flags & (PF_R |
On Mon, Jul 17, 2017 at 3:29 PM, Avi Kivity wrote:
>
>
> On 07/17/2017 10:10 AM, Nadav Har'El wrote:
>
> Oops, this is bad, and unfortunately I committed too many patches
> yesterday to understand which one brought this bug...
>
>
>
> On Sun, Jul 16, 2017 at 7
Thanks. One question below:
On Tue, Jul 18, 2017 at 12:16 PM, Justin Cinkelj
wrote:
> nginx doesn't start if ioctl with cmd=FIOASYNC fails. Stub it, so that
> nginx can start unpatched.
>
> Signed-off-by: Justin Cinkelj
> ---
> libc/af_local.cc | 18 +-
> 1 file changed, 13 ins
--
On Tue, Jul 18, 2017 at 12:50 PM, Justin Cinkelj
wrote:
> sigsuspend is used in nginx. With current (not yet commited) patch, it
> is never really called. Nginx master process code is used to execute
> worker code, and this prevents master code to actually call sigsuspend.
> So stub implementa
0x10c18733 in main (argc=, argv=)
at src/core/nginx.c:375
--
Nadav Har'El
n...@scylladb.com
On Tue, Jul 18, 2017 at 1:16 PM, Justin Cinkelj
wrote:
> Based on @myechuri work,
> https://github.com/myechuri/osv-apps/commits/nginx-osv
>
> A single nginx worker thread is
On Wed, Jul 19, 2017 at 3:31 AM, Justin Cinkelj
wrote:
> I also tried with/without cli, debug/release build, and did clean it
> before rebuild
> ./scripts/build -j4 mode=release clean
> ./scripts/build -j4 mode=release image=nginx
>
> And it still works
> justin_cinkelj@jcpc:~/devel/mikelangelo/o
This patch is missing the aligned_new.hh file with the actual (rather
simple) implementation.
I'll need to send the patch again.
--
Nadav Har'El
n...@scylladb.com
On Thu, Jul 13, 2017 at 12:46 AM, Nadav Har'El wrote:
> OSv has various types which have special alignment
though it's not really needed for nginx. but maybe for the multi-threaded
one it is needed?)
--
Nadav Har'El
n...@scylladb.com
On Tue, Jul 18, 2017 at 5:36 PM, Commit Bot
wrote:
> From: Justin Cinkelj
> Committer: Nadav Har'El
> Branch: master
>
> af_local:
But the patch leaves the line commented out - I don't suppose this can help
:-)
On Jul 20, 2017 15:08, "Justin Cinkelj" wrote:
At least a symbol needs to be defined for python2.7.
Signed-off-by: Justin Cinkelj
---
libc/libc.cc | 2 ++
1 file changed, 2 insertions(+)
diff --git a/libc/libc.cc
abled?
I don't think it is reasonable that this is a memory problem on my build
machine. I've been using this machine successfully for several years, and
don't recall any other suspicious SIGSEGV, and OSv was built and run
successfully on this machine many times before I upgraded t
On Sat, Jul 22, 2017 at 1:51 PM, Benoît Canet <
benoit.canet.cont...@gmail.com> wrote:
> To be complete misc.bin should be updated and
> openssl switched to version 1.1 in openssl manifest.
>
I committed the previous version of this patch, it looked the same right?
I think the openssl version di
n't really
care about what it did.
I'll open an issue on TCXONC
--
Nadav Har'El
n...@scylladb.com
On Fri, Jul 21, 2017 at 1:45 PM, Justin Cinkelj
wrote:
> At least a symbol needs to be defined for python2.7.
>
> Signed-off-by: Justin Cinkelj
> ---
> Makefile | 2 +
On Fri, Jul 21, 2017 at 1:45 PM, Justin Cinkelj
wrote:
> At least a symbol needs to be defined for python2.7.
> File mbrlen.c is moved from musl/ to libc/, as the weak alias declaration
> must be in same file as function definition.
>
I don't think this is true - the alias declaration does NOT n
On Sun, Jul 23, 2017 at 11:17 AM, Nadav Har'El wrote:
>
> On Fri, Jul 21, 2017 at 1:45 PM, Justin Cinkelj
> wrote:
>
>> At least a symbol needs to be defined for python2.7.
>> File mbrlen.c is moved from musl/ to libc/, as the weak alias declaration
>&g
On Sun, Jul 23, 2017 at 2:01 AM, Justin Cinkelj
wrote:
> Based on @gberginc work.
>
> Usage example:
> ./scripts/run.py -e "/python"
> ./scripts/run.py -e "/python -c \"aa={1:22,3:44}; print aa; print 'asdf'\""
>
Thanks. I'll commit this as is.
However, looking at python on my own machine (Fedo
On Sun, Jul 23, 2017 at 12:20 PM, Nadav Har'El wrote:
>
> By the way, when I try to build this myself, something at the end doesn't
> work:
>
> ./copy-required-system-libraries.sh
> ldd: ROOTFS/lib/python2.7/lib-dynload/readline.so: No such file or
> directory
&
On Sun, Jul 23, 2017 at 2:10 AM, Justin Cinkelj
wrote:
> I didn't want to include every it-might-be-actually-required library from
> the build system.
>
But did you, or did you not?
> Complete list from my system is:
> justin_cinkelj@jcpc:~/devel/mikelangelo/osv-fc25/osv/apps/python27$ ldd
> R
On Sat, Jul 22, 2017 at 6:58 AM, wrote:
> Hi all, I'm trying to port Aerospike server to OSv. But when I tried to
> run the server (asd.so), it failed to resolve the "je_chunksize_mask"
> symbol from libjemalloc.so.2 even when libjemalloc.so.2 was included into
> root folder. Anyone can give me a
On Sat, Jul 22, 2017 at 7:32 PM, Benoît Canet wrote:
>
> Maybe a better solution than this patch would be to avoid the mix of host
> openssl and
> internal openssl so this patch would be useless.
>
I agree that modules/openssl should take stuff from the build machine, not
from external/ - which
nction is ruined by using SSE
registers. So this patch adds a tag on this function telling gcc to not
use SSE when compiling this function.
P.S. This patch has to be a new world record. Zero lines of code,
and 50 lines of explanations :-)
Signed-off-by: Nadav Har'El
---
arch/x64/arch-cpu.
On Sat, Jul 22, 2017 at 1:51 PM, Benoît Canet <
benoit.canet.cont...@gmail.com> wrote:
> The LuaRock repository is not up to date.
>
Hi,
I am far from a Lua or LuaRock expert, but a quick look in
https://luarocks.org/modules/brunoos/luasec
shows me that they *do* have version 0.6-1, which appare
cations larger than a page are
page-aligned in any case).
Fixes #894.
Signed-off-by: Nadav Har'El
---
arch/x64/arch-switch.hh | 12 +---
core/elf.cc | 3 +++
2 files changed, 12 insertions(+), 3 deletions(-)
diff --git a/arch/x64/arch-switch.hh b/arch/x64/arch-switch
and sent patches (and I asked some questions about them).
--
Nadav Har'El
n...@scylladb.com
On Mon, Jul 24, 2017 at 11:05 AM, Nadav Har'El wrote:
> When the initial-exec TLS model is used, as in our tst-tls.so, the
> application's TLS area moves OSv's own TLS area b
On Mon, Jul 24, 2017 at 1:11 PM, Benoît Canet wrote:
>
>
> 2017-07-24 8:02 GMT+02:00 Nadav Har'El :
>
>>
>> On Sat, Jul 22, 2017 at 1:51 PM, Benoît Canet <
>> benoit.canet.cont...@gmail.com> wrote:
>>
>>> The LuaRock repository is no
com/cloudius-systems/osv/issues/821 (about running
applications compiled with different libraries than the OSv kernel)
Again, it doesn't mean we need to do this now. I'll leave it up to you to
decide.
>
> Best regards
>
> Benoît
>
> 2017-07-24 12:11 GMT+02:00 Benoît C
a/5.2/ssl.so: failed looking up symbol lsec_load_curves
--
Nadav Har'El
n...@scylladb.com
On Sat, Jul 22, 2017 at 1:51 PM, Benoît Canet <
benoit.canet.cont...@gmail.com> wrote:
> The LuaRock repository is not up to date.
> Use the github repository.
>
> Signed-of-by: Be
On Mon, Jul 24, 2017 at 3:27 PM, Nadav Har'El wrote:
> Hi, I committed this, but please continue to having module/openssl find
> and take the build machine's openssl, because as it stands, with your patch
> the default build (the CLI) no longer works even on my older Fedor
ra 26 you have both openssl 1.1 and 1.0
libraries. What we're playing with here is just the "devel" package (the
header files), only needed for compiling things.
Now the image built by default by "scripts/build" (httpserver and cli)
works well in Fedora 26 too.
--
Nada
On Tue, Jul 25, 2017 at 1:10 PM, Benoît Canet <
benoit.canet.cont...@gmail.com> wrote:
> If OpenSSL version is 1.1 prompt a message explaining
> to the user how to install the compat headers.
>
Thanks! Some requests below:
>
> Signed-off-by: Benoît Canet
> ---
> modules/lua/Makefile |
On Tue, Jul 25, 2017 at 2:51 PM, Benoît Canet <
benoit.canet.cont...@gmail.com> wrote:
> external/ is an old filesystem snapshot
> several years old.
> This means the openssl and crypto libraries
> in it are obsolettes.
>
> Get rid of this particular crypto dependencies
> by using the host librari
chine switched from the BFD
linker to GOLD and we changed the Makefile to revert a few failing tests to
BFD (but not tst-n). So I don't know yet if we have a serious regression,
or some environmental change (such as the linker) caused a new bug to
appear or an old bug to be exposed...
--
Nada
On Fri, Jul 28, 2017 at 8:34 PM, Benoît Canet wrote:
>
> Thanks
>
> I worked on creating this usr.manifest generator for openssl.
>
> Incidentally do you know if manifest file have a way to create symlinks ?
>
Yes, it does. You have two options:
1. Ask in the manifest to upload an actual file o
application threads the allocation will
*also* be a waste (most applications will not even use the syscall
instruction!), at least it will be a minor waste because application
threads normally have large stacks (by default 1MB) so the overhead of 4K
more will be negligible.
More comments inline, b
On Tue, Aug 1, 2017 at 2:16 PM, Justin Cinkelj
wrote:
> When we try to reuse python from the build system (instead of
> compiling it from source), libpython2.7.so requires additional
> libc/glibc symbols. The patch was prepared on Fedora 25.
>
> One problem with musl/src/linux/wait4.c remains. It
s the case.
> I didn't even open any of the .c files (except wait4.c, because it didn't
> compile).
> Will send v2 patch based on you comments.
> Thank you, Justin
>
>
> On 08/01/2017 02:23 PM, Nadav Har'El wrote:
>
>
> On Tue, Aug 1, 2017 at 2:16 PM,
On Tue, Aug 1, 2017 at 2:41 PM, Justin Cinkelj
wrote:
> Instead of compiling python from source, we reuse libpython2.7.so
> from the build system. Only the python executable is recompiled as
> shared object. The python.c file is almost literal copy from the git
> source (only 'include "Python.h"'
On Wed, Aug 2, 2017 at 7:18 AM, Hawx Chen wrote:
> Hi Nadav, thank you for your reply.
> I already finished the Patch V3 and then will update it soon. Then I want
> to summarize the tests and related modification in Patch V3.
>
> I tested the Patch in different ways:
> 1. I inserted "abort()" in
On Wed, Aug 2, 2017 at 9:30 AM, Benoît Canet wrote:
>
> simply porting the default hello world to OSv and loading it with wrk
> would help exibit the issue.
>
What default "hello world"?
Our apps/golang-example prints "hello world" (partly in Chinese...) but
cannot be "loaded with wrk".
I assume
on 3.6 in (stock) fedora 25.
> And with other/older distros, it will also be a bit problematic to build
> newer pythons.
>
Yes.
Perhaps we should have two apps, one for taking the system's existing
python and one from compiling from source like you previously had? Since
you already did t
On Wed, Aug 2, 2017 at 12:38 AM, Nadav Har'El wrote:
>
> On Tue, Aug 1, 2017 at 2:41 PM, Justin Cinkelj
> wrote:
>
>> Instead of compiling python from source, we reuse libpython2.7.so
>> from the build system. Only the python executable is recompiled as
>>
On Wed, Aug 9, 2017 at 2:47 PM, Waldemar Kozaczuk
wrote:
> This patch enhances OSv build scripts to allow exporting files in addition
> of uploading them to an image. It addresses 3 usage scenarios described in
> #900 and demostrated by examples below:
>
> - ./scripts/build image=empty export=al
On Mon, Aug 7, 2017 at 12:32 PM, Justin Cinkelj
wrote:
> Busybox ash shell is more powerful than OSv --runscript option.
> Use ash when fancy shell is required.
>
> While busybox config file includes default options (e.g. nearly all
> options are enabled), the usr.manifiest doesn't try to create
On Wed, Aug 16, 2017 at 12:45 PM, wrote:
>
> Thank you Waldek for trying to upstream this, I'm looking forward to be
> removing the current patches from the capstan-packages build platform in
> favor of using upstreamed code! :D
>
> Unfortunately I'm not very familiar with the scripts that we're
On Wed, Aug 16, 2017 at 1:33 PM, Nadav Har'El wrote:
>
> On Wed, Aug 16, 2017 at 12:45 PM, wrote:
>
>>
>> Thank you Waldek for trying to upstream this, I'm looking forward to be
>> removing the current patches from the capstan-packages build platform in
On Wed, Aug 16, 2017 at 2:12 PM, wrote:
> Perhaps you were right anyway - we are using some patches to
> upload_manifest, so we might have broken the stripping. So if I understand
> you right, the stripped cpiod (that is as small as 220 KB) should work the
> same as the unstripped one, or will th
On Thu, Aug 17, 2017 at 5:56 AM, Rick Payne (Offshore) wrote:
>
> Is there an equivalent to SO_BINDTODEVICE? I’m trying to get an
> implementation of VRRP working and the first issue is that the sending
> socket requires to be bound to the appropriate interface.
>
> After that, I’ll probably run
Thanks. I committed this patch.
One thing I didn't understand is the comment: "The only current limitation
is support for links in OSv". I see you did handle symlinks... Is this an
old comment?
--
Nadav Har'El
n...@scylladb.com
On Sat, Aug 19, 2017 at 7:51 PM, Waldemar Koza
On Sun, Aug 20, 2017 at 6:15 PM, Nadav Har'El wrote:
>
>
> 1. osv_execve() should recognize the case where the files it is trying to
> load doesn't have the executable magic header, and fail with ENOEXEC. This
> will hopefully allow ash to run scripts without a shebang
I'll commit this patch, but I have some minor questions below. You can send
a new incremental patch if you want to change something.
--
Nadav Har'El
n...@scylladb.com
On Fri, Aug 18, 2017 at 3:01 PM, Justin Cinkelj
wrote:
> On failure execve() returns -1 and sets errno. This pat
]: Leaving directory
'/home/nyh/osv.tmp2/osv/apps/busybox/build/busybox'
/usr/bin/ld: cannot find -lbusybox
collect2: error: ld returned 1 exit status
Any idea what might be wrong?
--
Nadav Har'El
n...@scylladb.com
On Fri, Aug 18, 2017 at 3:02 PM, Justin Cinkelj
wrote:
> 0
On Thu, Aug 17, 2017 at 10:54 PM, Rick Payne (Offshore) <
ri...@rossfell.co.uk> wrote:
>
> > On 17 Aug 2017, at 06:43, Rick Payne (Offshore)
> wrote:
> >
> > Does carp run on Freebsd or is it just OpenBSD? That has to solve a
> similar issue, namely sending announcements to a multicast address on
On Mon, Aug 21, 2017 at 7:01 PM, Justin Cinkelj
wrote:
> The strtok_r doesn't allow escaping the input parameters, which means
> each space is considered a separator. The boost::tokenizer supports
> escaping. This commit replaces strtok_r with boost::tokenizer. The
> final __argv is still continu
On Tue, Aug 22, 2017 at 2:48 PM, Justin Cinkelj
wrote:
> > What does it not fix?
> The $892 mentions environ varibales with spaces or empty one. And command
> line parsing is done in loader.cc to get loader options, and in
> commands.cc, to get argv/argc for app, and to implement runscript. This
On Sat, Aug 26, 2017 at 11:27 PM, Rick Payne (Offshore) <
ri...@rossfell.co.uk> wrote:
>
> > I saw that. I also noticed that multicast won’t work because the
> SIOCADDMULTI call does nothing in the hypervisor drivers (so the multicast
> mac addresses are never added to the filter tables for the in
On Mon, Aug 21, 2017 at 9:01 PM, Rick Payne (Offshore) wrote:
>
> I don’t know about vmxnet3 driver as I’m not familiar with the hypervisor
> that provides that. Can you point me at the code for the ‘other side’ for
> that?
>
vmxnet3 is VMWare's driver. I don't think we have the code for the oth
On Fri, Sep 1, 2017 at 7:08 PM, Waldek Kozaczuk
wrote:
> I changed
> int sendfile(int out_fd, int in_fd, off_t *_offset, size_t count)
> to
> ssize_t sendfile(int out_fd, int in_fd, off_t *_offset, size_t count)
>
> and all works like a charm.
>
Wow, thanks, great detective work. When you first
On Sun, Sep 3, 2017 at 10:26 AM, Nadav Har'El wrote:
>
> On Fri, Sep 1, 2017 at 7:08 PM, Waldek Kozaczuk
> wrote:
>
>> I changed
>> int sendfile(int out_fd, int in_fd, off_t *_offset, size_t count)
>> to
>> ssize_t sendfile(int out_fd, int in_fd, of
On Sun, Sep 3, 2017 at 10:55 AM, Commit Bot
wrote:
> From: Waldemar Kozaczuk
> Committer: Nadav Har'El
> Branch: master
>
> Waldek Kozaczuk discovered that nginx, using sendfile(), is experiencing
> various failures when multiple concurrent connections are attempted.
On Thu, Sep 7, 2017 at 8:50 PM, Waldek Kozaczuk
wrote:
> Is linuxkit from Docker (https://github.com/linuxkit/linuxkit) going to
> kill unikernels altogether? What do you think?
>
Hi,
This approach can definitely work - I have done something like this 17
years ago (even before virtualization),
On Thu, Sep 7, 2017 at 8:50 PM, Waldek Kozaczuk wrote:
> Is linuxkit from Docker (https://github.com/linuxkit/linuxkit) going to
>>> kill unikernels altogether? What do you think?
>>>
>>
>> Hi,
>>
>> This approach can definitely work - I have done something like this 17
>> years ago (even before
Thanks, I committed (sorry for the delay).
As usual when git submodules are involved, I had somewhat of a mess
applying this patch (needed to manually remove some old directories and
"git submodule update"), I hope users and jenkins will not be bit by this.
--
Nadav Har'El
n...@s
Hi, sorry, I didn't see this patch, and committed a different patch
(c8b1906cf59d73a7cb39926c53dc4e136611d4f5) at about the same time,
doing the same thing in a different way.
--
Nadav Har'El
n...@scylladb.com
On Tue, Aug 29, 2017 at 11:37 AM, Justin Cinkelj wrote:
> iperf failed
On Thu, Aug 24, 2017 at 8:29 PM, Justin Cinkelj wrote:
> memcpy had of-by-one error.
> Also, remove unneeded defines.
>
> Signed-off-by: Justin Cinkelj
> ---
> libc/string/stresep.c | 16 ++--
> 1 file changed, 2 insertions(+), 14 deletions(-)
>
> diff --git a/libc/string/stresep.c b
On Thu, Aug 24, 2017 at 8:29 PM, Justin Cinkelj wrote:
> loader_parse_cmdline accepts input str - OSv commandline, say:
> --env=AA=aa --env=BB=bb1\ bb2 app.so arg1 arg2
>
> The loader options are parsed and saved into argv, up to first
> not-loader-option
> token. argc is set to number of loader
On Thu, Aug 24, 2017 at 7:56 PM, Justin Cinkelj wrote:
> Parse loader options with loader_parse_cmdline instead of strtok_r.
> The cmdline passed in prepare_commands to osv::parse_command_line
> is not reassebled from already parsed argv anymore, as app_cmdline
> contains unmodifed commandline str
On Wed, Sep 13, 2017 at 3:05 PM, Justin Cinkelj wrote:
> Compared to v2 patch:
> The stresep.c file is now first only added, and included into Makefile
> after removing unneeded define and include namespace.h.
> The memcpy bug from v2 patch was reported upstrem to netbsd, and current file
> now in
lease test :-)
--
Nadav Har'El
n...@scylladb.com
On Wed, Sep 13, 2017 at 8:56 PM, Waldek Kozaczuk
wrote:
> Currently httpserver returns disk usage information in blocks like this
> snippet demonstrates:
>
> static void fill_dfstat(DFStat& dfstat, const osv::mount_desc&am
On Fri, Sep 15, 2017 at 3:58 PM, Benoît Canet <
benoit.canet.cont...@gmail.com> wrote:
> This one does compile on recent host.
>
Three weeks ago Justin noticed the same problem, and we fixed this by a
trivial fix to iperf's autoconf (commit
fd93d5b996720e60e617e894dccb028ef89e8ccb in apps.git).
I
rprised... I should do this too in one of my pet projects...)? Is it a
new fork by other developers? Or what?
> 2017-09-18 11:51 GMT+02:00 Nadav Har'El :
>
>> On Fri, Sep 15, 2017 at 3:58 PM, Benoît Canet <
>> benoit.canet.cont...@gmail.com> wrote:
>>
>>&g
On Mon, Sep 18, 2017 at 1:28 PM, Benoît Canet wrote:
>
> I think it's an enthousiast that package iperf for easy distribution.
>
> What worry me is that the official mirror
> https://github.com/esnet/iperf does not distribute version 2 that we need
> for osv.
>
I just noticed that http://source
On Mon, Sep 18, 2017 at 2:51 PM, Nadav Har'El wrote:
>
> On Mon, Sep 18, 2017 at 1:28 PM, Benoît Canet wrote:
>
>>
>> I think it's an enthousiast that package iperf for easy distribution.
>>
>> What worry me is that the official mirror
>> h
I found https://sourceforge.net/projects/iperf2/files/
--
Nadav Har'El
n...@scylladb.com
On Mon, Sep 18, 2017 at 2:53 PM, Nadav Har'El wrote:
>
> On Mon, Sep 18, 2017 at 2:51 PM, Nadav Har'El wrote:
>
>>
>> On Mon, Sep 18, 2017 at 1:28 PM, Benoît Canet
process the read data, even though it wasn't, and
crashed on this processing or on the brelse() at its end.
In this case, the read_partition_table() should just do nothing, and
certainly not crash.
Refs #903.
Signed-off-by: Nadav Har'El
---
fs/devfs/device.cc | 5 -
1 file
401 - 500 of 1373 matches
Mail list logo
