| Message ID | 20200507095829.16894-1-chriscool@tuxfamily.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [RFC] upload-pack: fix filter options scope | expand |
On Thu, May 07, 2020 at 11:58:29AM +0200, Christian Couder wrote: > I am not sure why upload_pack_v2() is called twice. Also it seems > that another similar issue might not be fixed by this. So this patch > is RFC for now. It's because of the request/response model of v2. The client side of the conversation looks something like: - client connects transport to server - client issues command=ls-refs to get the ref advertisement - client issues command=fetch and sends wants/have; the server may respond with a pack when the negotiation is finished, but may also respond with acks, etc, asking for more rounds of negotiation - if there's no pack, the client then issues another command=fetch, repeating until we get a pack In stateless git-over-http, each of those request/response pairs happens in its own server-side process, because the webserver kicks off a separate CGI for each request. But v2 adapts git://, ssh://, and direct-pipe connections to use the same request/response model, but all connected to a single Git process on the other side. So each upload_pack_v2() call needs to start with a clean slate; it can't assume anything about previous rounds, and it needs to clean up any allocated data from those rounds. So your patch is going in the right direction, but we already have other similar data handled by upload_pack_data, which has its own init/clear functions. Probably the filter data should go in there, too. This is an easy bug to introduce, and I suspect we'll see equivalent ones in the future (if there aren't already more lurking; there's a lot of global data in Git, and I really have no idea what subtle interactions one could see). One thing we could do is to truly run each v2 command request in its own process, just like git-over-http does. But there are a lot of complications there around holding the pipe/socket open, how the parent v2 serving process interacts with the child requests (e.g., noticing errors), etc. So I'd worry that going that direction would be just as likely to introduce bugs as fix them. > diff --git a/upload-pack.c b/upload-pack.c > index 902d0ad5e1..4e22e46294 100644 > --- a/upload-pack.c > +++ b/upload-pack.c > @@ -68,7 +68,6 @@ static const char *pack_objects_hook; > static int filter_capability_requested; > static int allow_filter; > static int allow_ref_in_want; > -static struct list_objects_filter_options filter_options; Just looking at nearby bits of global data that might want similar treatment: - allow_filter and allow_ref_in_want are set by the server-side config. So they're persisted between upload_pack_v2() calls, but we'd overwrite them by calling git_config() in each incarnation - filter_capability_requested is set based on client input, but it's only used in v1 But boy there are a lot of global variables there to trace through. Probably a more fruitful way to find similar problems is to look at the v2 process_args() to see what it touches. It looks like options like ofs_delta probably ought to be part of upload_pack_data, too. We didn't notice because it would require a client doing something like: command=fetch ofs-delta [claims to support ofs-delta] 0001 [wants and haves that don't result in a pack yet...] 0000 command=fetch [do not claim to support ofs-delta!] 0001 [wants and haves that result in a pack] 0000 The server _shouldn't_ use ofs-delta there (and wouldn't over http, where the stateless server side that generates the pack knows only about that second request). But in git-over-ssh, it would. A client who does this is stupid and wrong (and I didn't check the protocol spec, but it could very well be violating the spec). So I don't think it's _that_ big a deal. But it would be nice if all of those globals got moved into upload_pack_data, and both v1 and v2 just used it consistently. -Peff
On Thu, May 7, 2020 at 1:36 PM Jeff King <peff@peff.net> wrote: > > On Thu, May 07, 2020 at 11:58:29AM +0200, Christian Couder wrote: > > > I am not sure why upload_pack_v2() is called twice. Also it seems > > that another similar issue might not be fixed by this. So this patch > > is RFC for now. > > It's because of the request/response model of v2. The client side of the > conversation looks something like: > > - client connects transport to server > > - client issues command=ls-refs to get the ref advertisement > > - client issues command=fetch and sends wants/have; the server may > respond with a pack when the negotiation is finished, but may also > respond with acks, etc, asking for more rounds of negotiation > > - if there's no pack, the client then issues another command=fetch, > repeating until we get a pack > > In stateless git-over-http, each of those request/response pairs happens in > its own server-side process, because the webserver kicks off a separate > CGI for each request. But v2 adapts git://, ssh://, and direct-pipe > connections to use the same request/response model, but all connected to > a single Git process on the other side. > > So each upload_pack_v2() call needs to start with a clean slate; it > can't assume anything about previous rounds, and it needs to clean up > any allocated data from those rounds. > > So your patch is going in the right direction, but we already have other > similar data handled by upload_pack_data, which has its own init/clear > functions. Probably the filter data should go in there, too. Thank you for the detailed explanations! It looks like 'struct upload_pack_data' is indeed used by upload_pack_v2(). It's not used by upload_pack() though. So if I move 'struct list_objects_filter_options filter_options' there, I would need to also make upload_pack() either have its own filter_options or use 'struct upload_pack_data' too. > This is an easy bug to introduce, and I suspect we'll see equivalent > ones in the future (if there aren't already more lurking; there's a lot > of global data in Git, and I really have no idea what subtle > interactions one could see). One thing we could do is to truly run each > v2 command request in its own process, just like git-over-http does. But > there are a lot of complications there around holding the pipe/socket > open, how the parent v2 serving process interacts with the child > requests (e.g., noticing errors), etc. So I'd worry that going that > direction would be just as likely to introduce bugs as fix them. I agree that it seems better to just get rid of the global data used by the upload pack functions. > > diff --git a/upload-pack.c b/upload-pack.c > > index 902d0ad5e1..4e22e46294 100644 > > --- a/upload-pack.c > > +++ b/upload-pack.c > > @@ -68,7 +68,6 @@ static const char *pack_objects_hook; > > static int filter_capability_requested; > > static int allow_filter; > > static int allow_ref_in_want; > > -static struct list_objects_filter_options filter_options; > > Just looking at nearby bits of global data that might want similar > treatment: > > - allow_filter and allow_ref_in_want are set by the server-side config. > So they're persisted between upload_pack_v2() calls, but we'd > overwrite them by calling git_config() in each incarnation > > - filter_capability_requested is set based on client input, but it's > only used in v1 > > But boy there are a lot of global variables there to trace through. Yeah, unfortunately. > Probably a more fruitful way to find similar problems is to look at the > v2 process_args() to see what it touches. It looks like options like > ofs_delta probably ought to be part of upload_pack_data, too. We didn't > notice because it would require a client doing something like: > > command=fetch > ofs-delta [claims to support ofs-delta] > 0001 > [wants and haves that don't result in a pack yet...] > 0000 > command=fetch > [do not claim to support ofs-delta!] > 0001 > [wants and haves that result in a pack] > 0000 > > The server _shouldn't_ use ofs-delta there (and wouldn't over http, > where the stateless server side that generates the pack knows only about > that second request). But in git-over-ssh, it would. > > A client who does this is stupid and wrong (and I didn't check the > protocol spec, but it could very well be violating the spec). So I don't > think it's _that_ big a deal. But it would be nice if all of those > globals got moved into upload_pack_data, and both v1 and v2 just used it > consistently. Unfortunately as I discuss a bit above 'struct upload_pack_data' is not used by v1, only by v2. I think making v1 use upload_pack_data might be nice, but it's a separate issue. So I am tempted to just improve the commit message, adding some information from you and from this discussion, and then re-sending. Another intermediate solution would be to add filter_options to 'struct upload_pack_data' for v2 and to use filter_options directly for v1.
On Thu, May 07, 2020 at 02:32:39PM +0200, Christian Couder wrote: > > A client who does this is stupid and wrong (and I didn't check the > > protocol spec, but it could very well be violating the spec). So I don't > > think it's _that_ big a deal. But it would be nice if all of those > > globals got moved into upload_pack_data, and both v1 and v2 just used it > > consistently. > > Unfortunately as I discuss a bit above 'struct upload_pack_data' is > not used by v1, only by v2. I think making v1 use upload_pack_data > might be nice, but it's a separate issue. So I am tempted to just > improve the commit message, adding some information from you and from > this discussion, and then re-sending. > > Another intermediate solution would be to add filter_options to > 'struct upload_pack_data' for v2 and to use filter_options directly > for v1. I think we do want the v1 path to use upload_pack_data in the long run, just to keep things less confusing. But if you don't want to go that far now, I'd strongly prefer the second option, pushing v2 towards how we ultimately want it to look, and plumbing a local variable through v1 paths as necessary. Or perhaps just renaming the global to filter_options_v1 or something, so that v2 code paths aren't tempted to use it. -Peff
On Thu, May 07, 2020 at 10:47:10AM -0400, Jeff King wrote: > On Thu, May 07, 2020 at 02:32:39PM +0200, Christian Couder wrote: > > > > A client who does this is stupid and wrong (and I didn't check the > > > protocol spec, but it could very well be violating the spec). So I don't > > > think it's _that_ big a deal. But it would be nice if all of those > > > globals got moved into upload_pack_data, and both v1 and v2 just used it > > > consistently. > > > > Unfortunately as I discuss a bit above 'struct upload_pack_data' is > > not used by v1, only by v2. I think making v1 use upload_pack_data > > might be nice, but it's a separate issue. So I am tempted to just > > improve the commit message, adding some information from you and from > > this discussion, and then re-sending. > > > > Another intermediate solution would be to add filter_options to > > 'struct upload_pack_data' for v2 and to use filter_options directly > > for v1. > > I think we do want the v1 path to use upload_pack_data in the long run, > just to keep things less confusing. > > But if you don't want to go that far now, I'd strongly prefer the second > option, pushing v2 towards how we ultimately want it to look, and > plumbing a local variable through v1 paths as necessary. Or perhaps just > renaming the global to filter_options_v1 or something, so that v2 code > paths aren't tempted to use it. I would be very much in favor of that direction. Thanks for CC-ing me on this thread, since any changes here will obviously create merge conflicts on my end when I send the "configure allowed object filters" patches. > -Peff Thanks, Taylor
diff --git a/t/t5616-partial-clone.sh b/t/t5616-partial-clone.sh index 88002b24af..b73c38c29a 100755 --- a/t/t5616-partial-clone.sh +++ b/t/t5616-partial-clone.sh @@ -389,7 +389,7 @@ test_expect_success 'fetch lazy-fetches only to resolve deltas, protocol v2' ' # repository did not have tags during clone, but has tags # in the fetch. -test_expect_failure 'verify fetch succeeds when asking for new tags' ' +test_expect_success 'verify fetch succeeds when asking for new tags' ' git clone --filter=blob:none "file://$(pwd)/srv.bare" tag-test && for i in I J K do diff --git a/upload-pack.c b/upload-pack.c index 902d0ad5e1..4e22e46294 100644 --- a/upload-pack.c +++ b/upload-pack.c @@ -68,7 +68,6 @@ static const char *pack_objects_hook; static int filter_capability_requested; static int allow_filter; static int allow_ref_in_want; -static struct list_objects_filter_options filter_options; static int allow_sideband_all; @@ -103,7 +102,8 @@ static int write_one_shallow(const struct commit_graft *graft, void *cb_data) } static void create_pack_file(const struct object_array *have_obj, - const struct object_array *want_obj) + const struct object_array *want_obj, + struct list_objects_filter_options *filter_options) { struct child_process pack_objects = CHILD_PROCESS_INIT; char data[8193], progress[128]; @@ -140,9 +140,9 @@ static void create_pack_file(const struct object_array *have_obj, argv_array_push(&pack_objects.args, "--delta-base-offset"); if (use_include_tag) argv_array_push(&pack_objects.args, "--include-tag"); - if (filter_options.choice) { + if (filter_options->choice) { const char *spec = - expand_list_objects_filter_spec(&filter_options); + expand_list_objects_filter_spec(filter_options); if (pack_objects.use_shell) { struct strbuf buf = STRBUF_INIT; sq_quote_buf(&buf, spec); @@ -848,7 +848,9 @@ static int process_deepen_not(const char *line, struct string_list *deepen_not, return 0; } -static void receive_needs(struct packet_reader *reader, struct object_array *want_obj) +static void receive_needs(struct packet_reader *reader, + struct object_array *want_obj, + struct list_objects_filter_options *filter_options) { struct object_array shallows = OBJECT_ARRAY_INIT; struct string_list deepen_not = STRING_LIST_INIT_DUP; @@ -883,8 +885,8 @@ static void receive_needs(struct packet_reader *reader, struct object_array *wan if (skip_prefix(reader->line, "filter ", &arg)) { if (!filter_capability_requested) die("git upload-pack: filtering capability not negotiated"); - list_objects_filter_die_if_populated(&filter_options); - parse_list_objects_filter(&filter_options, arg); + list_objects_filter_die_if_populated(filter_options); + parse_list_objects_filter(filter_options, arg); continue; } @@ -1087,11 +1089,14 @@ void upload_pack(struct upload_pack_options *options) struct string_list symref = STRING_LIST_INIT_DUP; struct object_array want_obj = OBJECT_ARRAY_INIT; struct packet_reader reader; + struct list_objects_filter_options filter_options; stateless_rpc = options->stateless_rpc; timeout = options->timeout; daemon_mode = options->daemon_mode; + memset(&filter_options, 0, sizeof(filter_options)); + git_config(upload_pack_config, NULL); head_ref_namespaced(find_symref, &symref); @@ -1114,12 +1119,14 @@ void upload_pack(struct upload_pack_options *options) PACKET_READ_CHOMP_NEWLINE | PACKET_READ_DIE_ON_ERR_PACKET); - receive_needs(&reader, &want_obj); + receive_needs(&reader, &want_obj, &filter_options); if (want_obj.nr) { struct object_array have_obj = OBJECT_ARRAY_INIT; get_common_commits(&reader, &have_obj, &want_obj); - create_pack_file(&have_obj, &want_obj); + create_pack_file(&have_obj, &want_obj, &filter_options); } + + list_objects_filter_release(&filter_options); } struct upload_pack_data { @@ -1250,7 +1257,8 @@ static int parse_have(const char *line, struct oid_array *haves) static void process_args(struct packet_reader *request, struct upload_pack_data *data, - struct object_array *want_obj) + struct object_array *want_obj, + struct list_objects_filter_options *filter_options) { while (packet_reader_read(request) == PACKET_READ_NORMAL) { const char *arg = request->line; @@ -1306,8 +1314,8 @@ static void process_args(struct packet_reader *request, } if (allow_filter && skip_prefix(arg, "filter ", &p)) { - list_objects_filter_die_if_populated(&filter_options); - parse_list_objects_filter(&filter_options, p); + list_objects_filter_die_if_populated(filter_options); + parse_list_objects_filter(filter_options, p); continue; } @@ -1470,6 +1478,7 @@ int upload_pack_v2(struct repository *r, struct argv_array *keys, struct upload_pack_data data; struct object_array have_obj = OBJECT_ARRAY_INIT; struct object_array want_obj = OBJECT_ARRAY_INIT; + struct list_objects_filter_options filter_options; clear_object_flags(ALL_FLAGS); @@ -1478,10 +1487,12 @@ int upload_pack_v2(struct repository *r, struct argv_array *keys, upload_pack_data_init(&data); use_sideband = LARGE_PACKET_MAX; + memset(&filter_options, 0, sizeof(filter_options)); + while (state != FETCH_DONE) { switch (state) { case FETCH_PROCESS_ARGS: - process_args(request, &data, &want_obj); + process_args(request, &data, &want_obj, &filter_options); if (!want_obj.nr) { /* @@ -1514,7 +1525,7 @@ int upload_pack_v2(struct repository *r, struct argv_array *keys, send_shallow_info(&data, &want_obj); packet_writer_write(&data.writer, "packfile\n"); - create_pack_file(&have_obj, &want_obj); + create_pack_file(&have_obj, &want_obj, &filter_options); state = FETCH_DONE; break; case FETCH_DONE: @@ -1525,6 +1536,7 @@ int upload_pack_v2(struct repository *r, struct argv_array *keys, upload_pack_data_clear(&data); object_array_clear(&have_obj); object_array_clear(&want_obj); + list_objects_filter_release(&filter_options); return 0; }
The upload_pack_v2() function is sometimes called twice in the same process while 'struct list_objects_filter_options filter_options' was declared as static at the beginning of 'upload-pack.c'. This made the check in list_objects_filter_die_if_populated(), which is called by process_args(), fail the second time upload_pack_v2() is called, as filter_options had already been populated the first time. To fix that, filter_options is not static any more, but instead it's now owned by upload_pack_v2() and upload_pack(). This fixes the first of the 2 bugs documented by d0badf8797 (partial-clone: demonstrate bugs in partial fetch, 2020-02-21). Helped-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Christian Couder <chriscool@tuxfamily.org> --- The issue was originally discussed in: https://lore.kernel.org/git/dbc1bdcae16f8b9941add514264b0fe04cda48c0.1582129312.git.gitgitgadget@gmail.com/ I am not sure why upload_pack_v2() is called twice. Also it seems that another similar issue might not be fixed by this. So this patch is RFC for now. t/t5616-partial-clone.sh | 2 +- upload-pack.c | 40 ++++++++++++++++++++++++++-------------- 2 files changed, 27 insertions(+), 15 deletions(-)