rebase: ParseAcceptLanguage takes a long time to parse complex tags

A vulnerability was found in golang.org/x/text/language package which
could cause a denial of service. An attacker can craft an
Accept-Language header which ParseAcceptLanguage will take significant
time to parse.
Version v0.3.8 of golang.org/x/text fixes a vulnerability.

See-also: https://go.dev/issue/56152
See-also: https://bugzilla.redhat.com/CVE-2022-32149
Signed-off-by: Niels de Vos <ndevos@redhat.com>
This commit is contained in:
Niels de Vos
2022-10-17 08:49:59 +02:00
committed by mergify[bot]
parent b3837d44ce
commit e08005f402
27 changed files with 667 additions and 583 deletions

View File

@ -1178,21 +1178,31 @@ var widthIndex = [1408]uint8{
}
// inverseData contains 4-byte entries of the following format:
// <length> <modified UTF-8-encoded rune> <0 padding>
//
// <length> <modified UTF-8-encoded rune> <0 padding>
//
// The last byte of the UTF-8-encoded rune is xor-ed with the last byte of the
// UTF-8 encoding of the original rune. Mappings often have the following
// pattern:
// -> A (U+FF21 -> U+0041)
// -> B (U+FF22 -> U+0042)
// ...
//
// -> A (U+FF21 -> U+0041)
// -> B (U+FF22 -> U+0042)
// ...
//
// By xor-ing the last byte the same entry can be shared by many mappings. This
// reduces the total number of distinct entries by about two thirds.
// The resulting entry for the aforementioned mappings is
// { 0x01, 0xE0, 0x00, 0x00 }
//
// { 0x01, 0xE0, 0x00, 0x00 }
//
// Using this entry to map U+FF21 (UTF-8 [EF BC A1]), we get
// E0 ^ A1 = 41.
//
// E0 ^ A1 = 41.
//
// Similarly, for U+FF22 (UTF-8 [EF BC A2]), we get
// E0 ^ A2 = 42.
//
// E0 ^ A2 = 42.
//
// Note that because of the xor-ing, the byte sequence stored in the entry is
// not valid UTF-8.
var inverseData = [150][4]byte{