net/http: first byte of outgoing request body is sometimes lost #17071

jacobsa · 2016-09-12T10:54:25Z

% go version
go version devel +8086e7c Mon Sep 12 06:12:35 2016 +0000 darwin/amd64

(Apologies for not having been able to reproduce this hermetically. I suspect maybe it has something to do with SSL, which I don't know how to set up locally. I'm pretty sure the bug does lie in the Go client though; please keep reading.)

Here is a program that makes a bunch of calls to the Google Cloud Storage API. The calls are all identical. Although this is a heavily-reduced test case and they are not valid according to the GCS API (in particular they're missing credentials), they all do supply valid JSON in the request body.

Despite this, every time I run the program, after at most a few tens of seconds I receive an HTTP 400 parse error from GCS:

% go run http2_bug.go
[...]
Req body:
{"bucket":"foo","name":"bar"}
[...]
Resp body:
{
 "error": {
  "errors": [
   {
    "domain": "global",
    "reason": "parseError",
    "message": "Parse Error"
   }
  ],
  "code": 400,
  "message": "Parse Error"
 }
}

Status code: 400

I happen to have insider knowledge that this charmingly terse error message from GCS indicates that the JSON in the request body could not be parsed. (If you want details, please contact me internally within Google.) In particular, the server on the other end does not receive the opening { character.

Go's own HTTP2 debugging agrees.

% GODEBUG=http2debug=2 go run http2_bug.go

With this invocation, among lots of other output I see

2016/09/12 20:46:54 http2: Framer 0xc420242180: wrote HEADERS flags=END_HEADERS stream=455 len=7
2016/09/12 20:46:54 http2: Framer 0xc420242180: wrote DATA stream=455 len=1 data="\""
2016/09/12 20:46:54 http2: Framer 0xc420242180: wrote DATA stream=455 len=27 data="bucket\":\"foo\",\"name\":\"bar\"}"
2016/09/12 20:46:54 http2: Framer 0xc420242180: wrote DATA flags=END_STREAM stream=455 len=0 data=""
2016/09/12 20:46:54 http2: Framer 0xc420242180: read HEADERS flags=END_HEADERS stream=455 len=94
2016/09/12 20:46:54 http2: Framer 0xc420242180: read DATA flags=END_STREAM stream=455 len=171 data="{\n \"error\": {\n  \"errors\": [\n   {\n    \"domain\": \"global\",\n    \

The http2 package has forgotten to write the opening curly brace. In contrast, everything works fine if I use GODEBUG=http2client=0 to downgrade to HTTP1.

FYI, I have observed this from Sydney (Google office and at home, multiple computers) and from a computer within a Google data center in the US. I've seen this in Go 1.7.1 and Go 1.6.1, as well as at HEAD.

The text was updated successfully, but these errors were encountered:

bradfitz · 2016-09-12T14:47:36Z

Well, crap. https://golang.org/cl/29074

Thanks for the report.

gopherbot · 2016-09-12T15:01:14Z

CL https://golang.org/cl/29074 mentions this issue.

bradfitz · 2016-09-12T16:01:08Z

This probably warrants a Go 1.7.2 since it's sporadic data corruption. It happens whenever a Request.ContentLength == 0 && Request.Body != nil && the connection is either closed or temporarily at max-streams capacity.

Two possible workarounds, though:

set Request.ContentLength
disabled HTTP/2

… a conn bodyAndLength mutates Request.Body if Request.ContentLength == 0, reading the first byte to determine whether it's actually empty or just undeclared. But we did that before we checked whether our connection was overloaded, which meant the caller could retry the request on an new or lesser-loaded connection, but then lose the first byte of the request. Updates golang/go#17071 (needs bundle into std before fixed) Change-Id: I996a92ad037b45bc49e7cf0801a2027bbbb3c920 Reviewed-on: https://go-review.googlesource.com/29074 Reviewed-by: Gleb Stepanov <glebstepanov1992@gmail.com> Reviewed-by: Chris Broadfoot <cbro@golang.org> Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>

gopherbot · 2016-09-12T20:00:58Z

CL https://golang.org/cl/29081 mentions this issue.

jacobsa · 2016-09-12T21:21:09Z

Thanks for the quick fix. Please also consider a patch release for Go 1.6. Note that this bug was actually introduced in a patch release, since Go 1.6 didn't have HTTP2 on by default but 406752b made it into Go 1.6.2. (I may be confused about the details.)

bradfitz · 2016-09-12T21:26:17Z

The first Go 1.6 included HTTP/2 by default for all user-created http.Transports but not for the http.DefaultTransport. Go 1.6.2 fixed that. All the Go 1.6 betas before Go 1.6 had it enabled for the DefaultTransport too.

In any case, we don't issue point releases for two releases back. Only Go 1.7.x at this point.

jacobsa · 2016-09-12T22:13:51Z

Okay, sounds good. Thanks for clarifying.

gopherbot · 2016-09-13T19:00:47Z

CL https://golang.org/cl/29110 mentions this issue.

…Transport until we have a conn bodyAndLength mutates Request.Body if Request.ContentLength == 0, reading the first byte to determine whether it's actually empty or just undeclared. But we did that before we checked whether our connection was overloaded, which meant the caller could retry the request on an new or lesser-loaded connection, but then lose the first byte of the request. Updates golang/go#17071 (needs bundle into std before fixed) Change-Id: I996a92ad037b45bc49e7cf0801a2027bbbb3c920 Reviewed-on: https://go-review.googlesource.com/29074 Reviewed-by: Gleb Stepanov <glebstepanov1992@gmail.com> Reviewed-by: Chris Broadfoot <cbro@golang.org> Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-on: https://go-review.googlesource.com/29081

gopherbot · 2016-09-13T20:00:30Z

CL https://golang.org/cl/29111 mentions this issue.

This works around golang/go#17071 in particular, and is more robust to future similar issues. The sniffing magic performed by http.NewRequest has now been hoisted to the one place where it's actually needed.

In order to work around corruption issues caused by golang/go#17071.

Updates bundled http2 for x/net/http2 git rev 8d4d01f0 for: [release-branch.go1.7] http2: don't sniff first Request.Body byte in Transport until we have a conn https://golang.org/cl/29074 Fixes #17071 Change-Id: I37fef5c2c0fdf499545f9af08abd5f9edb2da4c0 Reviewed-on: https://go-review.googlesource.com/29111 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Chris Broadfoot <cbro@golang.org>

yonderblue · 2016-10-14T02:02:14Z

Is this going to be released?

bradfitz · 2016-10-14T05:49:43Z

Yes. @broady is planning on Monday.

yonderblue · 2016-10-17T21:42:13Z

@bradfitz Forgot to mention, we had to lob this commit off because it caused a deadlock. Some mutex between RoundTrip and CanTakeNewRequest. Sorry I don't have an exact senario for you.

broady · 2016-10-17T21:45:03Z

@Gaillard sorry, can you elaborate? this commit is bad?

broady · 2016-10-17T22:27:51Z

Holding off the 1.7.2 release announcement to get more information from @Gaillard.

yonderblue · 2016-10-18T07:24:25Z

I don't have much more information. Just that it locked up in the client in a concurrent scenario and appeared to be sitting in those two places from the goroutine dumps. Tested with and without the commit but it was not a fully isolated test however seems highly likely since the commit moved that call past a lock.

bradfitz · 2016-10-18T08:00:28Z

Was your request body an io.Pipe? Sounds like #17480

bradfitz · 2016-10-18T08:00:47Z

Please post stack traces of the hang if not.

bradfitz · 2016-10-18T09:02:46Z

Okay, better fix: https://go-review.googlesource.com/31326

@broady, we'll need to cherry-pick that to the 1.7 branch of the "net" repo, and then re-run "go generate" from the "go" repo's 1.7 branch.

gopherbot · 2016-10-18T09:06:12Z

CL https://golang.org/cl/31326 mentions this issue.

…Length Updates golang/go#17480 (fixes after vendor to std) Updates golang/go#17071 (a better fix than the original one) Change-Id: Ibb49d2cb825e28518e688b5c3e0952956a305050 Reviewed-on: https://go-review.googlesource.com/31326 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Chris Broadfoot <cbro@golang.org>

gopherbot · 2016-10-18T19:20:39Z

CL https://golang.org/cl/31361 mentions this issue.

…rt to determine ContentLength Updates golang/go#17480 (fixes after vendor to std) Updates golang/go#17071 (a better fix than the original one) Change-Id: Ibb49d2cb825e28518e688b5c3e0952956a305050 Reviewed-on: https://go-review.googlesource.com/31326 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Chris Broadfoot <cbro@golang.org> Reviewed-on: https://go-review.googlesource.com/31361 Run-TryBot: Chris Broadfoot <cbro@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

gopherbot · 2016-10-18T19:55:12Z

CL https://golang.org/cl/31432 mentions this issue.

yonderblue · 2016-10-18T21:10:29Z

I do remember a pipe in the stacks, sorry I don't have them anymore.

gopherbot · 2016-10-18T21:45:23Z

CL https://golang.org/cl/31434 mentions this issue.

The way to send an explicitly-zero Content-Length is to set a nil Body. Fix this test to do that, rather than relying on type sniffing. Updates #17480 Updates #17071 Change-Id: I6a38e20f17013c88ec4ea69d73c507e4ed886947 Reviewed-on: https://go-review.googlesource.com/31434 TryBot-Result: Gobot Gobot <gobot@golang.org> Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> Reviewed-by: Chris Broadfoot <cbro@golang.org>

gopherbot · 2016-10-18T23:40:10Z

CL https://golang.org/cl/31437 mentions this issue.

…0 Body more reliably The way to send an explicitly-zero Content-Length is to set a nil Body. Fix this test to do that, rather than relying on type sniffing. Updates #17480 Updates #17071 Change-Id: I6a38e20f17013c88ec4ea69d73c507e4ed886947 Reviewed-on: https://go-review.googlesource.com/31434 TryBot-Result: Gobot Gobot <gobot@golang.org> Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> Reviewed-by: Chris Broadfoot <cbro@golang.org> Reviewed-on: https://go-review.googlesource.com/31437 Run-TryBot: Chris Broadfoot <cbro@golang.org>

Updates bundled http2 for x/net/http2 git rev d4c55e66 for: [release-branch.go1.7] http2: never Read from Request.Body in Transport to determine ContentLength https://golang.org/cl/31361 Updates #17480 Updates #17071 Change-Id: I2231adaed3cb5b368927a9654dcf7e69a8b664b6 Reviewed-on: https://go-review.googlesource.com/31432 Run-TryBot: Chris Broadfoot <cbro@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Chris Broadfoot <cbro@golang.org>

gopherbot · 2016-10-19T12:05:32Z

CL https://golang.org/cl/31445 mentions this issue.

… Transport This CL makes NewRequest set Body nil for known-zero bodies, and makes the http1 Transport not peek-Read a byte to determine whether there's a body. Background: Many fields of the Request struct have different meanings for whether they're outgoing (via the Transport) or incoming (via the Server). For outgoing requests, ContentLength and Body are documented as: // Body is the request's body. // // For client requests a nil body means the request has no // body, such as a GET request. The HTTP Client's Transport // is responsible for calling the Close method. Body io.ReadCloser // ContentLength records the length of the associated content. // The value -1 indicates that the length is unknown. // Values >= 0 indicate that the given number of bytes may // be read from Body. // For client requests, a value of 0 with a non-nil Body is // also treated as unknown. ContentLength int64 Because of the ambiguity of what ContentLength==0 means, the http1 and http2 Transports previously Read the first byte of a non-nil Body when the ContentLength was 0 to determine whether there was an actual body (with a non-zero length) and ContentLength just wasn't populated, or it was actually empty. That byte-sniff has been problematic and gross (see #17480, #17071) and was removed for http2 in a previous commit. That means, however, that users doing: req, _ := http.NewRequest("POST", url, strings.NewReader("")) ... would not send a Content-Length header in their http2 request, because the size of the reader (even though it was known, being one of the three common recognized types from NewRequest) was zero, and so the HTTP Transport thought it was simply unset. To signal explicitly-zero vs unset-zero, this CL changes NewRequest to signal explicitly-zero by setting the Body to nil, instead of the strings.NewReader("") or other zero-byte reader. This CL also removes the byte sniff from the http1 Transport, like https://golang.org/cl/31326 did for http2. Updates #17480 Updates #17071 Change-Id: I329f02f124659bf7d8bc01e2c9951ebdd236b52a Reviewed-on: https://go-review.googlesource.com/31445 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>

gopherbot · 2016-10-21T21:00:38Z

CL https://golang.org/cl/31737 mentions this issue.

Updates http2 to x/net/http2 git rev 40a0a18 for: http2: fix Server race with concurrent Read/Close http2: make Server reuse 64k request body buffer between requests http2: never Read from Request.Body in Transport to determine ContentLength Fixes #17480 Updates #17071 Change-Id: If142925764a2e148f95957f559637cfc1785ad21 Reviewed-on: https://go-review.googlesource.com/31737 Reviewed-by: Ian Lance Taylor <iant@golang.org> Run-TryBot: Ian Lance Taylor <iant@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>

Updates bundled http2 for x/net/http2 git rev 8d4d01f0 for: [release-branch.go1.7] http2: don't sniff first Request.Body byte in Transport until we have a conn https://golang.org/cl/29074 Fixes golang#17071 Change-Id: I37fef5c2c0fdf499545f9af08abd5f9edb2da4c0 Reviewed-on: https://go-review.googlesource.com/29111 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Chris Broadfoot <cbro@golang.org>

…0 Body more reliably The way to send an explicitly-zero Content-Length is to set a nil Body. Fix this test to do that, rather than relying on type sniffing. Updates golang#17480 Updates golang#17071 Change-Id: I6a38e20f17013c88ec4ea69d73c507e4ed886947 Reviewed-on: https://go-review.googlesource.com/31434 TryBot-Result: Gobot Gobot <gobot@golang.org> Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> Reviewed-by: Chris Broadfoot <cbro@golang.org> Reviewed-on: https://go-review.googlesource.com/31437 Run-TryBot: Chris Broadfoot <cbro@golang.org>

Updates bundled http2 for x/net/http2 git rev d4c55e66 for: [release-branch.go1.7] http2: never Read from Request.Body in Transport to determine ContentLength https://golang.org/cl/31361 Updates golang#17480 Updates golang#17071 Change-Id: I2231adaed3cb5b368927a9654dcf7e69a8b664b6 Reviewed-on: https://go-review.googlesource.com/31432 Run-TryBot: Chris Broadfoot <cbro@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Chris Broadfoot <cbro@golang.org>

gopherbot · 2016-12-22T06:01:00Z

CL https://golang.org/cl/34668 mentions this issue.

In Go 1.8, we'd removed the Transport's Request.Body one-byte-Read-sniffing to disambiguate between non-nil Request.Body with a ContentLength of 0 or -1. Previously, we tried to see whether a ContentLength of 0 meant actually zero, or just an unset by reading a single byte of the Request.Body and then stitching any read byte back together with the original Request.Body. That historically has caused many problems due to either data races, blocking forever (#17480), or losing bytes (#17071). Thus, we removed it in both HTTP/1 and HTTP/2 in Go 1.8. Unfortunately, during the Go 1.8 beta, we've found that a few people have gotten bitten by the behavior change on requests with methods typically not containing request bodies (e.g. GET, HEAD, DELETE). The most popular example is the aws-go SDK, which always set http.Request.Body to a non-nil value, even on such request methods. That was causing Go 1.8 to send such requests with Transfer-Encoding chunked bodies, with zero bytes, confusing popular servers (including but limited to AWS). This CL partially reverts the no-byte-sniffing behavior and restores it only for GET/HEAD/DELETE/etc requests, and only when there's no Transfer-Encoding set, and the Content-Length is 0 or -1. Updates #18257 (aws-go) bug And also private bug reports about non-AWS issues. Updates #18407 also, but haven't yet audited things enough to declare it fixed. Change-Id: Ie5284d3e067c181839b31faf637eee56e5738a6a Reviewed-on: https://go-review.googlesource.com/34668 Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>

jacobsa changed the title ~~net/http: first character of outgoing request body is sometimes lost~~ net/http: first byte of outgoing request body is sometimes lost Sep 12, 2016

bradfitz self-assigned this Sep 12, 2016

bradfitz added this to the Go1.7.2 milestone Sep 12, 2016

bradfitz mentioned this issue Sep 13, 2016

net/http: backport http2 first byte loss CL to go 1.6.x maybe #17091

Closed

gopherbot closed this in 802cb59 Sep 13, 2016

jacobsa added a commit to jacobsa/comeback that referenced this issue Sep 14, 2016

Update to jacobsa/gcloud@17f7008.

Loading status checks…

d252af5

In order to work around corruption issues caused by golang/go#17071.

fsouza mentioned this issue Oct 18, 2016

Go 1.7.2 Homebrew/homebrew-core#6033

Closed

edvakf mentioned this issue Oct 20, 2016

ベンチマーカー：GoのHTTP2クライアントで最初のバイトがロストするときがあるらしい isucon/isucon6-final#230

Closed

glycerine mentioned this issue Oct 22, 2016

is non-http2 transport available? cockroachdb/cockroach#10150

Closed

jba mentioned this issue Oct 24, 2016

googleapi: Error 400: Parse Error, parseError googleapis/google-cloud-go#389

Closed

bradfitz mentioned this issue May 5, 2017

net/http: client do not set Content-Length: 0 with an io.Reader with no bytes. #20257

Closed

golang locked and limited conversation to collaborators Dec 22, 2017

gopherbot added the FrozenDueToAge label Dec 22, 2017

golang / go

net/http: first byte of outgoing request body is sometimes lost #17071

net/http: first byte of outgoing request body is sometimes lost #17071

jacobsa commented Sep 12, 2016 •

edited

bradfitz commented Sep 12, 2016

gopherbot commented Sep 12, 2016

bradfitz commented Sep 12, 2016

gopherbot commented Sep 12, 2016

jacobsa commented Sep 12, 2016 •

edited

bradfitz commented Sep 12, 2016

jacobsa commented Sep 12, 2016

gopherbot commented Sep 13, 2016

gopherbot commented Sep 13, 2016

yonderblue commented Oct 14, 2016

bradfitz commented Oct 14, 2016

yonderblue commented Oct 17, 2016 •

edited

broady commented Oct 17, 2016

broady commented Oct 17, 2016

yonderblue commented Oct 18, 2016

bradfitz commented Oct 18, 2016

bradfitz commented Oct 18, 2016

bradfitz commented Oct 18, 2016

gopherbot commented Oct 18, 2016

gopherbot commented Oct 18, 2016

gopherbot commented Oct 18, 2016

yonderblue commented Oct 18, 2016

gopherbot commented Oct 18, 2016

gopherbot commented Oct 18, 2016

gopherbot commented Oct 19, 2016

gopherbot commented Oct 21, 2016

gopherbot commented Dec 22, 2016

golang / go

net/http: first byte of outgoing request body is sometimes lost #17071

net/http: first byte of outgoing request body is sometimes lost #17071

Comments

jacobsa commented Sep 12, 2016 • edited

bradfitz commented Sep 12, 2016

gopherbot commented Sep 12, 2016

bradfitz commented Sep 12, 2016

gopherbot commented Sep 12, 2016

jacobsa commented Sep 12, 2016 • edited

bradfitz commented Sep 12, 2016

jacobsa commented Sep 12, 2016

gopherbot commented Sep 13, 2016

gopherbot commented Sep 13, 2016

yonderblue commented Oct 14, 2016

bradfitz commented Oct 14, 2016

yonderblue commented Oct 17, 2016 • edited

broady commented Oct 17, 2016

broady commented Oct 17, 2016

yonderblue commented Oct 18, 2016

bradfitz commented Oct 18, 2016

bradfitz commented Oct 18, 2016

bradfitz commented Oct 18, 2016

gopherbot commented Oct 18, 2016

gopherbot commented Oct 18, 2016

gopherbot commented Oct 18, 2016

yonderblue commented Oct 18, 2016

gopherbot commented Oct 18, 2016

gopherbot commented Oct 18, 2016

gopherbot commented Oct 19, 2016

gopherbot commented Oct 21, 2016

gopherbot commented Dec 22, 2016

jacobsa commented Sep 12, 2016 •

edited

jacobsa commented Sep 12, 2016 •

edited

yonderblue commented Oct 17, 2016 •

edited