fix(s3): harden Snowball archive extraction (#6942)

* fix(s3): harden Snowball extract error boundaries

* fix(s3): close Snowball extract compatibility gaps

* fix(s3): verify Snowball request body completion

* test(s3): reject forged Snowball streaming signatures

* build(deps): pin Snowball archive parser limits

* fix(s3): preserve Snowball trailer and member errors

* docs(architecture): register Snowball tar fork cleanup

* refactor(s3): route Snowball errors through object boundary

* ci(deps): allow pinned tokio-tar source

* ci(e2e): refresh Snowball smoke selection
This commit is contained in:
cxymds
2026-08-31 19:18:09 +08:00
committed by GitHub
parent 35456bcede
commit ff28b79088
12 changed files with 1650 additions and 169 deletions
+1 -1
View File
@@ -1 +1 @@
sha256=294350518743cac8d7c41880a2835216e4b697908d7b0b1bc92b62816d94c59d sha256=dbebfbab9b9efd4eff31211e69dd32235dc00e207f2ab0dd919a1b2ac9e724c2
-11
View File
@@ -481,23 +481,12 @@ path = "junit.xml"
# parallel-safe — the same property e2e-smoke relies on. The exceptions are the # parallel-safe — the same property e2e-smoke relies on. The exceptions are the
# 4-disk reliability / degraded-read fault-injection tests and the fixed-port # 4-disk reliability / degraded-read fault-injection tests and the fixed-port
# Vault tests, both serialized below. # Vault tests, both serialized below.
# KNOWN-FAILURE EXCLUSIONS (characterization run 29381309848, 2026-07-15:
# 341 ran / 32 failed on the suites' first automated run ever). Deterministic
# product failures cannot be quarantined away with retries, so each family is
# excluded here with its tracking issue, under the same discipline as the
# ci-profile quarantine (docs/testing/README.md): every entry MUST cite one
# OPEN issue, and the fixing PR MUST delete the exclusion. The passing
# negative-path siblings of each family stay in as regression guards.
# * rustfs#4843 — over-limit archive entry paths hard-reject the whole
# archive even under ignore-errors semantics.
[profile.e2e-full] [profile.e2e-full]
default-filter = """ default-filter = """
package(e2e_test) package(e2e_test)
& !test(/^protocols::/) & !test(/^protocols::/)
& !test(/^(admin_timeout_regression_test|cluster_concurrency_test|cluster_multidrive_pool_test|heal_erasure_disk_rebuild_test|namespace_lock_quorum_test|object_lambda_test|stale_multipart_cleanup_cluster_test)::/) & !test(/^(admin_timeout_regression_test|cluster_concurrency_test|cluster_multidrive_pool_test|heal_erasure_disk_rebuild_test|namespace_lock_quorum_test|object_lambda_test|stale_multipart_cleanup_cluster_test)::/)
& !test(/^replication_extension_test::/) & !test(/^replication_extension_test::/)
& !test(/^multipart_auth_test::test_signed_put_object_extract_skips_invalid_entry_when_ignore_errors_enabled$/)
& !test(/^snowball_auto_extract_test::tests::snowball_auto_extract_(ignores_invalid_entries_when_requested|supports_standard_headers_with_combined_extract_options)$/)
""" """
fail-fast = false fail-fast = false
Generated
+1 -2
View File
@@ -627,8 +627,7 @@ dependencies = [
[[package]] [[package]]
name = "astral-tokio-tar" name = "astral-tokio-tar"
version = "0.7.0" version = "0.7.0"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "git+https://github.com/cxymds/tokio-tar.git?rev=603756478b7668436e464519c77ccac22a99ba96#603756478b7668436e464519c77ccac22a99ba96"
checksum = "6f2e989b33246fe9240d39accf4dd9a01e0b6c1f3ce9dd095e0a47fa02505523"
dependencies = [ dependencies = [
"futures-core", "futures-core",
"libc", "libc",
+2 -1
View File
@@ -232,7 +232,8 @@ tokio-postgres-rustls = "0.14.0"
# Utilities and Tools # Utilities and Tools
anyhow = "1.0.104" anyhow = "1.0.104"
arc-swap = "1.9.2" arc-swap = "1.9.2"
astral-tokio-tar = "0.7.0" # RUSTFS_COMPAT_TODO(tokio-tar-extension-limits): keep the fork pin until bounded extension parsing is released upstream. Remove after astral-sh/tokio-tar#118 is merged and a published tokio-tar release exposes the extension limits used here.
astral-tokio-tar = { git = "https://github.com/cxymds/tokio-tar.git", rev = "603756478b7668436e464519c77ccac22a99ba96" }
atoi = "3.1.0" atoi = "3.1.0"
atomic_enum = "0.3.0" atomic_enum = "0.3.0"
aws-config = { version = "1.11.0" } aws-config = { version = "1.11.0" }
@@ -3456,6 +3456,62 @@ async fn test_signed_put_object_extract_expands_tar_entries_with_prefix_headers(
Ok(()) Ok(())
} }
#[tokio::test]
async fn test_signed_put_object_extract_ignore_dirs_skips_unauthorized_directory()
-> Result<(), Box<dyn std::error::Error + Send + Sync>> {
init_logging();
let mut env = RustFSTestEnvironment::new().await?;
env.start_rustfs_server(vec![]).await?;
let bucket = "signed-extract-ignore-dirs-auth";
let archive_key = "bundle.tar";
let allowed_member = "allowed/member.txt";
let denied_directory = "denied/";
let username = "snowball-ignore-dirs";
let secret_key = "snowball-ignore-dirs-secret";
let expected_body = b"allowed-body";
let admin_client = env.create_s3_client();
admin_client.create_bucket().bucket(bucket).send().await?;
create_restricted_user(&env, username, secret_key).await?;
let policy = serde_json::json!({
"Version": "2012-10-17",
"Statement": [{
"Effect": "Allow",
"Principal": { "AWS": [username] },
"Action": ["s3:PutObject"],
"Resource": [
format!("arn:aws:s3:::{bucket}/{archive_key}"),
format!("arn:aws:s3:::{bucket}/{allowed_member}")
]
}]
})
.to_string();
admin_client.put_bucket_policy().bucket(bucket).policy(policy).send().await?;
let restricted_client = restricted_user_client(&env, username, secret_key);
let tar_bytes = make_tar(&[(allowed_member, expected_body)], &[denied_directory]).await;
restricted_client
.put_object()
.bucket(bucket)
.key(archive_key)
.body(ByteStream::from(tar_bytes))
.customize()
.mutate_request(|req| {
req.headers_mut().insert("x-amz-meta-snowball-auto-extract", "true");
req.headers_mut().insert("x-amz-meta-snowball-ignore-dirs", "true");
})
.send()
.await?;
let stored = admin_client.get_object().bucket(bucket).key(allowed_member).send().await?;
assert_eq!(stored.body.collect().await?.into_bytes().as_ref(), expected_body);
Ok(())
}
#[tokio::test] #[tokio::test]
async fn test_signed_put_object_extract_preserves_request_metadata_on_extracted_objects() async fn test_signed_put_object_extract_preserves_request_metadata_on_extracted_objects()
-> Result<(), Box<dyn std::error::Error + Send + Sync>> { -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+182 -9
View File
@@ -36,9 +36,10 @@
use crate::common::{RustFSTestEnvironment, init_logging, local_http_client}; use crate::common::{RustFSTestEnvironment, init_logging, local_http_client};
use aws_sdk_s3::error::ProvideErrorMetadata; use aws_sdk_s3::error::ProvideErrorMetadata;
use aws_sdk_s3::primitives::ByteStream; use aws_sdk_s3::primitives::ByteStream;
use rustfs_signer::constants::UNSIGNED_PAYLOAD; use rustfs_signer::constants::{UNSIGNED_PAYLOAD, UNSIGNED_PAYLOAD_TRAILER};
use rustfs_signer::request_signature_v4::{SIGN_V4_ALGORITHM, get_scope, get_signature, get_signing_key}; use rustfs_signer::request_signature_v4::{SIGN_V4_ALGORITHM, get_scope, get_signature, get_signing_key};
use std::fmt::Write as _; use std::fmt::Write as _;
use std::io::Cursor;
use time::macros::format_description; use time::macros::format_description;
use time::{Duration, OffsetDateTime}; use time::{Duration, OffsetDateTime};
use tracing::info; use tracing::info;
@@ -98,15 +99,37 @@ impl SigV4 {
/// header AND folded into the canonical request — pass the hash of the /// header AND folded into the canonical request — pass the hash of the
/// body you *claim* to send, which may differ from what you actually send. /// body you *claim* to send, which may differ from what you actually send.
fn sign(&self, method: &str, path: &str, canonical_query: &str, content_sha256: &str) -> SignedHeaders { fn sign(&self, method: &str, path: &str, canonical_query: &str, content_sha256: &str) -> SignedHeaders {
let amz_date = amz_datetime(self.time); self.sign_with_extra_headers(method, path, canonical_query, content_sha256, &[])
let signed_headers = "host;x-amz-content-sha256;x-amz-date"; }
let canonical_headers = format!( /// Sign additional request headers while preserving SigV4's lowercase,
"host:{host}\nx-amz-content-sha256:{sha}\nx-amz-date:{date}\n", /// lexicographically sorted canonical-header representation.
host = self.host, fn sign_with_extra_headers(
sha = content_sha256, &self,
date = amz_date, method: &str,
); path: &str,
canonical_query: &str,
content_sha256: &str,
extra_signed_headers: &[(&str, &str)],
) -> SignedHeaders {
let amz_date = amz_datetime(self.time);
let mut canonical_header_values = vec![
("host", self.host.as_str()),
("x-amz-content-sha256", content_sha256),
("x-amz-date", amz_date.as_str()),
];
canonical_header_values.extend(extra_signed_headers.iter().copied());
canonical_header_values.sort_unstable_by(|left, right| left.0.cmp(right.0));
let signed_headers = canonical_header_values
.iter()
.map(|(name, _)| *name)
.collect::<Vec<_>>()
.join(";");
let mut canonical_headers = String::new();
for (name, value) in canonical_header_values {
let _ = writeln!(canonical_headers, "{name}:{value}");
}
let canonical_request = let canonical_request =
format!("{method}\n{path}\n{canonical_query}\n{canonical_headers}\n{signed_headers}\n{content_sha256}"); format!("{method}\n{path}\n{canonical_query}\n{canonical_headers}\n{signed_headers}\n{content_sha256}");
@@ -179,6 +202,34 @@ async fn setup(env: &mut RustFSTestEnvironment) -> Result<(), Box<dyn std::error
Ok(()) Ok(())
} }
async fn build_single_member_archive(
member_key: &str,
member_body: &[u8],
) -> Result<Vec<u8>, Box<dyn std::error::Error + Send + Sync>> {
let mut builder = tokio_tar::Builder::new(Cursor::new(Vec::new()));
let mut header = tokio_tar::Header::new_gnu();
header.set_size(member_body.len() as u64);
header.set_mode(0o644);
header.set_cksum();
builder.append_data(&mut header, member_key, Cursor::new(member_body)).await?;
Ok(builder.into_inner().await?.into_inner())
}
fn sha256_base64(data: &[u8]) -> String {
use sha2::{Digest, Sha256};
base64_simd::STANDARD.encode_to_string(Sha256::digest(data))
}
fn encode_unsigned_aws_chunked_with_sha256_trailer(decoded: &[u8]) -> Vec<u8> {
let checksum = sha256_base64(decoded);
let mut encoded = format!("{:x}\r\n", decoded.len()).into_bytes();
encoded.extend_from_slice(decoded);
encoded.extend_from_slice(b"\r\n0\r\n\r\n");
encoded.extend_from_slice(format!("x-amz-checksum-sha256:{checksum}").as_bytes());
encoded
}
/// Positive control: a correctly hand-signed request must succeed. Without /// Positive control: a correctly hand-signed request must succeed. Without
/// this, every negative assertion below could pass for the wrong reason (a /// this, every negative assertion below could pass for the wrong reason (a
/// broken signer that never produces a valid signature). /// broken signer that never produces a valid signature).
@@ -249,6 +300,128 @@ async fn tampered_signature_returns_signature_does_not_match() -> Result<(), Box
Ok(()) Ok(())
} }
/// `STREAMING-UNSIGNED-PAYLOAD-TRAILER` disables per-chunk signatures, not the
/// seed/header SigV4 signature. A forged request must be rejected before the
/// Snowball handler can publish any archive member.
#[tokio::test]
async fn snowball_streaming_unsigned_trailer_rejects_forged_signature() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
init_logging();
let mut env = RustFSTestEnvironment::new().await?;
setup(&mut env).await?;
let archive_key = "forged-streaming-snowball.tar";
let member_key = "must-not-be-published.txt";
let archive = build_single_member_archive(member_key, b"forged request payload").await?;
let decoded_content_length = archive.len().to_string();
let encoded_body = encode_unsigned_aws_chunked_with_sha256_trailer(&archive);
let path = format!("/{BUCKET}/{archive_key}");
let mut signer = SigV4::new(&env);
signer.secret_key = "wrong-secret-for-forged-streaming-request".to_string();
let extra_signed_headers = [
("content-encoding", "aws-chunked"),
("x-amz-decoded-content-length", decoded_content_length.as_str()),
("x-amz-meta-snowball-auto-extract", "true"),
("x-amz-trailer", "x-amz-checksum-sha256"),
];
let headers = signer.sign_with_extra_headers("PUT", &path, "", UNSIGNED_PAYLOAD_TRAILER, &extra_signed_headers);
let response = local_http_client()
.put(format!("{}{}", env.url, path))
.header("authorization", &headers.authorization)
.header("content-encoding", "aws-chunked")
.header("x-amz-content-sha256", &headers.content_sha256)
.header("x-amz-date", &headers.amz_date)
.header("x-amz-decoded-content-length", &decoded_content_length)
.header("x-amz-meta-snowball-auto-extract", "true")
.header("x-amz-trailer", "x-amz-checksum-sha256")
.body(encoded_body)
.send()
.await?;
let status = response.status();
let body = response.text().await?;
assert_eq!(status.as_u16(), 403, "forged streaming signature must be 403, body:\n{body}");
assert_error_code(&body, "SignatureDoesNotMatch");
let absent = env
.create_s3_client()
.get_object()
.bucket(BUCKET)
.key(member_key)
.send()
.await
.expect_err("a forged streaming request must not publish a Snowball member");
assert_eq!(absent.raw_response().map(|response| response.status().as_u16()), Some(404));
assert_eq!(absent.as_service_error().and_then(ProvideErrorMetadata::code), Some("NoSuchKey"));
env.stop_server();
Ok(())
}
/// Snowball must consume the complete aws-chunked body before reading the
/// trailing checksum exported by s3s into the PutObject response.
#[tokio::test]
async fn snowball_streaming_unsigned_trailer_returns_sha256_checksum() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
init_logging();
let mut env = RustFSTestEnvironment::new().await?;
setup(&mut env).await?;
let archive_key = "valid-streaming-snowball.tar";
let member_key = "streaming-checksum-member.txt";
let member_body = b"valid streaming Snowball payload";
let archive = build_single_member_archive(member_key, member_body).await?;
let expected_checksum = sha256_base64(&archive);
let decoded_content_length = archive.len().to_string();
let encoded_body = encode_unsigned_aws_chunked_with_sha256_trailer(&archive);
let path = format!("/{BUCKET}/{archive_key}");
let signer = SigV4::new(&env);
let extra_signed_headers = [
("content-encoding", "aws-chunked"),
("x-amz-decoded-content-length", decoded_content_length.as_str()),
("x-amz-meta-snowball-auto-extract", "true"),
("x-amz-sdk-checksum-algorithm", "SHA256"),
("x-amz-trailer", "x-amz-checksum-sha256"),
];
let headers = signer.sign_with_extra_headers("PUT", &path, "", UNSIGNED_PAYLOAD_TRAILER, &extra_signed_headers);
let response = local_http_client()
.put(format!("{}{}", env.url, path))
.header("authorization", &headers.authorization)
.header("content-encoding", "aws-chunked")
.header("x-amz-content-sha256", &headers.content_sha256)
.header("x-amz-date", &headers.amz_date)
.header("x-amz-decoded-content-length", &decoded_content_length)
.header("x-amz-meta-snowball-auto-extract", "true")
.header("x-amz-sdk-checksum-algorithm", "SHA256")
.header("x-amz-trailer", "x-amz-checksum-sha256")
.body(encoded_body)
.send()
.await?;
let status = response.status();
let response_checksum = response
.headers()
.get("x-amz-checksum-sha256")
.and_then(|value| value.to_str().ok())
.map(str::to_owned);
let response_body = response.text().await?;
assert_eq!(status.as_u16(), 200, "valid streaming Snowball PUT failed, body:\n{response_body}");
assert_eq!(response_checksum.as_deref(), Some(expected_checksum.as_str()));
let member = env
.create_s3_client()
.get_object()
.bucket(BUCKET)
.key(member_key)
.send()
.await?;
let stored = member.body.collect().await?.into_bytes();
assert_eq!(stored.as_ref(), member_body);
env.stop_server();
Ok(())
}
/// (b) A valid AccessKeyId paired with the wrong secret key must be rejected /// (b) A valid AccessKeyId paired with the wrong secret key must be rejected
/// with SignatureDoesNotMatch / 403. /// with SignatureDoesNotMatch / 403.
#[tokio::test] #[tokio::test]
@@ -17,8 +17,9 @@ mod tests {
use crate::common::{RustFSTestEnvironment, init_logging}; use crate::common::{RustFSTestEnvironment, init_logging};
use aws_sdk_s3::error::ProvideErrorMetadata; use aws_sdk_s3::error::ProvideErrorMetadata;
use aws_sdk_s3::primitives::ByteStream; use aws_sdk_s3::primitives::ByteStream;
use flate2::{Compression, write::GzEncoder};
use std::error::Error; use std::error::Error;
use std::io::Cursor; use std::io::{Cursor, Write};
async fn build_test_archive() -> Result<Vec<u8>, Box<dyn Error + Send + Sync>> { async fn build_test_archive() -> Result<Vec<u8>, Box<dyn Error + Send + Sync>> {
let mut builder = tokio_tar::Builder::new(Cursor::new(Vec::new())); let mut builder = tokio_tar::Builder::new(Cursor::new(Vec::new()));
@@ -69,12 +70,50 @@ mod tests {
Ok(builder.into_inner().await?.into_inner()) Ok(builder.into_inner().await?.into_inner())
} }
fn build_archive_with_parent_dir_entry(victim_bucket: &str) -> Vec<u8> { async fn build_archive_with_invalid_checksum() -> Result<Vec<u8>, Box<dyn Error + Send + Sync>> {
let path = format!("../{victim_bucket}/evil-injected.txt"); let mut archive = build_test_archive().await?;
let data = b"injected-body"; archive[0] ^= 1;
Ok(archive)
}
async fn build_archive_with_negative_gnu_mtime() -> Result<Vec<u8>, Box<dyn Error + Send + Sync>> {
let mut builder = tokio_tar::Builder::new(Cursor::new(Vec::new()));
let mut header = tokio_tar::Header::new_gnu();
header.set_size(b"negative-mtime-body".len() as u64);
header.set_mode(0o644);
header.as_old_mut().mtime.fill(0xff);
builder
.append_data(&mut header, "negative-mtime.txt", Cursor::new(b"negative-mtime-body".as_slice()))
.await?;
Ok(builder.into_inner().await?.into_inner())
}
fn gzip_member(payload: &[u8]) -> Result<Vec<u8>, Box<dyn Error + Send + Sync>> {
let mut encoder = GzEncoder::new(Vec::new(), Compression::default());
encoder.write_all(payload)?;
Ok(encoder.finish()?)
}
async fn build_concatenated_gzip_archive() -> Result<Vec<u8>, Box<dyn Error + Send + Sync>> {
let archive = build_test_archive().await?;
let split_at = archive.len() / 2;
let mut encoded = gzip_member(&archive[..split_at])?;
encoded.extend(gzip_member(&archive[split_at..])?);
Ok(encoded)
}
async fn build_gzip_archive_with_invalid_crc() -> Result<Vec<u8>, Box<dyn Error + Send + Sync>> {
let mut encoded = gzip_member(&build_test_archive().await?)?;
let crc_offset = encoded.len().checked_sub(8).expect("gzip fixture must contain a trailer");
encoded[crc_offset] ^= 1;
Ok(encoded)
}
fn append_raw_tar_entry_with_type(archive: &mut Vec<u8>, path: &[u8], data: &[u8], entry_type: u8) {
assert!(path.len() <= 100, "raw TAR fixture path must fit in the name field");
let mut header = [0u8; 512]; let mut header = [0u8; 512];
header[..path.len()].copy_from_slice(path.as_bytes()); header[..path.len()].copy_from_slice(path);
header[100..108].copy_from_slice(b"0000644\0"); header[100..108].copy_from_slice(b"0000644\0");
header[108..116].copy_from_slice(b"0000000\0"); header[108..116].copy_from_slice(b"0000000\0");
header[116..124].copy_from_slice(b"0000000\0"); header[116..124].copy_from_slice(b"0000000\0");
@@ -82,7 +121,7 @@ mod tests {
header[124..136].copy_from_slice(size.as_bytes()); header[124..136].copy_from_slice(size.as_bytes());
header[136..148].copy_from_slice(b"00000000000\0"); header[136..148].copy_from_slice(b"00000000000\0");
header[148..156].fill(b' '); header[148..156].fill(b' ');
header[156] = b'0'; header[156] = entry_type;
header[257..263].copy_from_slice(b"ustar\0"); header[257..263].copy_from_slice(b"ustar\0");
header[263..265].copy_from_slice(b"00"); header[263..265].copy_from_slice(b"00");
@@ -90,11 +129,36 @@ mod tests {
let checksum = format!("{:06o}\0 ", checksum); let checksum = format!("{:06o}\0 ", checksum);
header[148..156].copy_from_slice(checksum.as_bytes()); header[148..156].copy_from_slice(checksum.as_bytes());
let mut archive = Vec::new();
archive.extend_from_slice(&header); archive.extend_from_slice(&header);
archive.extend_from_slice(data); archive.extend_from_slice(data);
let padding = (512 - (data.len() % 512)) % 512; let padding = (512 - (data.len() % 512)) % 512;
archive.extend(std::iter::repeat_n(0, padding)); archive.extend(std::iter::repeat_n(0, padding));
}
fn append_raw_tar_entry(archive: &mut Vec<u8>, path: &[u8], data: &[u8]) {
append_raw_tar_entry_with_type(archive, path, data, b'0');
}
fn build_archive_with_parent_dir_entry(victim_bucket: &str) -> Vec<u8> {
let path = format!("../{victim_bucket}/evil-injected.txt");
let mut archive = Vec::new();
append_raw_tar_entry(&mut archive, path.as_bytes(), b"injected-body");
archive.extend_from_slice(&[0u8; 1024]);
archive
}
fn build_archive_with_invalid_utf8_entry() -> Vec<u8> {
let mut archive = Vec::new();
append_raw_tar_entry(&mut archive, b"invalid-\xff.txt", b"ignored-body");
append_raw_tar_entry(&mut archive, b"valid.txt", b"valid-body");
archive.extend_from_slice(&[0u8; 1024]);
archive
}
fn build_archive_with_invalid_utf8_symlink() -> Vec<u8> {
let mut archive = Vec::new();
append_raw_tar_entry_with_type(&mut archive, b"invalid-\xff-link", b"", b'2');
append_raw_tar_entry(&mut archive, b"valid.txt", b"valid-body");
archive.extend_from_slice(&[0u8; 1024]); archive.extend_from_slice(&[0u8; 1024]);
archive archive
} }
@@ -263,6 +327,113 @@ mod tests {
Ok(()) Ok(())
} }
#[tokio::test]
async fn snowball_auto_extract_accepts_negative_gnu_mtime() -> Result<(), Box<dyn Error + Send + Sync>> {
init_logging();
let mut env = RustFSTestEnvironment::new().await?;
env.start_rustfs_server(vec![]).await?;
let client = env.create_s3_client();
let bucket = "snowball-negative-mtime";
client.create_bucket().bucket(bucket).send().await?;
client
.put_object()
.bucket(bucket)
.key("fixture.tar")
.metadata("Snowball-Auto-Extract", "true")
.body(ByteStream::from(build_archive_with_negative_gnu_mtime().await?))
.send()
.await?;
let object = client.get_object().bucket(bucket).key("negative-mtime.txt").send().await?;
assert_eq!(object.body.collect().await?.into_bytes().as_ref(), b"negative-mtime-body");
env.stop_server();
Ok(())
}
#[tokio::test]
async fn snowball_auto_extract_consumes_concatenated_gzip_members() -> Result<(), Box<dyn Error + Send + Sync>> {
init_logging();
let mut env = RustFSTestEnvironment::new().await?;
env.start_rustfs_server(vec![]).await?;
let client = env.create_s3_client();
let bucket = "snowball-concatenated-gzip";
client.create_bucket().bucket(bucket).send().await?;
client
.put_object()
.bucket(bucket)
.key("fixture.tar.gz")
.metadata("Snowball-Auto-Extract", "true")
.body(ByteStream::from(build_concatenated_gzip_archive().await?))
.send()
.await?;
let object = client.get_object().bucket(bucket).key("root.txt").send().await?;
assert_eq!(object.body.collect().await?.into_bytes().as_ref(), b"root payload\n");
env.stop_server();
Ok(())
}
#[tokio::test]
async fn snowball_auto_extract_rejects_gzip_crc_error_when_ignore_errors_enabled() -> Result<(), Box<dyn Error + Send + Sync>>
{
init_logging();
let mut env = RustFSTestEnvironment::new().await?;
env.start_rustfs_server(vec![]).await?;
let client = env.create_s3_client();
let bucket = "snowball-gzip-crc-ignore-errors";
client.create_bucket().bucket(bucket).send().await?;
let err = client
.put_object()
.bucket(bucket)
.key("fixture.tar.gz")
.metadata("Snowball-Auto-Extract", "true")
.metadata("Minio-Snowball-Ignore-Errors", "true")
.body(ByteStream::from(build_gzip_archive_with_invalid_crc().await?))
.send()
.await
.expect_err("gzip integrity failures must remain fatal under ignore-errors");
assert_eq!(err.into_service_error().code(), Some("InvalidArgument"));
env.stop_server();
Ok(())
}
#[tokio::test]
async fn snowball_auto_extract_rejects_mismatched_content_md5() -> Result<(), Box<dyn Error + Send + Sync>> {
init_logging();
let mut env = RustFSTestEnvironment::new().await?;
env.start_rustfs_server(vec![]).await?;
let client = env.create_s3_client();
let bucket = "snowball-content-md5";
client.create_bucket().bucket(bucket).send().await?;
let err = client
.put_object()
.bucket(bucket)
.key("fixture.tar")
.metadata("Snowball-Auto-Extract", "true")
.content_md5("AAAAAAAAAAAAAAAAAAAAAA==")
.body(ByteStream::from(build_test_archive().await?))
.send()
.await
.expect_err("mismatched Content-MD5 must fail after the raw body reaches EOF");
assert_eq!(err.into_service_error().code(), Some("BadDigest"));
env.stop_server();
Ok(())
}
#[tokio::test] #[tokio::test]
async fn snowball_auto_extract_ignores_invalid_entries_when_requested() -> Result<(), Box<dyn Error + Send + Sync>> { async fn snowball_auto_extract_ignores_invalid_entries_when_requested() -> Result<(), Box<dyn Error + Send + Sync>> {
init_logging(); init_logging();
@@ -299,7 +470,100 @@ mod tests {
} }
#[tokio::test] #[tokio::test]
async fn snowball_auto_extract_rejects_parent_dir_entry_without_cross_bucket_write() async fn snowball_auto_extract_skips_non_utf8_symlink_without_ignore_errors() -> Result<(), Box<dyn Error + Send + Sync>> {
init_logging();
let mut env = RustFSTestEnvironment::new().await?;
env.start_rustfs_server(vec![]).await?;
let client = env.create_s3_client();
let bucket = "snowball-invalid-utf8-link";
client.create_bucket().bucket(bucket).send().await?;
client
.put_object()
.bucket(bucket)
.key("fixture.tar")
.metadata("Snowball-Auto-Extract", "true")
.body(ByteStream::from(build_archive_with_invalid_utf8_symlink()))
.send()
.await?;
let valid = client.get_object().bucket(bucket).key("valid.txt").send().await?;
assert_eq!(valid.body.collect().await?.into_bytes().as_ref(), b"valid-body");
let listed = client.list_objects_v2().bucket(bucket).send().await?;
let keys: Vec<_> = listed.contents().iter().filter_map(|entry| entry.key()).collect();
assert_eq!(keys, vec!["valid.txt"]);
env.stop_server();
Ok(())
}
#[tokio::test]
async fn snowball_auto_extract_skips_non_utf8_member_without_lossy_key_collision() -> Result<(), Box<dyn Error + Send + Sync>>
{
init_logging();
let mut env = RustFSTestEnvironment::new().await?;
env.start_rustfs_server(vec![]).await?;
let client = env.create_s3_client();
let bucket = "snowball-invalid-utf8";
client.create_bucket().bucket(bucket).send().await?;
client
.put_object()
.bucket(bucket)
.key("fixture.tar")
.metadata("Snowball-Auto-Extract", "true")
.metadata("Minio-Snowball-Ignore-Errors", "true")
.body(ByteStream::from(build_archive_with_invalid_utf8_entry()))
.send()
.await?;
let valid = client.get_object().bucket(bucket).key("valid.txt").send().await?;
assert_eq!(valid.body.collect().await?.into_bytes().as_ref(), b"valid-body");
let listed = client.list_objects_v2().bucket(bucket).send().await?;
let keys: Vec<_> = listed.contents().iter().filter_map(|entry| entry.key()).collect();
assert_eq!(keys, vec!["valid.txt"]);
env.stop_server();
Ok(())
}
#[tokio::test]
async fn snowball_auto_extract_rejects_corrupt_tar_when_ignore_errors_enabled() -> Result<(), Box<dyn Error + Send + Sync>> {
init_logging();
let mut env = RustFSTestEnvironment::new().await?;
env.start_rustfs_server(vec![]).await?;
let client = env.create_s3_client();
let bucket = "snowball-corrupt-ignore-errors";
let archive = build_archive_with_invalid_checksum().await?;
client.create_bucket().bucket(bucket).send().await?;
let err = client
.put_object()
.bucket(bucket)
.key("fixture.tar")
.metadata("Snowball-Auto-Extract", "true")
.metadata("Minio-Snowball-Ignore-Errors", "true")
.body(ByteStream::from(archive))
.send()
.await
.expect_err("corrupt TAR structure must remain fatal under ignore-errors");
assert_eq!(err.into_service_error().code(), Some("InvalidArgument"));
let listed = client.list_objects_v2().bucket(bucket).send().await?;
assert!(listed.contents().is_empty(), "corrupt archive must not produce objects");
env.stop_server();
Ok(())
}
#[tokio::test]
async fn snowball_auto_extract_rejects_parent_dir_entry_even_when_ignore_errors_enabled()
-> Result<(), Box<dyn Error + Send + Sync>> { -> Result<(), Box<dyn Error + Send + Sync>> {
init_logging(); init_logging();
@@ -319,6 +583,7 @@ mod tests {
.bucket(attacker_bucket) .bucket(attacker_bucket)
.key("fixture.tar") .key("fixture.tar")
.metadata("Snowball-Auto-Extract", "true") .metadata("Snowball-Auto-Extract", "true")
.metadata("Minio-Snowball-Ignore-Errors", "true")
.body(ByteStream::from(archive)) .body(ByteStream::from(archive))
.send() .send()
.await .await
+59 -5
View File
@@ -47,7 +47,12 @@ pub struct ArchiveLimits {
pub max_entries: usize, pub max_entries: usize,
pub max_entry_size: u64, pub max_entry_size: u64,
pub max_total_unpacked_size: u64, pub max_total_unpacked_size: u64,
pub max_decoded_size: u64,
pub max_path_length: usize, pub max_path_length: usize,
pub max_pax_metadata_size: u64,
pub max_total_pax_metadata_size: u64,
pub max_pax_metadata_records: usize,
pub max_total_pax_metadata_records: usize,
pub validate_entry_paths: bool, pub validate_entry_paths: bool,
} }
@@ -57,7 +62,12 @@ impl Default for ArchiveLimits {
max_entries: 100_000, max_entries: 100_000,
max_entry_size: 1_073_741_824, max_entry_size: 1_073_741_824,
max_total_unpacked_size: 10_737_418_240, max_total_unpacked_size: 10_737_418_240,
max_decoded_size: 11_811_160_064,
max_path_length: 1024, max_path_length: 1024,
max_pax_metadata_size: 1_048_576,
max_total_pax_metadata_size: 67_108_864,
max_pax_metadata_records: 4_096,
max_total_pax_metadata_records: 100_000,
validate_entry_paths: true, validate_entry_paths: true,
} }
} }
@@ -100,11 +110,31 @@ impl CompressionFormat {
let reader = BufReader::new(input); let reader = BufReader::new(input);
let decoder: Box<dyn AsyncRead + Send + Unpin + 'static> = match self { let decoder: Box<dyn AsyncRead + Send + Unpin + 'static> = match self {
CompressionFormat::Gzip => Box::new(GzipDecoder::new(reader)), CompressionFormat::Gzip => {
CompressionFormat::Bzip2 => Box::new(BzDecoder::new(reader)), let mut decoder = GzipDecoder::new(reader);
CompressionFormat::Zlib => Box::new(ZlibDecoder::new(reader)), decoder.multiple_members(true);
CompressionFormat::Xz => Box::new(XzDecoder::new(reader)), Box::new(decoder)
CompressionFormat::Zstd => Box::new(ZstdDecoder::new(reader)), }
CompressionFormat::Bzip2 => {
let mut decoder = BzDecoder::new(reader);
decoder.multiple_members(true);
Box::new(decoder)
}
CompressionFormat::Zlib => {
let mut decoder = ZlibDecoder::new(reader);
decoder.multiple_members(true);
Box::new(decoder)
}
CompressionFormat::Xz => {
let mut decoder = XzDecoder::new(reader);
decoder.multiple_members(true);
Box::new(decoder)
}
CompressionFormat::Zstd => {
let mut decoder = ZstdDecoder::new(reader);
decoder.multiple_members(true);
Box::new(decoder)
}
CompressionFormat::Tar => Box::new(reader), CompressionFormat::Tar => Box::new(reader),
CompressionFormat::Zip => { CompressionFormat::Zip => {
return Err(ZipError::UnsupportedFormat { return Err(ZipError::UnsupportedFormat {
@@ -160,6 +190,30 @@ mod tests {
assert_eq!(decoded, b"payload"); assert_eq!(decoded, b"payload");
} }
#[tokio::test]
async fn test_get_decoder_consumes_concatenated_gzip_members() {
async fn gzip_member(payload: &[u8]) -> Vec<u8> {
let mut encoder = GzipEncoder::new(Vec::new());
encoder.write_all(payload).await.expect("gzip encode should succeed");
encoder.shutdown().await.expect("gzip encoder shutdown should succeed");
encoder.into_inner()
}
let mut encoded = gzip_member(b"first-").await;
encoded.extend(gzip_member(b"second").await);
let mut decoder = CompressionFormat::Gzip
.get_decoder(std::io::Cursor::new(encoded))
.expect("gzip decoder should be created");
let mut decoded = Vec::new();
decoder
.read_to_end(&mut decoded)
.await
.expect("concatenated gzip members should decode");
assert_eq!(decoded, b"first-second");
}
#[tokio::test] #[tokio::test]
async fn test_get_decoder_rejects_zip_and_unknown_formats() { async fn test_get_decoder_rejects_zip_and_unknown_formats() {
let zip_err = CompressionFormat::Zip let zip_err = CompressionFormat::Zip
+4
View File
@@ -36,6 +36,10 @@ unknown-registry = "deny"
unknown-git = "deny" unknown-git = "deny"
allow-registry = ["https://github.com/rust-lang/crates.io-index"] allow-registry = ["https://github.com/rust-lang/crates.io-index"]
allow-git = [ allow-git = [
# Temporary tokio-tar fork pinned to the reviewed bounded extension parser
# change while astral-sh/tokio-tar#118 awaits an upstream release.
# owner: cxymds review: 2026-10
"https://github.com/cxymds/tokio-tar.git",
# Official s3s repository. Temporarily pinned to the merged generic REST # Official s3s repository. Temporarily pinned to the merged generic REST
# SigV4 payload-checksum fix until it is available in a crates.io release. # SigV4 payload-checksum fix until it is available in a crates.io release.
# owner: marshawcoco review: 2026-10 # owner: marshawcoco review: 2026-10
@@ -12,6 +12,7 @@ for later deletion.
## Open Items ## Open Items
- `tokio-tar-extension-limits` bounded archive extension parsing: Snowball extraction depends on explicit limits for GNU long-name, GNU long-link, and PAX extension payloads, while the released tokio-tar API does not expose those limits. Keep the reviewed fork pin so untrusted archives cannot allocate unbounded extension metadata. Remove the fork pin after astral-sh/tokio-tar#118 is merged and a published tokio-tar release exposes the extension limits used here.
- `backlog-2102` rc.2/rc.3 empty scanner usage floor recovery: old DeleteBucket cleanup could synthesize an empty incomplete v2 usage primary/backup before leadership added an epoch, while newer scanners require a durable authoritative baseline identity. New scanners recognize only that exact serialized empty-fence shape, preserve its epoch through a CAS-protected recovery marker, and rebuild namespace coverage without treating zero usage as authoritative. Remove this recovery path and marker after rc.2 and rc.3 are no longer supported direct-upgrade sources. - `backlog-2102` rc.2/rc.3 empty scanner usage floor recovery: old DeleteBucket cleanup could synthesize an empty incomplete v2 usage primary/backup before leadership added an epoch, while newer scanners require a durable authoritative baseline identity. New scanners recognize only that exact serialized empty-fence shape, preserve its epoch through a CAS-protected recovery marker, and rebuild namespace coverage without treating zero usage as authoritative. Remove this recovery path and marker after rc.2 and rc.3 are no longer supported direct-upgrade sources.
- `s3gate-metadata-xml` persisted bucket XML migration: mixed-version site-replication peers, retained `.metadata.bin` objects, and backup archives can all carry XML written by the s3s codec, so the gateway migration must keep the legacy codec available until every stored form has crossed a verified rewrite boundary. Remove the legacy s3s parser and serializer only after the minimum supported direct-upgrade release reads and writes every persisted XML configuration family through the gateway codec, the four-way D1-D5 gate has remained clean for one full support window, every supported mixed-version site-replication topology has completed its writer upgrade, and migration tooling has verified or rewritten every retained bucket metadata object and restorable backup archive. - `s3gate-metadata-xml` persisted bucket XML migration: mixed-version site-replication peers, retained `.metadata.bin` objects, and backup archives can all carry XML written by the s3s codec, so the gateway migration must keep the legacy codec available until every stored form has crossed a verified rewrite boundary. Remove the legacy s3s parser and serializer only after the minimum supported direct-upgrade release reads and writes every persisted XML configuration family through the gateway codec, the four-way D1-D5 gate has remained clean for one full support window, every supported mixed-version site-replication topology has completed its writer upgrade, and migration tooling has verified or rewritten every retained bucket metadata object and restorable backup archive.
- `rustfs-6339` legacy bucket policy ID casing: earlier RustFS releases persisted the top-level policy identifier as "ID", while current writes use the S3-compatible "Id" spelling. Readers accept both spellings so retained bucket metadata remains usable after upgrade. Remove the legacy alias after migration tooling has rewritten every retained bucket policy using "ID". - `rustfs-6339` legacy bucket policy ID casing: earlier RustFS releases persisted the top-level policy identifier as "ID", while current writes use the S3-compatible "Id" spelling. Readers accept both spellings so retained bucket metadata remains usable after upgrade. Remove the legacy alias after migration tooling has rewritten every retained bucket policy using "ID".
File diff suppressed because it is too large Load Diff
+4
View File
@@ -178,6 +178,10 @@ use s3s::header::{X_AMZ_RESTORE, X_AMZ_RESTORE_OUTPUT_PATH};
use s3s::stream::{ByteStream, DynByteStream, RemainingLength}; use s3s::stream::{ByteStream, DynByteStream, RemainingLength};
use s3s::{S3Error, S3ErrorCode, S3Request, S3Response, S3Result, s3_error}; use s3s::{S3Error, S3ErrorCode, S3Request, S3Response, S3Result, s3_error};
fn object_s3_error(code: S3ErrorCode, message: impl Into<std::borrow::Cow<'static, str>>) -> S3Error {
S3Error::with_message(code, message)
}
mod copy; mod copy;
mod delete; mod delete;
mod extract; mod extract;