* [PATCH 1/4] push options: {pre,post}-receive hook learns about push options
2016-07-09 0:31 [PATCHv4 0/4] Push options Stefan Beller
@ 2016-07-09 0:31 ` Stefan Beller
2016-07-09 0:31 ` [PATCH 2/4] receive-pack: implement advertising and receiving " Stefan Beller
` (2 subsequent siblings)
3 siblings, 0 replies; 9+ messages in thread
From: Stefan Beller @ 2016-07-09 0:31 UTC (permalink / raw)
To: gitster; +Cc: git, e, peff, dwwang, dennis, jrnieder, Stefan Beller
The environment variable GIT_PUSH_OPTION_COUNT is set to the number of
push options sent, and GIT_PUSH_OPTION_{0,1,..} is set to the transmitted
option.
The code is not executed as the push options are set to NULL, nor is the
new capability advertised.
There was some discussion back and forth how to present these push options
to the user as there are some ways to do it:
Keep all options in one environment variable
============================================
+ easiest way to implement in Git
- This would make things hard to parse correctly in the hook.
Put the options in files instead,
filenames are in GIT_PUSH_OPTION_FILES
======================================
+ After a discussion about environment variables and shells, we may not
want to put user data into an environment variable (see [1] for example).
+ We could transmit binaries, i.e. we're not bound to C strings as
we are when using environment variables to the user.
+ Maybe easier to parse than constructing environment variable names
GIT_PUSH_OPTION_{0,1,..} yourself
- cleanup of the temporary files is hard to do reliably
- we have race conditions with multiple clients pushing, hence we'd need
to use mkstemp. That's not too bad, but still.
Use environment variables, but restrict to key/value pairs
==========================================================
(When the user pushes a push option `foo=bar`, we'd
GIT_PUSH_OPTION_foo=bar)
+ very easy to parse for a simple model of push options
- it's not sufficient for more elaborate models, e.g.
it doesn't allow doubles (e.g. cc=reviewer@email)
Present the options in different environment variables
======================================================
(This is implemented)
* harder to parse as a user, but we have a sample hook for that.
- doesn't allow binary files
+ allows the same option twice, i.e. is not restrictive about
options, except for binary files.
+ doesn't clutter a remote directory with (possibly stale)
temporary files
As we first want to focus on getting simple strings to work
reliably, we go with the last option for now. If we want to
do transmission of binaries later, we can just attach a
'side-channel', e.g. "any push option that contains a '\0' is
put into a file instead of the environment variable and we'd
have new GIT_PUSH_OPTION_FILES, GIT_PUSH_OPTION_FILENAME_{0,1,..}
environment variables".
We limit the push options for now
* to not exceed an arbitrary count, and
* to not exceed an arbitrary size.
This serves two purposes:
* DoS protection (i.e. one connection can add no more than 32kB
now)
* We need to figure out how to handle large (>64kB). Jeff wrote:
> Yes, but people are also happy when they can use a flexible and
> standardized tool to do a thing. I'd be more frustrated when I found out
> that Git's data-pushing protocol has arbitrary limitations (like, say, I
> can't push a data item larger than a single 64K pkt-line), which would
> easily just work with something like HTTP POSTs.
So to keep a way open in the future to deal with large pay loads,
the size is restricted for now.
[1] 'Shellshock' https://lwn.net/Articles/614218/
Signed-off-by: Stefan Beller <sbeller@google.com>
---
Documentation/githooks.txt | 18 ++++++++++++++
builtin/receive-pack.c | 49 +++++++++++++++++++++++++++----------
templates/hooks--pre-receive.sample | 24 ++++++++++++++++++
3 files changed, 78 insertions(+), 13 deletions(-)
create mode 100644 templates/hooks--pre-receive.sample
diff --git a/Documentation/githooks.txt b/Documentation/githooks.txt
index d82e912..9565dc3 100644
--- a/Documentation/githooks.txt
+++ b/Documentation/githooks.txt
@@ -247,6 +247,15 @@ Both standard output and standard error output are forwarded to
'git send-pack' on the other end, so you can simply `echo` messages
for the user.
+The number of push options given on the command line of
+`git push --push-option=...` can be read from the environment
+variable `GIT_PUSH_OPTION_COUNT`, and the options themselves are
+found in `GIT_PUSH_OPTION_0`, `GIT_PUSH_OPTION_1`,...
+If it is negotiated to not use the push options phase, the
+environment variables will not be set. If the client selects
+to use push options, but doesn't transmit any, the count variable
+will be set to zero, `GIT_PUSH_OPTION_COUNT=0`.
+
[[update]]
update
~~~~~~
@@ -322,6 +331,15 @@ a sample script `post-receive-email` provided in the `contrib/hooks`
directory in Git distribution, which implements sending commit
emails.
+The number of push options given on the command line of
+`git push --push-option=...` can be read from the environment
+variable `GIT_PUSH_OPTION_COUNT`, and the options themselves are
+found in `GIT_PUSH_OPTION_0`, `GIT_PUSH_OPTION_1`,...
+If it is negotiated to not use the push options phase, the
+environment variables will not be set. If the client selects
+to use push options, but doesn't transmit any, the count variable
+will be set to zero, `GIT_PUSH_OPTION_COUNT=0`.
+
[[post-update]]
post-update
~~~~~~~~~~~
diff --git a/builtin/receive-pack.c b/builtin/receive-pack.c
index 15c323a..236417e 100644
--- a/builtin/receive-pack.c
+++ b/builtin/receive-pack.c
@@ -550,8 +550,16 @@ static void prepare_push_cert_sha1(struct child_process *proc)
}
}
+struct receive_hook_feed_state {
+ struct command *cmd;
+ int skip_broken;
+ struct strbuf buf;
+ const struct string_list *push_options;
+};
+
typedef int (*feed_fn)(void *, const char **, size_t *);
-static int run_and_feed_hook(const char *hook_name, feed_fn feed, void *feed_state)
+static int run_and_feed_hook(const char *hook_name, feed_fn feed,
+ struct receive_hook_feed_state *feed_state)
{
struct child_process proc = CHILD_PROCESS_INIT;
struct async muxer;
@@ -567,6 +575,16 @@ static int run_and_feed_hook(const char *hook_name, feed_fn feed, void *feed_sta
proc.argv = argv;
proc.in = -1;
proc.stdout_to_stderr = 1;
+ if (feed_state->push_options) {
+ int i;
+ for (i = 0; i < feed_state->push_options->nr; i++)
+ argv_array_pushf(&proc.env_array,
+ "GIT_PUSH_OPTION_%d=%s", i,
+ feed_state->push_options->items[i].string);
+ argv_array_pushf(&proc.env_array, "GIT_PUSH_OPTION_COUNT=%d",
+ feed_state->push_options->nr);
+ } else
+ argv_array_pushf(&proc.env_array, "GIT_PUSH_OPTION_COUNT");
if (use_sideband) {
memset(&muxer, 0, sizeof(muxer));
@@ -606,12 +624,6 @@ static int run_and_feed_hook(const char *hook_name, feed_fn feed, void *feed_sta
return finish_command(&proc);
}
-struct receive_hook_feed_state {
- struct command *cmd;
- int skip_broken;
- struct strbuf buf;
-};
-
static int feed_receive_hook(void *state_, const char **bufp, size_t *sizep)
{
struct receive_hook_feed_state *state = state_;
@@ -634,8 +646,10 @@ static int feed_receive_hook(void *state_, const char **bufp, size_t *sizep)
return 0;
}
-static int run_receive_hook(struct command *commands, const char *hook_name,
- int skip_broken)
+static int run_receive_hook(struct command *commands,
+ const char *hook_name,
+ int skip_broken,
+ const struct string_list *push_options)
{
struct receive_hook_feed_state state;
int status;
@@ -646,6 +660,7 @@ static int run_receive_hook(struct command *commands, const char *hook_name,
if (feed_receive_hook(&state, NULL, NULL))
return 0;
state.cmd = commands;
+ state.push_options = push_options;
status = run_and_feed_hook(hook_name, feed_receive_hook, &state);
strbuf_release(&state.buf);
return status;
@@ -1316,7 +1331,8 @@ cleanup:
static void execute_commands(struct command *commands,
const char *unpacker_error,
- struct shallow_info *si)
+ struct shallow_info *si,
+ const struct string_list *push_options)
{
struct command *cmd;
unsigned char sha1[20];
@@ -1335,7 +1351,7 @@ static void execute_commands(struct command *commands,
reject_updates_to_hidden(commands);
- if (run_receive_hook(commands, "pre-receive", 0)) {
+ if (run_receive_hook(commands, "pre-receive", 0, push_options)) {
for (cmd = commands; cmd; cmd = cmd->next) {
if (!cmd->error_string)
cmd->error_string = "pre-receive hook declined";
@@ -1756,6 +1772,7 @@ int cmd_receive_pack(int argc, const char **argv, const char *prefix)
if ((commands = read_head_info(&shallow)) != NULL) {
const char *unpack_status = NULL;
+ struct string_list *push_options = NULL;
prepare_shallow_info(&si, &shallow);
if (!si.nr_ours && !si.nr_theirs)
@@ -1764,13 +1781,19 @@ int cmd_receive_pack(int argc, const char **argv, const char *prefix)
unpack_status = unpack_with_sideband(&si);
update_shallow_info(commands, &si, &ref);
}
- execute_commands(commands, unpack_status, &si);
+ execute_commands(commands, unpack_status, &si,
+ push_options);
if (pack_lockfile)
unlink_or_warn(pack_lockfile);
if (report_status)
report(commands, unpack_status);
- run_receive_hook(commands, "post-receive", 1);
+ run_receive_hook(commands, "post-receive", 1,
+ push_options);
run_update_post_hook(commands);
+ if (push_options) {
+ string_list_clear(push_options, 0);
+ free(push_options);
+ }
if (auto_gc) {
const char *argv_gc_auto[] = {
"gc", "--auto", "--quiet", NULL,
diff --git a/templates/hooks--pre-receive.sample b/templates/hooks--pre-receive.sample
new file mode 100644
index 0000000..a1fd29e
--- /dev/null
+++ b/templates/hooks--pre-receive.sample
@@ -0,0 +1,24 @@
+#!/bin/sh
+#
+# An example hook script to make use of push options.
+# The example simply echoes all push options that start with 'echoback='
+# and rejects all pushes when the "reject" push option is used.
+#
+# To enable this hook, rename this file to "pre-receive".
+
+if test -n "$GIT_PUSH_OPTION_COUNT"
+then
+ i=0
+ while test "$i" -lt "$GIT_PUSH_OPTION_COUNT"
+ do
+ eval "value=\$GIT_PUSH_OPTION_$i"
+ case "$value" in
+ echoback=*)
+ echo "echo from the pre-receive-hook: ${value#*=}" >&2
+ ;;
+ reject)
+ exit 1
+ esac
+ i=$((i + 1))
+ done
+fi
--
2.9.0.247.g176c4f7
^ permalink raw reply related [flat|nested] 9+ messages in thread* [PATCH 2/4] receive-pack: implement advertising and receiving push options
2016-07-09 0:31 [PATCHv4 0/4] Push options Stefan Beller
2016-07-09 0:31 ` [PATCH 1/4] push options: {pre,post}-receive hook learns about push options Stefan Beller
@ 2016-07-09 0:31 ` Stefan Beller
2016-07-10 17:06 ` Shawn Pearce
2016-07-09 0:31 ` [PATCH 3/4] push: accept " Stefan Beller
2016-07-09 0:31 ` [PATCH 4/4] add a test for " Stefan Beller
3 siblings, 1 reply; 9+ messages in thread
From: Stefan Beller @ 2016-07-09 0:31 UTC (permalink / raw)
To: gitster; +Cc: git, e, peff, dwwang, dennis, jrnieder, Stefan Beller
The pre/post receive hook may be interested in more information from the
user. This information can be transmitted when both client and server
support the "push-options" capability, which when used is a phase directly
after update commands ended by a flush pkt.
Similar to the atomic option, the server capability can be disabled via
the `receive.advertisePushOptions` config variable. While documenting
this, fix a nit in the `receive.advertiseAtomic` wording.
Signed-off-by: Stefan Beller <sbeller@google.com>
---
Documentation/config.txt | 9 ++-
Documentation/technical/pack-protocol.txt | 10 ++--
Documentation/technical/protocol-capabilities.txt | 9 +++
builtin/receive-pack.c | 69 +++++++++++++++++++++++
4 files changed, 91 insertions(+), 6 deletions(-)
diff --git a/Documentation/config.txt b/Documentation/config.txt
index e208af1..25b5db1 100644
--- a/Documentation/config.txt
+++ b/Documentation/config.txt
@@ -2410,8 +2410,13 @@ rebase.instructionFormat
receive.advertiseAtomic::
By default, git-receive-pack will advertise the atomic push
- capability to its clients. If you don't want to this capability
- to be advertised, set this variable to false.
+ capability to its clients. If you don't want to advertise this
+ capability, set this variable to false.
+
+receive.advertisePushOptions::
+ By default, git-receive-pack will advertise the push options
+ capability to its clients. If you don't want to advertise this
+ capability, set this variable to false.
receive.autogc::
By default, git-receive-pack will run "git-gc --auto" after
diff --git a/Documentation/technical/pack-protocol.txt b/Documentation/technical/pack-protocol.txt
index 8b36343..7a2ed30 100644
--- a/Documentation/technical/pack-protocol.txt
+++ b/Documentation/technical/pack-protocol.txt
@@ -454,7 +454,8 @@ The reference discovery phase is done nearly the same way as it is in the
fetching protocol. Each reference obj-id and name on the server is sent
in packet-line format to the client, followed by a flush-pkt. The only
real difference is that the capability listing is different - the only
-possible values are 'report-status', 'delete-refs' and 'ofs-delta'.
+possible values are 'report-status', 'delete-refs', 'ofs-delta' and
+'push-options'.
Reference Update Request and Packfile Transfer
----------------------------------------------
@@ -465,9 +466,10 @@ that it wants to update, it sends a line listing the obj-id currently on
the server, the obj-id the client would like to update it to and the name
of the reference.
-This list is followed by a flush-pkt and then the packfile that should
-contain all the objects that the server will need to complete the new
-references.
+This list is followed by a flush-pkt. Then the push options are transmitted
+one per packet followed by another flush-pkt. After that the packfile that
+should contain all the objects that the server will need to complete the new
+references will be sent.
----
update-request = *shallow ( command-list | push-cert ) [packfile]
diff --git a/Documentation/technical/protocol-capabilities.txt b/Documentation/technical/protocol-capabilities.txt
index eaab6b4..4c28d3a 100644
--- a/Documentation/technical/protocol-capabilities.txt
+++ b/Documentation/technical/protocol-capabilities.txt
@@ -253,6 +253,15 @@ atomic pushes. If the pushing client requests this capability, the server
will update the refs in one atomic transaction. Either all refs are
updated or none.
+push-options
+------------
+
+If the server sends the 'push-options' capability it is able to accept
+push options after the update commands have been sent, but before the
+packfile is streamed. If the pushing client requests this capability,
+the server will pass the options to the pre- and post- receive hooks
+that process this push request.
+
allow-tip-sha1-in-want
----------------------
diff --git a/builtin/receive-pack.c b/builtin/receive-pack.c
index 236417e..754db6e 100644
--- a/builtin/receive-pack.c
+++ b/builtin/receive-pack.c
@@ -44,10 +44,12 @@ static struct strbuf fsck_msg_types = STRBUF_INIT;
static int receive_unpack_limit = -1;
static int transfer_unpack_limit = -1;
static int advertise_atomic_push = 1;
+static int advertise_push_options = 1;
static int unpack_limit = 100;
static int report_status;
static int use_sideband;
static int use_atomic;
+static int use_push_options;
static int quiet;
static int prefer_ofs_delta = 1;
static int auto_update_server_info;
@@ -193,6 +195,11 @@ static int receive_pack_config(const char *var, const char *value, void *cb)
return 0;
}
+ if (strcmp(var, "receive.advertisepushoptions") == 0) {
+ advertise_push_options = git_config_bool(var, value);
+ return 0;
+ }
+
return git_default_config(var, value, cb);
}
@@ -211,6 +218,8 @@ static void show_ref(const char *path, const unsigned char *sha1)
strbuf_addstr(&cap, " ofs-delta");
if (push_cert_nonce)
strbuf_addf(&cap, " push-cert=%s", push_cert_nonce);
+ if (advertise_push_options)
+ strbuf_addstr(&cap, " push-options");
strbuf_addf(&cap, " agent=%s", git_user_agent_sanitized());
packet_write(1, "%s %s%c%s\n",
sha1_to_hex(sha1), path, 0, cap.buf);
@@ -1455,6 +1464,9 @@ static struct command *read_head_info(struct sha1_array *shallow)
if (advertise_atomic_push
&& parse_feature_request(feature_list, "atomic"))
use_atomic = 1;
+ if (advertise_push_options
+ && parse_feature_request(feature_list, "push-options"))
+ use_push_options = 1;
}
if (!strcmp(line, "push-cert")) {
@@ -1487,6 +1499,60 @@ static struct command *read_head_info(struct sha1_array *shallow)
return commands;
}
+static struct string_list *read_push_options(void)
+{
+ int i;
+ struct string_list *ret = xmalloc(sizeof(*ret));
+ string_list_init(ret, 1);
+
+ /* NEEDSWORK: expose the limitations to be configurable. */
+ int max_options = 32;
+
+ /*
+ * NEEDSWORK: expose the limitations to be configurable;
+ * Once the limit can be lifted, include a way for payloads
+ * larger than one pkt, e.g allow a payload of up to
+ * LARGE_PACKET_MAX - 1 only, and reserve the last byte
+ * to indicate whether the next pkt continues with this
+ * push option.
+ */
+ int max_size = 1024;
+
+ for (i = 0; i < max_options; i++) {
+ char *line;
+ int len;
+
+ line = packet_read_line(0, &len);
+
+ if (!line)
+ break;
+
+ if (len > max_size) {
+ /*
+ * NEEDSWORK: The error message in die(..) is not
+ * transmitted in call cases, so ideally all die(..)
+ * calls are prefixed with rp_error and then we can
+ * combine rp_error && die into one helper function.
+ */
+ rp_error("protocol error: server configuration allows push "
+ "options of size up to %d bytes", max_size);
+ die("protocol error: push options too large");
+ }
+
+ len = strcspn(line, "\n");
+ line[len] = '\0';
+
+ string_list_append(ret, line);
+ }
+ if (i == max_options) {
+ rp_error("protocol error: server configuration only allows up "
+ "to %d push options", max_options);
+ die("protocol error: push options too large");
+ }
+
+ return ret;
+}
+
static const char *parse_pack_header(struct pack_header *hdr)
{
switch (read_pack_header(0, hdr)) {
@@ -1774,6 +1840,9 @@ int cmd_receive_pack(int argc, const char **argv, const char *prefix)
const char *unpack_status = NULL;
struct string_list *push_options = NULL;
+ if (use_push_options)
+ push_options = read_push_options();
+
prepare_shallow_info(&si, &shallow);
if (!si.nr_ours && !si.nr_theirs)
shallow_update = 0;
--
2.9.0.247.g176c4f7
^ permalink raw reply related [flat|nested] 9+ messages in thread* Re: [PATCH 2/4] receive-pack: implement advertising and receiving push options
2016-07-09 0:31 ` [PATCH 2/4] receive-pack: implement advertising and receiving " Stefan Beller
@ 2016-07-10 17:06 ` Shawn Pearce
2016-07-10 18:05 ` Stefan Beller
2016-07-12 5:24 ` Jeff King
0 siblings, 2 replies; 9+ messages in thread
From: Shawn Pearce @ 2016-07-10 17:06 UTC (permalink / raw)
To: Stefan Beller
Cc: Junio C Hamano, git, e, Jeff King, dwwang, Dennis Kaarsemaker,
Jonathan Nieder
On Fri, Jul 8, 2016 at 5:31 PM, Stefan Beller <sbeller@google.com> wrote:
> +
> + /* NEEDSWORK: expose the limitations to be configurable. */
> + int max_options = 32;
> +
> + /*
> + * NEEDSWORK: expose the limitations to be configurable;
> + * Once the limit can be lifted, include a way for payloads
> + * larger than one pkt, e.g allow a payload of up to
> + * LARGE_PACKET_MAX - 1 only, and reserve the last byte
> + * to indicate whether the next pkt continues with this
> + * push option.
> + */
> + int max_size = 1024;
Instead of this, what about a new config variable
receive.maxCommandBytes[1] that places a limit on the number of bytes
of pkt-line data the client can supply in both the command list ("old
new ref"), push signature framing, and option list?
Memory demands for the server are proportional to the data sent. A
simple byte limit lets the user make the decision about how this gets
used. Longer ref names or option values means fewer refs or options
can be sent. Shorter ref names or option values means more values or
options can be sent.
I studied a lot of repositories[2] and most use ref names under 200
bytes in length. A 3 MiB default for receive.maxCommandBytes gives
users something like 11,115 references in a single git push invocation
if they used all 200 bytes in every name. Most users don't have ref
names this long. Unlike a cap on each ref, it allows users to use the
full 65449 bytes in a reference name available in pkt-line, but you
can only send 48 such references. Likewise for options. :)
[1] I may propose this to JGit.
[2] More than 3M, but maybe Peff has access to more.
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [PATCH 2/4] receive-pack: implement advertising and receiving push options
2016-07-10 17:06 ` Shawn Pearce
@ 2016-07-10 18:05 ` Stefan Beller
2016-07-12 4:53 ` Shawn Pearce
2016-07-12 5:24 ` Jeff King
1 sibling, 1 reply; 9+ messages in thread
From: Stefan Beller @ 2016-07-10 18:05 UTC (permalink / raw)
To: Shawn Pearce
Cc: Junio C Hamano, git, Eric Wong, Jeff King, Dan Wang,
Dennis Kaarsemaker, Jonathan Nieder
On Sun, Jul 10, 2016 at 10:06 AM, Shawn Pearce <spearce@spearce.org> wrote:
> On Fri, Jul 8, 2016 at 5:31 PM, Stefan Beller <sbeller@google.com> wrote:
>> +
>> + /* NEEDSWORK: expose the limitations to be configurable. */
>> + int max_options = 32;
>> +
>> + /*
>> + * NEEDSWORK: expose the limitations to be configurable;
>> + * Once the limit can be lifted, include a way for payloads
>> + * larger than one pkt, e.g allow a payload of up to
>> + * LARGE_PACKET_MAX - 1 only, and reserve the last byte
>> + * to indicate whether the next pkt continues with this
>> + * push option.
>> + */
>> + int max_size = 1024;
>
> Instead of this, what about a new config variable
> receive.maxCommandBytes[1] that places a limit on the number of bytes
> of pkt-line data the client can supply in both the command list ("old
> new ref"), push signature framing, and option list?
Including the whole command list is pretty smart as it actually tackles the
DoS problem as a whole. We shortly discussed having just one upper bound
limit for the push options alone, but we were distracted by the discussion
on whether to advertise this number or just reject it on the server side
after it filled up so much data.
The design here with two bounds was used to not care about the oversized
push options for now. (I mean we can still just reject larger push
options even when
having a receive.maxCommandBytes setting.)
>
> Memory demands for the server are proportional to the data sent. A
> simple byte limit lets the user make the decision about how this gets
> used. Longer ref names or option values means fewer refs or options
> can be sent. Shorter ref names or option values means more values or
> options can be sent.
>
> I studied a lot of repositories[2] and most use ref names under 200
> bytes in length. A 3 MiB default for receive.maxCommandBytes gives
> users something like 11,115 references in a single git push invocation
> if they used all 200 bytes in every name. Most users don't have ref
> names this long. Unlike a cap on each ref, it allows users to use the
> full 65449 bytes in a reference name available in pkt-line, but you
> can only send 48 such references. Likewise for options. :)
In an earlier discussion Jeff said roughly "either make it work well,
or don't make it work at all, i.e. why are git push options better
than a `git push .. && curl <server>/REST-API` thing?"
And by having this design we could punt on the corner cases with
transmitting arbitrary large push options/binaries for now and claim
it's another next step that needs to be done when adding the config
option for it. By having a single receive.maxCommandBytes setting
we would sweep that problem under the rug and people could wonder
why it fails with the large push option.
As said in an earlier email as a side note, we could think about introducing
a v2 pkt line format which starts with a variable int to indicate the packet
size, such that the payload is not bound up to 64k.
I think 3MiB is a bit much for everyday use though and not enough for
corner cases?
>
>
> [1] I may propose this to JGit.
> [2] More than 3M, but maybe Peff has access to more.
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [PATCH 2/4] receive-pack: implement advertising and receiving push options
2016-07-10 18:05 ` Stefan Beller
@ 2016-07-12 4:53 ` Shawn Pearce
0 siblings, 0 replies; 9+ messages in thread
From: Shawn Pearce @ 2016-07-12 4:53 UTC (permalink / raw)
To: Stefan Beller
Cc: Junio C Hamano, git, Eric Wong, Jeff King, Dan Wang,
Dennis Kaarsemaker, Jonathan Nieder
On Sun, Jul 10, 2016 at 11:05 AM, Stefan Beller <sbeller@google.com> wrote:
> On Sun, Jul 10, 2016 at 10:06 AM, Shawn Pearce <spearce@spearce.org> wrote:
>> On Fri, Jul 8, 2016 at 5:31 PM, Stefan Beller <sbeller@google.com> wrote:
>>> +
>>> + /* NEEDSWORK: expose the limitations to be configurable. */
>>> + int max_options = 32;
>>> +
>>> + /*
>>> + * NEEDSWORK: expose the limitations to be configurable;
>>> + * Once the limit can be lifted, include a way for payloads
>>> + * larger than one pkt, e.g allow a payload of up to
>>> + * LARGE_PACKET_MAX - 1 only, and reserve the last byte
>>> + * to indicate whether the next pkt continues with this
>>> + * push option.
>>> + */
>>> + int max_size = 1024;
>>
>> Instead of this, what about a new config variable
>> receive.maxCommandBytes[1] that places a limit on the number of bytes
>> of pkt-line data the client can supply in both the command list ("old
>> new ref"), push signature framing, and option list?
FYI https://git.eclipse.org/r/77108 is my proposal in JGit.
> The design here with two bounds was used to not care about the oversized
> push options for now. (I mean we can still just reject larger push
> options even when
> having a receive.maxCommandBytes setting.)
I don't really see a problem with a single option being nearly 64 KiB
in size. The pack file is going to (probably) need a lot more memory
than that to be unpacked and have SHA-1s generated. The
maxCommandBytes setting at 3 MiB neatly allows a few options to be up
to the pkt-line 64 KiB size, which is 1/10th of what anyone would ever
need. :)
> In an earlier discussion Jeff said roughly "either make it work well,
> or don't make it work at all, i.e. why are git push options better
> than a `git push .. && curl <server>/REST-API` thing?"
One reason this is less optimal is you need to setup authentication
twice. git push might be working over SSH, or over a custom
git-remote-NNN transport. But curl might not speak the same
authentication. Even if git and curl are both using .netrc or an HTTP
cookie file you need to pass the right options to make sure both use
the same auth.
Another is we can do validation of the incoming Git data in context of
what options you specified to us and refuse it while discarding the
pack if we don't like what was asked. With `git push && curl` we have
to save the data without necessarily knowing exactly what we should do
with it, just in case the user comes back with a corresponding REST
API call on another connection.
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH 2/4] receive-pack: implement advertising and receiving push options
2016-07-10 17:06 ` Shawn Pearce
2016-07-10 18:05 ` Stefan Beller
@ 2016-07-12 5:24 ` Jeff King
1 sibling, 0 replies; 9+ messages in thread
From: Jeff King @ 2016-07-12 5:24 UTC (permalink / raw)
To: Shawn Pearce
Cc: Stefan Beller, Junio C Hamano, git, e, dwwang,
Dennis Kaarsemaker, Jonathan Nieder
On Sun, Jul 10, 2016 at 10:06:12AM -0700, Shawn Pearce wrote:
> On Fri, Jul 8, 2016 at 5:31 PM, Stefan Beller <sbeller@google.com> wrote:
> > +
> > + /* NEEDSWORK: expose the limitations to be configurable. */
> > + int max_options = 32;
> > +
> > + /*
> > + * NEEDSWORK: expose the limitations to be configurable;
> > + * Once the limit can be lifted, include a way for payloads
> > + * larger than one pkt, e.g allow a payload of up to
> > + * LARGE_PACKET_MAX - 1 only, and reserve the last byte
> > + * to indicate whether the next pkt continues with this
> > + * push option.
> > + */
> > + int max_size = 1024;
>
> Instead of this, what about a new config variable
> receive.maxCommandBytes[1] that places a limit on the number of bytes
> of pkt-line data the client can supply in both the command list ("old
> new ref"), push signature framing, and option list?
Hmm, I hadn't thought about that. Just for RAM usage, I don't really see
a huge need for individual limits Sometimes meta-limits like this can be
confusing to set ("I have a huge repo with 50,000 refs that I want to
mirror-push; what do I set the byte-limit to?"), but for the most part
people shouldn't need to be touching it. And I wonder if this could
simply be done at the pkt-line level.
OTOH, if the true goal is just to limit memory usage, I wonder if git
should be involved at all. Sadly, things like ulimit or cgroups are not
greated for limiting heap memory. ulimit doesn't count non-brk() memory
on most operating systems, and cgroups is too eager to count shared
packfile mmaps. You can try hooking xmalloc/free to keep stats, but you
have to know the size of each allocation in free(), which means either
being intimate with the underlying malloc implementation, or doing dirty
tricks (e.g., increasing the allocation by a few bytes and shoving the
size there). So it feels like there should be a way to ask the malloc
implementation not to allocate more than N bytes at a time, but AFAIK
glibc malloc does not have such a feature. Maybe other allocators do.
I also think there are some reasons besides RAM usage to limit things.
For instance, there is definitely some per-ref work that receive-pack
does (and hooks may do, as well). That remains proportional to the
amount of data sent, but an annoying client can min-max the parameters
by sending a bunch of short ref names, maximizing the size of the "# of
refs" parameter.
Similarly, I'm as much interested in total RAM usage as I am in making
sure we don't hit weird pathological cases. For instance, places where
tree entry names are so long that they overflow "int" and cause bogus
size computations and access memory outside of the array. That probably
wouldn't be an issue here, if the total data size is on the order of
megabytes (so your worst-case is somebody minimizes all parameters but
one, and then that maximizes the other one; but if it still caps out at
a few megabytes, that's not going to overflow anything).
> I studied a lot of repositories[2] and most use ref names under 200
> bytes in length. A 3 MiB default for receive.maxCommandBytes gives
> users something like 11,115 references in a single git push invocation
> if they used all 200 bytes in every name. Most users don't have ref
> names this long. Unlike a cap on each ref, it allows users to use the
> full 65449 bytes in a reference name available in pkt-line, but you
> can only send 48 such references. Likewise for options. :)
My usual "worst case" for refs is:
https://github.com/JetBrains/intellij-community
A mirror push of that needs more like 4MiB. I'd probably say something
like 10MiB is a reasonable default.
-Peff
^ permalink raw reply [flat|nested] 9+ messages in thread
* [PATCH 3/4] push: accept push options
2016-07-09 0:31 [PATCHv4 0/4] Push options Stefan Beller
2016-07-09 0:31 ` [PATCH 1/4] push options: {pre,post}-receive hook learns about push options Stefan Beller
2016-07-09 0:31 ` [PATCH 2/4] receive-pack: implement advertising and receiving " Stefan Beller
@ 2016-07-09 0:31 ` Stefan Beller
2016-07-09 0:31 ` [PATCH 4/4] add a test for " Stefan Beller
3 siblings, 0 replies; 9+ messages in thread
From: Stefan Beller @ 2016-07-09 0:31 UTC (permalink / raw)
To: gitster; +Cc: git, e, peff, dwwang, dennis, jrnieder, Stefan Beller
This implements everything that is required on the client side to make use
of push options from the porcelain push command.
Signed-off-by: Stefan Beller <sbeller@google.com>
---
Documentation/git-push.txt | 8 +++++++-
builtin/push.c | 21 ++++++++++++++++++---
send-pack.c | 27 +++++++++++++++++++++++++++
send-pack.h | 3 +++
transport.c | 1 +
transport.h | 7 +++++++
6 files changed, 63 insertions(+), 4 deletions(-)
diff --git a/Documentation/git-push.txt b/Documentation/git-push.txt
index 19f46b6..e960258 100644
--- a/Documentation/git-push.txt
+++ b/Documentation/git-push.txt
@@ -11,7 +11,7 @@ SYNOPSIS
[verse]
'git push' [--all | --mirror | --tags] [--follow-tags] [--atomic] [-n | --dry-run] [--receive-pack=<git-receive-pack>]
[--repo=<repository>] [-f | --force] [-d | --delete] [--prune] [-v | --verbose]
- [-u | --set-upstream]
+ [-u | --set-upstream] [--push-option=<string>]
[--[no-]signed|--sign=(true|false|if-asked)]
[--force-with-lease[=<refname>[:<expect>]]]
[--no-verify] [<repository> [<refspec>...]]
@@ -156,6 +156,12 @@ already exists on the remote side.
Either all refs are updated, or on error, no refs are updated.
If the server does not support atomic pushes the push will fail.
+-o::
+--push-option::
+ Transmit the given string to the server, which passes them to
+ the pre-receive as well as the post-receive hook. The given string
+ must not contain a NUL or LF character.
+
--receive-pack=<git-receive-pack>::
--exec=<git-receive-pack>::
Path to the 'git-receive-pack' program on the remote
diff --git a/builtin/push.c b/builtin/push.c
index 4e9e4db..3bb9d6b 100644
--- a/builtin/push.c
+++ b/builtin/push.c
@@ -353,7 +353,8 @@ static int push_with_options(struct transport *transport, int flags)
return 1;
}
-static int do_push(const char *repo, int flags)
+static int do_push(const char *repo, int flags,
+ const struct string_list *push_options)
{
int i, errs;
struct remote *remote = pushremote_get(repo);
@@ -376,6 +377,9 @@ static int do_push(const char *repo, int flags)
if (remote->mirror)
flags |= (TRANSPORT_PUSH_MIRROR|TRANSPORT_PUSH_FORCE);
+ if (push_options->nr)
+ flags |= TRANSPORT_PUSH_OPTIONS;
+
if ((flags & TRANSPORT_PUSH_ALL) && refspec) {
if (!strcmp(*refspec, "refs/tags/*"))
return error(_("--all and --tags are incompatible"));
@@ -406,13 +410,16 @@ static int do_push(const char *repo, int flags)
for (i = 0; i < url_nr; i++) {
struct transport *transport =
transport_get(remote, url[i]);
+ if (flags & TRANSPORT_PUSH_OPTIONS)
+ transport->push_options = push_options;
if (push_with_options(transport, flags))
errs++;
}
} else {
struct transport *transport =
transport_get(remote, NULL);
-
+ if (flags & TRANSPORT_PUSH_OPTIONS)
+ transport->push_options = push_options;
if (push_with_options(transport, flags))
errs++;
}
@@ -500,6 +507,9 @@ int cmd_push(int argc, const char **argv, const char *prefix)
int push_cert = -1;
int rc;
const char *repo = NULL; /* default repository */
+ static struct string_list push_options = STRING_LIST_INIT_DUP;
+ static struct string_list_item *item;
+
struct option options[] = {
OPT__VERBOSITY(&verbosity),
OPT_STRING( 0 , "repo", &repo, N_("repository"), N_("repository")),
@@ -533,6 +543,7 @@ int cmd_push(int argc, const char **argv, const char *prefix)
0, "signed", &push_cert, "yes|no|if-asked", N_("GPG sign the push"),
PARSE_OPT_OPTARG, option_parse_push_signed },
OPT_BIT(0, "atomic", &flags, N_("request atomic transaction on remote side"), TRANSPORT_PUSH_ATOMIC),
+ OPT_STRING_LIST('o', "push-option", &push_options, N_("server-specific"), N_("option to transmit")),
OPT_SET_INT('4', "ipv4", &family, N_("use IPv4 addresses only"),
TRANSPORT_FAMILY_IPV4),
OPT_SET_INT('6', "ipv6", &family, N_("use IPv6 addresses only"),
@@ -563,7 +574,11 @@ int cmd_push(int argc, const char **argv, const char *prefix)
set_refspecs(argv + 1, argc - 1, repo);
}
- rc = do_push(repo, flags);
+ for_each_string_list_item(item, &push_options)
+ if (strchr(item->string, '\n'))
+ die(_("push options must not have new line characters"));
+
+ rc = do_push(repo, flags, &push_options);
if (rc == -1)
usage_with_options(push_usage, options);
else
diff --git a/send-pack.c b/send-pack.c
index 299d303..3a842ac 100644
--- a/send-pack.c
+++ b/send-pack.c
@@ -260,6 +260,7 @@ static int generate_push_cert(struct strbuf *req_buf,
const char *push_cert_nonce)
{
const struct ref *ref;
+ struct string_list_item *item;
char *signing_key = xstrdup(get_signing_key());
const char *cp, *np;
struct strbuf cert = STRBUF_INIT;
@@ -276,6 +277,9 @@ static int generate_push_cert(struct strbuf *req_buf,
}
if (push_cert_nonce[0])
strbuf_addf(&cert, "nonce %s\n", push_cert_nonce);
+ if (args->push_options)
+ for_each_string_list_item(item, args->push_options)
+ strbuf_addf(&cert, "push-option %s\n", item->string);
strbuf_addstr(&cert, "\n");
for (ref = remote_refs; ref; ref = ref->next) {
@@ -370,6 +374,8 @@ int send_pack(struct send_pack_args *args,
int agent_supported = 0;
int use_atomic = 0;
int atomic_supported = 0;
+ int use_push_options = 0;
+ int push_options_supported = 0;
unsigned cmds_sent = 0;
int ret;
struct async demux;
@@ -392,6 +398,8 @@ int send_pack(struct send_pack_args *args,
args->use_thin_pack = 0;
if (server_supports("atomic"))
atomic_supported = 1;
+ if (server_supports("push-options"))
+ push_options_supported = 1;
if (args->push_cert != SEND_PACK_PUSH_CERT_NEVER) {
int len;
@@ -418,6 +426,11 @@ int send_pack(struct send_pack_args *args,
use_atomic = atomic_supported && args->atomic;
+ if (args->push_options && !push_options_supported)
+ die(_("the receiving end does not support push options"));
+
+ use_push_options = push_options_supported && args->push_options;
+
if (status_report)
strbuf_addstr(&cap_buf, " report-status");
if (use_sideband)
@@ -426,6 +439,8 @@ int send_pack(struct send_pack_args *args,
strbuf_addstr(&cap_buf, " quiet");
if (use_atomic)
strbuf_addstr(&cap_buf, " atomic");
+ if (use_push_options)
+ strbuf_addstr(&cap_buf, " push-options");
if (agent_supported)
strbuf_addf(&cap_buf, " agent=%s", git_user_agent_sanitized());
@@ -512,6 +527,18 @@ int send_pack(struct send_pack_args *args,
strbuf_release(&req_buf);
strbuf_release(&cap_buf);
+ if (use_push_options) {
+ struct string_list_item *item;
+ struct strbuf sb = STRBUF_INIT;
+
+ for_each_string_list_item(item, args->push_options)
+ packet_buf_write(&sb, "%s", item->string);
+
+ write_or_die(out, sb.buf, sb.len);
+ packet_flush(out);
+ strbuf_release(&sb);
+ }
+
if (use_sideband && cmds_sent) {
memset(&demux, 0, sizeof(demux));
demux.proc = sideband_demux;
diff --git a/send-pack.h b/send-pack.h
index 57f222a..67fc40f 100644
--- a/send-pack.h
+++ b/send-pack.h
@@ -1,6 +1,8 @@
#ifndef SEND_PACK_H
#define SEND_PACK_H
+#include "string-list.h"
+
/* Possible values for push_cert field in send_pack_args. */
#define SEND_PACK_PUSH_CERT_NEVER 0
#define SEND_PACK_PUSH_CERT_IF_ASKED 1
@@ -21,6 +23,7 @@ struct send_pack_args {
push_cert:2,
stateless_rpc:1,
atomic:1;
+ const struct string_list *push_options;
};
struct option;
diff --git a/transport.c b/transport.c
index 095e61f..0298be1 100644
--- a/transport.c
+++ b/transport.c
@@ -510,6 +510,7 @@ static int git_transport_push(struct transport *transport, struct ref *remote_re
args.dry_run = !!(flags & TRANSPORT_PUSH_DRY_RUN);
args.porcelain = !!(flags & TRANSPORT_PUSH_PORCELAIN);
args.atomic = !!(flags & TRANSPORT_PUSH_ATOMIC);
+ args.push_options = transport->push_options;
args.url = transport->url;
if (flags & TRANSPORT_PUSH_CERT_ALWAYS)
diff --git a/transport.h b/transport.h
index c681408..6fe3485 100644
--- a/transport.h
+++ b/transport.h
@@ -48,6 +48,12 @@ struct transport {
*/
unsigned cloning : 1;
+ /*
+ * These strings will be passed to the {pre, post}-receive hook,
+ * on the remote side, if both sides support the push options capability.
+ */
+ const struct string_list *push_options;
+
/**
* Returns 0 if successful, positive if the option is not
* recognized or is inapplicable, and negative if the option
@@ -134,6 +140,7 @@ struct transport {
#define TRANSPORT_PUSH_CERT_ALWAYS 2048
#define TRANSPORT_PUSH_CERT_IF_ASKED 4096
#define TRANSPORT_PUSH_ATOMIC 8192
+#define TRANSPORT_PUSH_OPTIONS 16384
#define TRANSPORT_SUMMARY_WIDTH (2 * DEFAULT_ABBREV + 3)
#define TRANSPORT_SUMMARY(x) (int)(TRANSPORT_SUMMARY_WIDTH + strlen(x) - gettext_width(x)), (x)
--
2.9.0.247.g176c4f7
^ permalink raw reply related [flat|nested] 9+ messages in thread* [PATCH 4/4] add a test for push options
2016-07-09 0:31 [PATCHv4 0/4] Push options Stefan Beller
` (2 preceding siblings ...)
2016-07-09 0:31 ` [PATCH 3/4] push: accept " Stefan Beller
@ 2016-07-09 0:31 ` Stefan Beller
3 siblings, 0 replies; 9+ messages in thread
From: Stefan Beller @ 2016-07-09 0:31 UTC (permalink / raw)
To: gitster; +Cc: git, e, peff, dwwang, dennis, jrnieder, Stefan Beller
The functions `mk_repo_pair` as well as `test_refs` are borrowed from
t5543-atomic-push, with additional hooks installed.
Signed-off-by: Stefan Beller <sbeller@google.com>
---
t/t5545-push-options.sh | 101 ++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 101 insertions(+)
create mode 100755 t/t5545-push-options.sh
diff --git a/t/t5545-push-options.sh b/t/t5545-push-options.sh
new file mode 100755
index 0000000..8dd3c8e
--- /dev/null
+++ b/t/t5545-push-options.sh
@@ -0,0 +1,101 @@
+#!/bin/sh
+
+test_description='pushing to a repository using push options'
+
+. ./test-lib.sh
+
+mk_repo_pair () {
+ rm -rf workbench upstream &&
+ test_create_repo upstream &&
+ test_create_repo workbench &&
+ (
+ cd upstream &&
+ git config receive.denyCurrentBranch warn &&
+ mkdir -p .git/hooks &&
+ cat >.git/hooks/pre-receive <<-'EOF' &&
+ #!/bin/sh
+ if test -n "$GIT_PUSH_OPTION_COUNT"; then
+ i=0
+ >hooks/pre-receive.push_options
+ while test "$i" -lt "$GIT_PUSH_OPTION_COUNT"; do
+ eval "value=\$GIT_PUSH_OPTION_$i"
+ echo $value >>hooks/pre-receive.push_options
+ i=$((i + 1))
+ done
+ fi
+ EOF
+ chmod u+x .git/hooks/pre-receive
+
+ cat >.git/hooks/post-receive <<-'EOF' &&
+ #!/bin/sh
+ if test -n "$GIT_PUSH_OPTION_COUNT"; then
+ i=0
+ >hooks/post-receive.push_options
+ while test "$i" -lt "$GIT_PUSH_OPTION_COUNT"; do
+ eval "value=\$GIT_PUSH_OPTION_$i"
+ echo $value >>hooks/post-receive.push_options
+ i=$((i + 1))
+ done
+ fi
+ EOF
+ chmod u+x .git/hooks/post-receive
+ ) &&
+ (
+ cd workbench &&
+ git remote add up ../upstream
+ )
+}
+
+# Compare the ref ($1) in upstream with a ref value from workbench ($2)
+# i.e. test_refs second HEAD@{2}
+test_refs () {
+ test $# = 2 &&
+ git -C upstream rev-parse --verify "$1" >expect &&
+ git -C workbench rev-parse --verify "$2" >actual &&
+ test_cmp expect actual
+}
+
+test_expect_success 'one push option works for a single branch' '
+ mk_repo_pair &&
+ (
+ cd workbench &&
+ test_commit one &&
+ git push --mirror up &&
+ test_commit two &&
+ git push --push-option=asdf up master
+ ) &&
+ test_refs master master &&
+ echo "asdf" >expect &&
+ test_cmp expect upstream/.git/hooks/pre-receive.push_options &&
+ test_cmp expect upstream/.git/hooks/post-receive.push_options
+'
+
+test_expect_success 'push option denied by remote' '
+ mk_repo_pair &&
+ git -C upstream config receive.advertisePushOptions false &&
+ (
+ cd workbench &&
+ test_commit one &&
+ git push --mirror up &&
+ test_commit two &&
+ test_must_fail git push --push-option=asdf up master
+ ) &&
+ test_refs master HEAD@{1}
+'
+
+test_expect_success 'two push options work' '
+ mk_repo_pair &&
+ (
+ cd workbench &&
+ test_commit one &&
+ git push --mirror up &&
+ test_commit two &&
+ git push --push-option=asdf --push-option="more structured text" up master
+ ) &&
+ test_refs master master &&
+ printf "asdf\nmore structured text\n" >expect &&
+ test_cmp expect upstream/.git/hooks/pre-receive.push_options &&
+ test_cmp expect upstream/.git/hooks/post-receive.push_options
+'
+
+test_done
--
2.9.0.247.g176c4f7
^ permalink raw reply related [flat|nested] 9+ messages in thread