github.com/ooni/oohttp@v0.7.2/client.go (about) 1 // Copyright 2009 The Go Authors. All rights reserved. 2 // Use of this source code is governed by a BSD-style 3 // license that can be found in the LICENSE file. 4 5 // HTTP client. See RFC 7230 through 7235. 6 // 7 // This is the high-level Client interface. 8 // The low-level implementation is in transport.go. 9 10 package http 11 12 import ( 13 "context" 14 "crypto/tls" 15 "encoding/base64" 16 "errors" 17 "fmt" 18 "io" 19 "log" 20 "net/url" 21 "reflect" 22 "sort" 23 "strings" 24 "sync" 25 "sync/atomic" 26 "time" 27 28 "github.com/ooni/oohttp/internal/ascii" 29 ) 30 31 // A Client is an HTTP client. Its zero value (DefaultClient) is a 32 // usable client that uses DefaultTransport. 33 // 34 // The Client's Transport typically has internal state (cached TCP 35 // connections), so Clients should be reused instead of created as 36 // needed. Clients are safe for concurrent use by multiple goroutines. 37 // 38 // A Client is higher-level than a RoundTripper (such as Transport) 39 // and additionally handles HTTP details such as cookies and 40 // redirects. 41 // 42 // When following redirects, the Client will forward all headers set on the 43 // initial Request except: 44 // 45 // • when forwarding sensitive headers like "Authorization", 46 // "WWW-Authenticate", and "Cookie" to untrusted targets. 47 // These headers will be ignored when following a redirect to a domain 48 // that is not a subdomain match or exact match of the initial domain. 49 // For example, a redirect from "foo.com" to either "foo.com" or "sub.foo.com" 50 // will forward the sensitive headers, but a redirect to "bar.com" will not. 51 // 52 // • when forwarding the "Cookie" header with a non-nil cookie Jar. 53 // Since each redirect may mutate the state of the cookie jar, 54 // a redirect may possibly alter a cookie set in the initial request. 55 // When forwarding the "Cookie" header, any mutated cookies will be omitted, 56 // with the expectation that the Jar will insert those mutated cookies 57 // with the updated values (assuming the origin matches). 58 // If Jar is nil, the initial cookies are forwarded without change. 59 type Client struct { 60 // Transport specifies the mechanism by which individual 61 // HTTP requests are made. 62 // If nil, DefaultTransport is used. 63 Transport RoundTripper 64 65 // CheckRedirect specifies the policy for handling redirects. 66 // If CheckRedirect is not nil, the client calls it before 67 // following an HTTP redirect. The arguments req and via are 68 // the upcoming request and the requests made already, oldest 69 // first. If CheckRedirect returns an error, the Client's Get 70 // method returns both the previous Response (with its Body 71 // closed) and CheckRedirect's error (wrapped in a url.Error) 72 // instead of issuing the Request req. 73 // As a special case, if CheckRedirect returns ErrUseLastResponse, 74 // then the most recent response is returned with its body 75 // unclosed, along with a nil error. 76 // 77 // If CheckRedirect is nil, the Client uses its default policy, 78 // which is to stop after 10 consecutive requests. 79 CheckRedirect func(req *Request, via []*Request) error 80 81 // Jar specifies the cookie jar. 82 // 83 // The Jar is used to insert relevant cookies into every 84 // outbound Request and is updated with the cookie values 85 // of every inbound Response. The Jar is consulted for every 86 // redirect that the Client follows. 87 // 88 // If Jar is nil, cookies are only sent if they are explicitly 89 // set on the Request. 90 Jar CookieJar 91 92 // Timeout specifies a time limit for requests made by this 93 // Client. The timeout includes connection time, any 94 // redirects, and reading the response body. The timer remains 95 // running after Get, Head, Post, or Do return and will 96 // interrupt reading of the Response.Body. 97 // 98 // A Timeout of zero means no timeout. 99 // 100 // The Client cancels requests to the underlying Transport 101 // as if the Request's Context ended. 102 // 103 // For compatibility, the Client will also use the deprecated 104 // CancelRequest method on Transport if found. New 105 // RoundTripper implementations should use the Request's Context 106 // for cancellation instead of implementing CancelRequest. 107 Timeout time.Duration 108 } 109 110 // DefaultClient is the default Client and is used by Get, Head, and Post. 111 var DefaultClient = &Client{} 112 113 // RoundTripper is an interface representing the ability to execute a 114 // single HTTP transaction, obtaining the Response for a given Request. 115 // 116 // A RoundTripper must be safe for concurrent use by multiple 117 // goroutines. 118 type RoundTripper interface { 119 // RoundTrip executes a single HTTP transaction, returning 120 // a Response for the provided Request. 121 // 122 // RoundTrip should not attempt to interpret the response. In 123 // particular, RoundTrip must return err == nil if it obtained 124 // a response, regardless of the response's HTTP status code. 125 // A non-nil err should be reserved for failure to obtain a 126 // response. Similarly, RoundTrip should not attempt to 127 // handle higher-level protocol details such as redirects, 128 // authentication, or cookies. 129 // 130 // RoundTrip should not modify the request, except for 131 // consuming and closing the Request's Body. RoundTrip may 132 // read fields of the request in a separate goroutine. Callers 133 // should not mutate or reuse the request until the Response's 134 // Body has been closed. 135 // 136 // RoundTrip must always close the body, including on errors, 137 // but depending on the implementation may do so in a separate 138 // goroutine even after RoundTrip returns. This means that 139 // callers wanting to reuse the body for subsequent requests 140 // must arrange to wait for the Close call before doing so. 141 // 142 // The Request's URL and Header fields must be initialized. 143 RoundTrip(*Request) (*Response, error) 144 } 145 146 // refererForURL returns a referer without any authentication info or 147 // an empty string if lastReq scheme is https and newReq scheme is http. 148 // If the referer was explicitly set, then it will continue to be used. 149 func refererForURL(lastReq, newReq *url.URL, explicitRef string) string { 150 // https://tools.ietf.org/html/rfc7231#section-5.5.2 151 // "Clients SHOULD NOT include a Referer header field in a 152 // (non-secure) HTTP request if the referring page was 153 // transferred with a secure protocol." 154 if lastReq.Scheme == "https" && newReq.Scheme == "http" { 155 return "" 156 } 157 if explicitRef != "" { 158 return explicitRef 159 } 160 161 referer := lastReq.String() 162 if lastReq.User != nil { 163 // This is not very efficient, but is the best we can 164 // do without: 165 // - introducing a new method on URL 166 // - creating a race condition 167 // - copying the URL struct manually, which would cause 168 // maintenance problems down the line 169 auth := lastReq.User.String() + "@" 170 referer = strings.Replace(referer, auth, "", 1) 171 } 172 return referer 173 } 174 175 // didTimeout is non-nil only if err != nil. 176 func (c *Client) send(req *Request, deadline time.Time) (resp *Response, didTimeout func() bool, err error) { 177 if c.Jar != nil { 178 for _, cookie := range c.Jar.Cookies(req.URL) { 179 req.AddCookie(cookie) 180 } 181 } 182 resp, didTimeout, err = send(req, c.transport(), deadline) 183 if err != nil { 184 return nil, didTimeout, err 185 } 186 if c.Jar != nil { 187 if rc := resp.Cookies(); len(rc) > 0 { 188 c.Jar.SetCookies(req.URL, rc) 189 } 190 } 191 return resp, nil, nil 192 } 193 194 func (c *Client) deadline() time.Time { 195 if c.Timeout > 0 { 196 return time.Now().Add(c.Timeout) 197 } 198 return time.Time{} 199 } 200 201 func (c *Client) transport() RoundTripper { 202 if c.Transport != nil { 203 return c.Transport 204 } 205 return DefaultTransport 206 } 207 208 // ErrSchemeMismatch is returned when a server returns an HTTP response to an HTTPS client. 209 var ErrSchemeMismatch = errors.New("http: server gave HTTP response to HTTPS client") 210 211 // send issues an HTTP request. 212 // Caller should close resp.Body when done reading from it. 213 func send(ireq *Request, rt RoundTripper, deadline time.Time) (resp *Response, didTimeout func() bool, err error) { 214 req := ireq // req is either the original request, or a modified fork 215 216 if rt == nil { 217 req.closeBody() 218 return nil, alwaysFalse, errors.New("http: no Client.Transport or DefaultTransport") 219 } 220 221 if req.URL == nil { 222 req.closeBody() 223 return nil, alwaysFalse, errors.New("http: nil Request.URL") 224 } 225 226 if req.RequestURI != "" { 227 req.closeBody() 228 return nil, alwaysFalse, errors.New("http: Request.RequestURI can't be set in client requests") 229 } 230 231 // forkReq forks req into a shallow clone of ireq the first 232 // time it's called. 233 forkReq := func() { 234 if ireq == req { 235 req = new(Request) 236 *req = *ireq // shallow clone 237 } 238 } 239 240 // Most the callers of send (Get, Post, et al) don't need 241 // Headers, leaving it uninitialized. We guarantee to the 242 // Transport that this has been initialized, though. 243 if req.Header == nil { 244 forkReq() 245 req.Header = make(Header) 246 } 247 248 if u := req.URL.User; u != nil && req.Header.Get("Authorization") == "" { 249 username := u.Username() 250 password, _ := u.Password() 251 forkReq() 252 req.Header = cloneOrMakeHeader(ireq.Header) 253 req.Header.Set("Authorization", "Basic "+basicAuth(username, password)) 254 } 255 256 if !deadline.IsZero() { 257 forkReq() 258 } 259 stopTimer, didTimeout := setRequestCancel(req, rt, deadline) 260 261 resp, err = rt.RoundTrip(req) 262 if err != nil { 263 stopTimer() 264 if resp != nil { 265 log.Printf("RoundTripper returned a response & error; ignoring response") 266 } 267 if tlsErr, ok := err.(tls.RecordHeaderError); ok { 268 // If we get a bad TLS record header, check to see if the 269 // response looks like HTTP and give a more helpful error. 270 // See golang.org/issue/11111. 271 if string(tlsErr.RecordHeader[:]) == "HTTP/" { 272 err = ErrSchemeMismatch 273 } 274 } 275 return nil, didTimeout, err 276 } 277 if resp == nil { 278 return nil, didTimeout, fmt.Errorf("http: RoundTripper implementation (%T) returned a nil *Response with a nil error", rt) 279 } 280 if resp.Body == nil { 281 // The documentation on the Body field says “The http Client and Transport 282 // guarantee that Body is always non-nil, even on responses without a body 283 // or responses with a zero-length body.” Unfortunately, we didn't document 284 // that same constraint for arbitrary RoundTripper implementations, and 285 // RoundTripper implementations in the wild (mostly in tests) assume that 286 // they can use a nil Body to mean an empty one (similar to Request.Body). 287 // (See https://golang.org/issue/38095.) 288 // 289 // If the ContentLength allows the Body to be empty, fill in an empty one 290 // here to ensure that it is non-nil. 291 if resp.ContentLength > 0 && req.Method != "HEAD" { 292 return nil, didTimeout, fmt.Errorf("http: RoundTripper implementation (%T) returned a *Response with content length %d but a nil Body", rt, resp.ContentLength) 293 } 294 resp.Body = io.NopCloser(strings.NewReader("")) 295 } 296 if !deadline.IsZero() { 297 resp.Body = &cancelTimerBody{ 298 stop: stopTimer, 299 rc: resp.Body, 300 reqDidTimeout: didTimeout, 301 } 302 } 303 return resp, nil, nil 304 } 305 306 // timeBeforeContextDeadline reports whether the non-zero Time t is 307 // before ctx's deadline, if any. If ctx does not have a deadline, it 308 // always reports true (the deadline is considered infinite). 309 func timeBeforeContextDeadline(t time.Time, ctx context.Context) bool { 310 d, ok := ctx.Deadline() 311 if !ok { 312 return true 313 } 314 return t.Before(d) 315 } 316 317 // knownRoundTripperImpl reports whether rt is a RoundTripper that's 318 // maintained by the Go team and known to implement the latest 319 // optional semantics (notably contexts). The Request is used 320 // to check whether this particular request is using an alternate protocol, 321 // in which case we need to check the RoundTripper for that protocol. 322 func knownRoundTripperImpl(rt RoundTripper, req *Request) bool { 323 switch t := rt.(type) { 324 case *Transport: 325 if altRT := t.alternateRoundTripper(req); altRT != nil { 326 return knownRoundTripperImpl(altRT, req) 327 } 328 return true 329 case *http2Transport, http2noDialH2RoundTripper: 330 return true 331 } 332 // There's a very minor chance of a false positive with this. 333 // Instead of detecting our golang.org/x/net/http2.Transport, 334 // it might detect a Transport type in a different http2 335 // package. But I know of none, and the only problem would be 336 // some temporarily leaked goroutines if the transport didn't 337 // support contexts. So this is a good enough heuristic: 338 if reflect.TypeOf(rt).String() == "*http2.Transport" { 339 return true 340 } 341 return false 342 } 343 344 // setRequestCancel sets req.Cancel and adds a deadline context to req 345 // if deadline is non-zero. The RoundTripper's type is used to 346 // determine whether the legacy CancelRequest behavior should be used. 347 // 348 // As background, there are three ways to cancel a request: 349 // First was Transport.CancelRequest. (deprecated) 350 // Second was Request.Cancel. 351 // Third was Request.Context. 352 // This function populates the second and third, and uses the first if it really needs to. 353 func setRequestCancel(req *Request, rt RoundTripper, deadline time.Time) (stopTimer func(), didTimeout func() bool) { 354 if deadline.IsZero() { 355 return nop, alwaysFalse 356 } 357 knownTransport := knownRoundTripperImpl(rt, req) 358 oldCtx := req.Context() 359 360 if req.Cancel == nil && knownTransport { 361 // If they already had a Request.Context that's 362 // expiring sooner, do nothing: 363 if !timeBeforeContextDeadline(deadline, oldCtx) { 364 return nop, alwaysFalse 365 } 366 367 var cancelCtx func() 368 req.ctx, cancelCtx = context.WithDeadline(oldCtx, deadline) 369 return cancelCtx, func() bool { return time.Now().After(deadline) } 370 } 371 initialReqCancel := req.Cancel // the user's original Request.Cancel, if any 372 373 var cancelCtx func() 374 if timeBeforeContextDeadline(deadline, oldCtx) { 375 req.ctx, cancelCtx = context.WithDeadline(oldCtx, deadline) 376 } 377 378 cancel := make(chan struct{}) 379 req.Cancel = cancel 380 381 doCancel := func() { 382 // The second way in the func comment above: 383 close(cancel) 384 // The first way, used only for RoundTripper 385 // implementations written before Go 1.5 or Go 1.6. 386 type canceler interface{ CancelRequest(*Request) } 387 if v, ok := rt.(canceler); ok { 388 v.CancelRequest(req) 389 } 390 } 391 392 stopTimerCh := make(chan struct{}) 393 var once sync.Once 394 stopTimer = func() { 395 once.Do(func() { 396 close(stopTimerCh) 397 if cancelCtx != nil { 398 cancelCtx() 399 } 400 }) 401 } 402 403 timer := time.NewTimer(time.Until(deadline)) 404 var timedOut atomic.Bool 405 406 go func() { 407 select { 408 case <-initialReqCancel: 409 doCancel() 410 timer.Stop() 411 case <-timer.C: 412 timedOut.Store(true) 413 doCancel() 414 case <-stopTimerCh: 415 timer.Stop() 416 } 417 }() 418 419 return stopTimer, timedOut.Load 420 } 421 422 // See 2 (end of page 4) https://www.ietf.org/rfc/rfc2617.txt 423 // "To receive authorization, the client sends the userid and password, 424 // separated by a single colon (":") character, within a base64 425 // encoded string in the credentials." 426 // It is not meant to be urlencoded. 427 func basicAuth(username, password string) string { 428 auth := username + ":" + password 429 return base64.StdEncoding.EncodeToString([]byte(auth)) 430 } 431 432 // Get issues a GET to the specified URL. If the response is one of 433 // the following redirect codes, Get follows the redirect, up to a 434 // maximum of 10 redirects: 435 // 436 // 301 (Moved Permanently) 437 // 302 (Found) 438 // 303 (See Other) 439 // 307 (Temporary Redirect) 440 // 308 (Permanent Redirect) 441 // 442 // An error is returned if there were too many redirects or if there 443 // was an HTTP protocol error. A non-2xx response doesn't cause an 444 // error. Any returned error will be of type *url.Error. The url.Error 445 // value's Timeout method will report true if the request timed out. 446 // 447 // When err is nil, resp always contains a non-nil resp.Body. 448 // Caller should close resp.Body when done reading from it. 449 // 450 // Get is a wrapper around DefaultClient.Get. 451 // 452 // To make a request with custom headers, use NewRequest and 453 // DefaultClient.Do. 454 // 455 // To make a request with a specified context.Context, use NewRequestWithContext 456 // and DefaultClient.Do. 457 func Get(url string) (resp *Response, err error) { 458 return DefaultClient.Get(url) 459 } 460 461 // Get issues a GET to the specified URL. If the response is one of the 462 // following redirect codes, Get follows the redirect after calling the 463 // Client's CheckRedirect function: 464 // 465 // 301 (Moved Permanently) 466 // 302 (Found) 467 // 303 (See Other) 468 // 307 (Temporary Redirect) 469 // 308 (Permanent Redirect) 470 // 471 // An error is returned if the Client's CheckRedirect function fails 472 // or if there was an HTTP protocol error. A non-2xx response doesn't 473 // cause an error. Any returned error will be of type *url.Error. The 474 // url.Error value's Timeout method will report true if the request 475 // timed out. 476 // 477 // When err is nil, resp always contains a non-nil resp.Body. 478 // Caller should close resp.Body when done reading from it. 479 // 480 // To make a request with custom headers, use NewRequest and Client.Do. 481 // 482 // To make a request with a specified context.Context, use NewRequestWithContext 483 // and Client.Do. 484 func (c *Client) Get(url string) (resp *Response, err error) { 485 req, err := NewRequest("GET", url, nil) 486 if err != nil { 487 return nil, err 488 } 489 return c.Do(req) 490 } 491 492 func alwaysFalse() bool { return false } 493 494 // ErrUseLastResponse can be returned by Client.CheckRedirect hooks to 495 // control how redirects are processed. If returned, the next request 496 // is not sent and the most recent response is returned with its body 497 // unclosed. 498 var ErrUseLastResponse = errors.New("net/http: use last response") 499 500 // checkRedirect calls either the user's configured CheckRedirect 501 // function, or the default. 502 func (c *Client) checkRedirect(req *Request, via []*Request) error { 503 fn := c.CheckRedirect 504 if fn == nil { 505 fn = defaultCheckRedirect 506 } 507 return fn(req, via) 508 } 509 510 // redirectBehavior describes what should happen when the 511 // client encounters a 3xx status code from the server. 512 func redirectBehavior(reqMethod string, resp *Response, ireq *Request) (redirectMethod string, shouldRedirect, includeBody bool) { 513 switch resp.StatusCode { 514 case 301, 302, 303: 515 redirectMethod = reqMethod 516 shouldRedirect = true 517 includeBody = false 518 519 // RFC 2616 allowed automatic redirection only with GET and 520 // HEAD requests. RFC 7231 lifts this restriction, but we still 521 // restrict other methods to GET to maintain compatibility. 522 // See Issue 18570. 523 if reqMethod != "GET" && reqMethod != "HEAD" { 524 redirectMethod = "GET" 525 } 526 case 307, 308: 527 redirectMethod = reqMethod 528 shouldRedirect = true 529 includeBody = true 530 531 if ireq.GetBody == nil && ireq.outgoingLength() != 0 { 532 // We had a request body, and 307/308 require 533 // re-sending it, but GetBody is not defined. So just 534 // return this response to the user instead of an 535 // error, like we did in Go 1.7 and earlier. 536 shouldRedirect = false 537 } 538 } 539 return redirectMethod, shouldRedirect, includeBody 540 } 541 542 // urlErrorOp returns the (*url.Error).Op value to use for the 543 // provided (*Request).Method value. 544 func urlErrorOp(method string) string { 545 if method == "" { 546 return "Get" 547 } 548 if lowerMethod, ok := ascii.ToLower(method); ok { 549 return method[:1] + lowerMethod[1:] 550 } 551 return method 552 } 553 554 // Do sends an HTTP request and returns an HTTP response, following 555 // policy (such as redirects, cookies, auth) as configured on the 556 // client. 557 // 558 // An error is returned if caused by client policy (such as 559 // CheckRedirect), or failure to speak HTTP (such as a network 560 // connectivity problem). A non-2xx status code doesn't cause an 561 // error. 562 // 563 // If the returned error is nil, the Response will contain a non-nil 564 // Body which the user is expected to close. If the Body is not both 565 // read to EOF and closed, the Client's underlying RoundTripper 566 // (typically Transport) may not be able to re-use a persistent TCP 567 // connection to the server for a subsequent "keep-alive" request. 568 // 569 // The request Body, if non-nil, will be closed by the underlying 570 // Transport, even on errors. 571 // 572 // On error, any Response can be ignored. A non-nil Response with a 573 // non-nil error only occurs when CheckRedirect fails, and even then 574 // the returned Response.Body is already closed. 575 // 576 // Generally Get, Post, or PostForm will be used instead of Do. 577 // 578 // If the server replies with a redirect, the Client first uses the 579 // CheckRedirect function to determine whether the redirect should be 580 // followed. If permitted, a 301, 302, or 303 redirect causes 581 // subsequent requests to use HTTP method GET 582 // (or HEAD if the original request was HEAD), with no body. 583 // A 307 or 308 redirect preserves the original HTTP method and body, 584 // provided that the Request.GetBody function is defined. 585 // The NewRequest function automatically sets GetBody for common 586 // standard library body types. 587 // 588 // Any returned error will be of type *url.Error. The url.Error 589 // value's Timeout method will report true if the request timed out. 590 func (c *Client) Do(req *Request) (*Response, error) { 591 return c.do(req) 592 } 593 594 var testHookClientDoResult func(retres *Response, reterr error) 595 596 func (c *Client) do(req *Request) (retres *Response, reterr error) { 597 if testHookClientDoResult != nil { 598 defer func() { testHookClientDoResult(retres, reterr) }() 599 } 600 if req.URL == nil { 601 req.closeBody() 602 return nil, &url.Error{ 603 Op: urlErrorOp(req.Method), 604 Err: errors.New("http: nil Request.URL"), 605 } 606 } 607 608 var ( 609 deadline = c.deadline() 610 reqs []*Request 611 resp *Response 612 copyHeaders = c.makeHeadersCopier(req) 613 reqBodyClosed = false // have we closed the current req.Body? 614 615 // Redirect behavior: 616 redirectMethod string 617 includeBody bool 618 ) 619 uerr := func(err error) error { 620 // the body may have been closed already by c.send() 621 if !reqBodyClosed { 622 req.closeBody() 623 } 624 var urlStr string 625 if resp != nil && resp.Request != nil { 626 urlStr = stripPassword(resp.Request.URL) 627 } else { 628 urlStr = stripPassword(req.URL) 629 } 630 return &url.Error{ 631 Op: urlErrorOp(reqs[0].Method), 632 URL: urlStr, 633 Err: err, 634 } 635 } 636 for { 637 // For all but the first request, create the next 638 // request hop and replace req. 639 if len(reqs) > 0 { 640 loc := resp.Header.Get("Location") 641 if loc == "" { 642 // While most 3xx responses include a Location, it is not 643 // required and 3xx responses without a Location have been 644 // observed in the wild. See issues #17773 and #49281. 645 return resp, nil 646 } 647 u, err := req.URL.Parse(loc) 648 if err != nil { 649 resp.closeBody() 650 return nil, uerr(fmt.Errorf("failed to parse Location header %q: %v", loc, err)) 651 } 652 host := "" 653 if req.Host != "" && req.Host != req.URL.Host { 654 // If the caller specified a custom Host header and the 655 // redirect location is relative, preserve the Host header 656 // through the redirect. See issue #22233. 657 if u, _ := url.Parse(loc); u != nil && !u.IsAbs() { 658 host = req.Host 659 } 660 } 661 ireq := reqs[0] 662 req = &Request{ 663 Method: redirectMethod, 664 Response: resp, 665 URL: u, 666 Header: make(Header), 667 Host: host, 668 Cancel: ireq.Cancel, 669 ctx: ireq.ctx, 670 } 671 if includeBody && ireq.GetBody != nil { 672 req.Body, err = ireq.GetBody() 673 if err != nil { 674 resp.closeBody() 675 return nil, uerr(err) 676 } 677 req.ContentLength = ireq.ContentLength 678 } 679 680 // Copy original headers before setting the Referer, 681 // in case the user set Referer on their first request. 682 // If they really want to override, they can do it in 683 // their CheckRedirect func. 684 copyHeaders(req) 685 686 // Add the Referer header from the most recent 687 // request URL to the new one, if it's not https->http: 688 if ref := refererForURL(reqs[len(reqs)-1].URL, req.URL, req.Header.Get("Referer")); ref != "" { 689 req.Header.Set("Referer", ref) 690 } 691 err = c.checkRedirect(req, reqs) 692 693 // Sentinel error to let users select the 694 // previous response, without closing its 695 // body. See Issue 10069. 696 if err == ErrUseLastResponse { 697 return resp, nil 698 } 699 700 // Close the previous response's body. But 701 // read at least some of the body so if it's 702 // small the underlying TCP connection will be 703 // re-used. No need to check for errors: if it 704 // fails, the Transport won't reuse it anyway. 705 const maxBodySlurpSize = 2 << 10 706 if resp.ContentLength == -1 || resp.ContentLength <= maxBodySlurpSize { 707 io.CopyN(io.Discard, resp.Body, maxBodySlurpSize) 708 } 709 resp.Body.Close() 710 711 if err != nil { 712 // Special case for Go 1 compatibility: return both the response 713 // and an error if the CheckRedirect function failed. 714 // See https://golang.org/issue/3795 715 // The resp.Body has already been closed. 716 ue := uerr(err) 717 ue.(*url.Error).URL = loc 718 return resp, ue 719 } 720 } 721 722 reqs = append(reqs, req) 723 var err error 724 var didTimeout func() bool 725 if resp, didTimeout, err = c.send(req, deadline); err != nil { 726 // c.send() always closes req.Body 727 reqBodyClosed = true 728 if !deadline.IsZero() && didTimeout() { 729 err = &httpError{ 730 err: err.Error() + " (Client.Timeout exceeded while awaiting headers)", 731 timeout: true, 732 } 733 } 734 return nil, uerr(err) 735 } 736 737 var shouldRedirect bool 738 redirectMethod, shouldRedirect, includeBody = redirectBehavior(req.Method, resp, reqs[0]) 739 if !shouldRedirect { 740 return resp, nil 741 } 742 743 req.closeBody() 744 } 745 } 746 747 // makeHeadersCopier makes a function that copies headers from the 748 // initial Request, ireq. For every redirect, this function must be called 749 // so that it can copy headers into the upcoming Request. 750 func (c *Client) makeHeadersCopier(ireq *Request) func(*Request) { 751 // The headers to copy are from the very initial request. 752 // We use a closured callback to keep a reference to these original headers. 753 var ( 754 ireqhdr = cloneOrMakeHeader(ireq.Header) 755 icookies map[string][]*Cookie 756 ) 757 if c.Jar != nil && ireq.Header.Get("Cookie") != "" { 758 icookies = make(map[string][]*Cookie) 759 for _, c := range ireq.Cookies() { 760 icookies[c.Name] = append(icookies[c.Name], c) 761 } 762 } 763 764 preq := ireq // The previous request 765 return func(req *Request) { 766 // If Jar is present and there was some initial cookies provided 767 // via the request header, then we may need to alter the initial 768 // cookies as we follow redirects since each redirect may end up 769 // modifying a pre-existing cookie. 770 // 771 // Since cookies already set in the request header do not contain 772 // information about the original domain and path, the logic below 773 // assumes any new set cookies override the original cookie 774 // regardless of domain or path. 775 // 776 // See https://golang.org/issue/17494 777 if c.Jar != nil && icookies != nil { 778 var changed bool 779 resp := req.Response // The response that caused the upcoming redirect 780 for _, c := range resp.Cookies() { 781 if _, ok := icookies[c.Name]; ok { 782 delete(icookies, c.Name) 783 changed = true 784 } 785 } 786 if changed { 787 ireqhdr.Del("Cookie") 788 var ss []string 789 for _, cs := range icookies { 790 for _, c := range cs { 791 ss = append(ss, c.Name+"="+c.Value) 792 } 793 } 794 sort.Strings(ss) // Ensure deterministic headers 795 ireqhdr.Set("Cookie", strings.Join(ss, "; ")) 796 } 797 } 798 799 // Copy the initial request's Header values 800 // (at least the safe ones). 801 for k, vv := range ireqhdr { 802 if shouldCopyHeaderOnRedirect(k, preq.URL, req.URL) { 803 req.Header[k] = vv 804 } 805 } 806 807 preq = req // Update previous Request with the current request 808 } 809 } 810 811 func defaultCheckRedirect(req *Request, via []*Request) error { 812 if len(via) >= 10 { 813 return errors.New("stopped after 10 redirects") 814 } 815 return nil 816 } 817 818 // Post issues a POST to the specified URL. 819 // 820 // Caller should close resp.Body when done reading from it. 821 // 822 // If the provided body is an io.Closer, it is closed after the 823 // request. 824 // 825 // Post is a wrapper around DefaultClient.Post. 826 // 827 // To set custom headers, use NewRequest and DefaultClient.Do. 828 // 829 // See the Client.Do method documentation for details on how redirects 830 // are handled. 831 // 832 // To make a request with a specified context.Context, use NewRequestWithContext 833 // and DefaultClient.Do. 834 func Post(url, contentType string, body io.Reader) (resp *Response, err error) { 835 return DefaultClient.Post(url, contentType, body) 836 } 837 838 // Post issues a POST to the specified URL. 839 // 840 // Caller should close resp.Body when done reading from it. 841 // 842 // If the provided body is an io.Closer, it is closed after the 843 // request. 844 // 845 // To set custom headers, use NewRequest and Client.Do. 846 // 847 // To make a request with a specified context.Context, use NewRequestWithContext 848 // and Client.Do. 849 // 850 // See the Client.Do method documentation for details on how redirects 851 // are handled. 852 func (c *Client) Post(url, contentType string, body io.Reader) (resp *Response, err error) { 853 req, err := NewRequest("POST", url, body) 854 if err != nil { 855 return nil, err 856 } 857 req.Header.Set("Content-Type", contentType) 858 return c.Do(req) 859 } 860 861 // PostForm issues a POST to the specified URL, with data's keys and 862 // values URL-encoded as the request body. 863 // 864 // The Content-Type header is set to application/x-www-form-urlencoded. 865 // To set other headers, use NewRequest and DefaultClient.Do. 866 // 867 // When err is nil, resp always contains a non-nil resp.Body. 868 // Caller should close resp.Body when done reading from it. 869 // 870 // PostForm is a wrapper around DefaultClient.PostForm. 871 // 872 // See the Client.Do method documentation for details on how redirects 873 // are handled. 874 // 875 // To make a request with a specified context.Context, use NewRequestWithContext 876 // and DefaultClient.Do. 877 func PostForm(url string, data url.Values) (resp *Response, err error) { 878 return DefaultClient.PostForm(url, data) 879 } 880 881 // PostForm issues a POST to the specified URL, 882 // with data's keys and values URL-encoded as the request body. 883 // 884 // The Content-Type header is set to application/x-www-form-urlencoded. 885 // To set other headers, use NewRequest and Client.Do. 886 // 887 // When err is nil, resp always contains a non-nil resp.Body. 888 // Caller should close resp.Body when done reading from it. 889 // 890 // See the Client.Do method documentation for details on how redirects 891 // are handled. 892 // 893 // To make a request with a specified context.Context, use NewRequestWithContext 894 // and Client.Do. 895 func (c *Client) PostForm(url string, data url.Values) (resp *Response, err error) { 896 return c.Post(url, "application/x-www-form-urlencoded", strings.NewReader(data.Encode())) 897 } 898 899 // Head issues a HEAD to the specified URL. If the response is one of 900 // the following redirect codes, Head follows the redirect, up to a 901 // maximum of 10 redirects: 902 // 903 // 301 (Moved Permanently) 904 // 302 (Found) 905 // 303 (See Other) 906 // 307 (Temporary Redirect) 907 // 308 (Permanent Redirect) 908 // 909 // Head is a wrapper around DefaultClient.Head. 910 // 911 // To make a request with a specified context.Context, use NewRequestWithContext 912 // and DefaultClient.Do. 913 func Head(url string) (resp *Response, err error) { 914 return DefaultClient.Head(url) 915 } 916 917 // Head issues a HEAD to the specified URL. If the response is one of the 918 // following redirect codes, Head follows the redirect after calling the 919 // Client's CheckRedirect function: 920 // 921 // 301 (Moved Permanently) 922 // 302 (Found) 923 // 303 (See Other) 924 // 307 (Temporary Redirect) 925 // 308 (Permanent Redirect) 926 // 927 // To make a request with a specified context.Context, use NewRequestWithContext 928 // and Client.Do. 929 func (c *Client) Head(url string) (resp *Response, err error) { 930 req, err := NewRequest("HEAD", url, nil) 931 if err != nil { 932 return nil, err 933 } 934 return c.Do(req) 935 } 936 937 // CloseIdleConnections closes any connections on its Transport which 938 // were previously connected from previous requests but are now 939 // sitting idle in a "keep-alive" state. It does not interrupt any 940 // connections currently in use. 941 // 942 // If the Client's Transport does not have a CloseIdleConnections method 943 // then this method does nothing. 944 func (c *Client) CloseIdleConnections() { 945 type closeIdler interface { 946 CloseIdleConnections() 947 } 948 if tr, ok := c.transport().(closeIdler); ok { 949 tr.CloseIdleConnections() 950 } 951 } 952 953 // cancelTimerBody is an io.ReadCloser that wraps rc with two features: 954 // 1. On Read error or close, the stop func is called. 955 // 2. On Read failure, if reqDidTimeout is true, the error is wrapped and 956 // marked as net.Error that hit its timeout. 957 type cancelTimerBody struct { 958 stop func() // stops the time.Timer waiting to cancel the request 959 rc io.ReadCloser 960 reqDidTimeout func() bool 961 } 962 963 func (b *cancelTimerBody) Read(p []byte) (n int, err error) { 964 n, err = b.rc.Read(p) 965 if err == nil { 966 return n, nil 967 } 968 if err == io.EOF { 969 return n, err 970 } 971 if b.reqDidTimeout() { 972 err = &httpError{ 973 err: err.Error() + " (Client.Timeout or context cancellation while reading body)", 974 timeout: true, 975 } 976 } 977 return n, err 978 } 979 980 func (b *cancelTimerBody) Close() error { 981 err := b.rc.Close() 982 b.stop() 983 return err 984 } 985 986 func shouldCopyHeaderOnRedirect(headerKey string, initial, dest *url.URL) bool { 987 switch CanonicalHeaderKey(headerKey) { 988 case "Authorization", "Www-Authenticate", "Cookie", "Cookie2": 989 // Permit sending auth/cookie headers from "foo.com" 990 // to "sub.foo.com". 991 992 // Note that we don't send all cookies to subdomains 993 // automatically. This function is only used for 994 // Cookies set explicitly on the initial outgoing 995 // client request. Cookies automatically added via the 996 // CookieJar mechanism continue to follow each 997 // cookie's scope as set by Set-Cookie. But for 998 // outgoing requests with the Cookie header set 999 // directly, we don't know their scope, so we assume 1000 // it's for *.domain.com. 1001 1002 ihost := idnaASCIIFromURL(initial) 1003 dhost := idnaASCIIFromURL(dest) 1004 return isDomainOrSubdomain(dhost, ihost) 1005 } 1006 // All other headers are copied: 1007 return true 1008 } 1009 1010 // isDomainOrSubdomain reports whether sub is a subdomain (or exact 1011 // match) of the parent domain. 1012 // 1013 // Both domains must already be in canonical form. 1014 func isDomainOrSubdomain(sub, parent string) bool { 1015 if sub == parent { 1016 return true 1017 } 1018 // If sub contains a :, it's probably an IPv6 address (and is definitely not a hostname). 1019 // Don't check the suffix in this case, to avoid matching the contents of a IPv6 zone. 1020 // For example, "::1%.www.example.com" is not a subdomain of "www.example.com". 1021 if strings.ContainsAny(sub, ":%") { 1022 return false 1023 } 1024 // If sub is "foo.example.com" and parent is "example.com", 1025 // that means sub must end in "."+parent. 1026 // Do it without allocating. 1027 if !strings.HasSuffix(sub, parent) { 1028 return false 1029 } 1030 return sub[len(sub)-len(parent)-1] == '.' 1031 } 1032 1033 func stripPassword(u *url.URL) string { 1034 _, passSet := u.User.Password() 1035 if passSet { 1036 return strings.Replace(u.String(), u.User.String()+"@", u.User.Username()+":***@", 1) 1037 } 1038 return u.String() 1039 }