[pbs-devel] [PATCH proxmox v3 2/4] proxmox-access-control: cache verified API token secrets

[Samuel Rufinatscha <s.rufinatscha@proxmox.com>]

Currently, every token-based API request reads the token.shadow file and
runs the expensive password hash verification for the given token
secret. This issue was first observed as part of profiling the PBS
/status endpoint (see bug #7017 [1]) and is required for the factored
out proxmox_access_control token_shadow implementation too.

This patch introduces an in-memory cache of successfully verified token
secrets. Subsequent requests for the same token+secret combination only
perform a comparison using openssl::memcmp::eq and avoid re-running the
password hash. The cache is updated when a token secret is set and
cleared when a token is deleted. Note, this does NOT include manual
config changes, which will be covered in a subsequent patch.

This patch is part of the series which fixes bug #7017 [1].

[1] https://bugzilla.proxmox.com/show_bug.cgi?id=7017

Signed-off-by: Samuel Rufinatscha <s.rufinatscha@proxmox.com>
---
Changes from v1 to v2:

* Replace OnceCell with LazyLock, and std::sync::RwLock with
parking_lot::RwLock.
* Add API_MUTATION_GENERATION and guard cache inserts
to prevent “zombie inserts” across concurrent set/delete.
* Refactor cache operations into cache_try_secret_matches,
cache_try_insert_secret, and centralize write-side behavior in
apply_api_mutation.
* Switch fast-path cache access to try_read/try_write (best-effort).

Changes from v2 to v3:

* Replaced process-local cache invalidation (AtomicU64
API_MUTATION_GENERATION) with a cross-process shared generation via
ConfigVersionCache.
* Validate shared generation before/after the constant-time secret
compare; only insert into cache if the generation is unchanged.
* invalidate_cache_state() on insert if shared generation changed.

 Cargo.toml                                 |   1 +
 proxmox-access-control/Cargo.toml          |   1 +
 proxmox-access-control/src/token_shadow.rs | 154 ++++++++++++++++++++-
 3 files changed, 155 insertions(+), 1 deletion(-)

diff --git a/Cargo.toml b/Cargo.toml
index 27a69afa..59a2ec93 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -112,6 +112,7 @@ native-tls = "0.2"
 nix = "0.29"
 openssl = "0.10"
 pam-sys = "0.5"
+parking_lot = "0.12"
 percent-encoding = "2.1"
 pin-utils = "0.1.0"
 proc-macro2 = "1.0"
diff --git a/proxmox-access-control/Cargo.toml b/proxmox-access-control/Cargo.toml
index ec189664..1de2842c 100644
--- a/proxmox-access-control/Cargo.toml
+++ b/proxmox-access-control/Cargo.toml
@@ -16,6 +16,7 @@ anyhow.workspace = true
 const_format.workspace = true
 nix = { workspace = true, optional = true }
 openssl = { workspace = true, optional = true }
+parking_lot.workspace = true
 regex.workspace = true
 hex = { workspace = true, optional = true }
 serde.workspace = true
diff --git a/proxmox-access-control/src/token_shadow.rs b/proxmox-access-control/src/token_shadow.rs
index c586d834..895309d2 100644
--- a/proxmox-access-control/src/token_shadow.rs
+++ b/proxmox-access-control/src/token_shadow.rs
@@ -1,13 +1,28 @@
 use std::collections::HashMap;
+use std::sync::LazyLock;
 
 use anyhow::{bail, format_err, Error};
+use parking_lot::RwLock;
 use serde_json::{from_value, Value};
 
 use proxmox_auth_api::types::Authid;
 use proxmox_product_config::{open_api_lockfile, replace_config, ApiLockGuard};
 
+use crate::init::access_conf;
 use crate::init::impl_feature::{token_shadow, token_shadow_lock};
 
+/// Global in-memory cache for successfully verified API token secrets.
+/// The cache stores plain text secrets for token Authids that have already been
+/// verified against the hashed values in `token.shadow`. This allows for cheap
+/// subsequent authentications for the same token+secret combination, avoiding
+/// recomputing the password hash on every request.
+static TOKEN_SECRET_CACHE: LazyLock<RwLock<ApiTokenSecretCache>> = LazyLock::new(|| {
+    RwLock::new(ApiTokenSecretCache {
+        secrets: HashMap::new(),
+        shared_gen: 0,
+    })
+});
+
 // Get exclusive lock
 fn lock_config() -> Result<ApiLockGuard, Error> {
     open_api_lockfile(token_shadow_lock(), None, true)
@@ -36,9 +51,27 @@ pub fn verify_secret(tokenid: &Authid, secret: &str) -> Result<(), Error> {
         bail!("not an API token ID");
     }
 
+    // Fast path
+    if cache_try_secret_matches(tokenid, secret) {
+        return Ok(());
+    }
+
+    // Slow path
+    // First, capture the shared generation before doing the hash verification.
+    let gen_before = token_shadow_shared_gen();
+
     let data = read_file()?;
     match data.get(tokenid) {
-        Some(hashed_secret) => proxmox_sys::crypt::verify_crypt_pw(secret, hashed_secret),
+        Some(hashed_secret) => {
+            proxmox_sys::crypt::verify_crypt_pw(secret, hashed_secret)?;
+
+            // Try to cache only if nothing changed while verifying the secret.
+            if let Some(gen) = gen_before {
+                cache_try_insert_secret(tokenid.clone(), secret.to_owned(), gen);
+            }
+
+            Ok(())
+        }
         None => bail!("invalid API token"),
     }
 }
@@ -56,6 +89,8 @@ pub fn set_secret(tokenid: &Authid, secret: &str) -> Result<(), Error> {
     data.insert(tokenid.clone(), hashed_secret);
     write_file(data)?;
 
+    apply_api_mutation(tokenid, Some(secret));
+
     Ok(())
 }
 
@@ -71,6 +106,8 @@ pub fn delete_secret(tokenid: &Authid) -> Result<(), Error> {
     data.remove(tokenid);
     write_file(data)?;
 
+    apply_api_mutation(tokenid, None);
+
     Ok(())
 }
 
@@ -81,3 +118,118 @@ pub fn generate_and_set_secret(tokenid: &Authid) -> Result<String, Error> {
     set_secret(tokenid, &secret)?;
     Ok(secret)
 }
+
+struct ApiTokenSecretCache {
+    /// Keys are token Authids, values are the corresponding plain text secrets.
+    /// Entries are added after a successful on-disk verification in
+    /// `verify_secret` or when a new token secret is generated by
+    /// `generate_and_set_secret`. Used to avoid repeated
+    /// password-hash computation on subsequent authentications.
+    secrets: HashMap<Authid, CachedSecret>,
+    /// Shared generation to detect mutations of the underlying token.shadow file.
+    shared_gen: usize,
+}
+
+/// Cached secret.
+struct CachedSecret {
+    secret: String,
+}
+
+fn cache_try_insert_secret(tokenid: Authid, secret: String, shared_gen_before: usize) {
+    let Some(mut cache) = TOKEN_SECRET_CACHE.try_write() else {
+        return;
+    };
+
+    let Some(shared_gen_now) = token_shadow_shared_gen() else {
+        return;
+    };
+
+    // If this process missed a generation bump, its cache is stale.
+    if cache.shared_gen != shared_gen_now {
+        invalidate_cache_state(&mut cache);
+        cache.shared_gen = shared_gen_now;
+    }
+
+    // If a mutation happened while we were verifying the secret, do not insert.
+    if shared_gen_now == shared_gen_before {
+        cache.secrets.insert(tokenid, CachedSecret { secret });
+    }
+}
+
+// Tries to match the given token secret against the cached secret.
+// Checks the generation before and after the constant-time compare to avoid a
+// TOCTOU window. If another process rotates/deletes a token while we're validating
+// the cached secret, the generation will change, and we
+// must not trust the cache for this request.
+fn cache_try_secret_matches(tokenid: &Authid, secret: &str) -> bool {
+    let Some(cache) = TOKEN_SECRET_CACHE.try_read() else {
+        return false;
+    };
+    let Some(entry) = cache.secrets.get(tokenid) else {
+        return false;
+    };
+
+    let cache_gen = cache.shared_gen;
+
+    let Some(gen1) = token_shadow_shared_gen() else {
+        return false;
+    };
+    if gen1 != cache_gen {
+        return false;
+    }
+
+    let eq = openssl::memcmp::eq(entry.secret.as_bytes(), secret.as_bytes());
+
+    let Some(gen2) = token_shadow_shared_gen() else {
+        return false;
+    };
+
+    eq && gen2 == cache_gen
+}
+
+fn apply_api_mutation(tokenid: &Authid, new_secret: Option<&str>) {
+    // Signal cache invalidation to other processes (best-effort).
+    let new_shared_gen = bump_token_shadow_shared_gen();
+
+    let mut cache = TOKEN_SECRET_CACHE.write();
+
+    // If we cannot read/bump the shared generation, we cannot safely trust the cache.
+    let Some(gen) = new_shared_gen else {
+        invalidate_cache_state(&mut cache);
+        cache.shared_gen = 0;
+        return;
+    };
+
+    // Update to the post-mutation generation.
+    cache.shared_gen = gen;
+
+    // Apply the new mutation.
+    match new_secret {
+        Some(secret) => {
+            cache.secrets.insert(
+                tokenid.clone(),
+                CachedSecret {
+                    secret: secret.to_owned(),
+                },
+            );
+        }
+        None => {
+            cache.secrets.remove(tokenid);
+        }
+    }
+}
+
+/// Get the current shared generation.
+fn token_shadow_shared_gen() -> Option<usize> {
+    access_conf().token_shadow_cache_generation()
+}
+
+/// Bump and return the new shared generation.
+fn bump_token_shadow_shared_gen() -> Option<usize> {
+    access_conf().increment_token_shadow_cache_generation().ok().map(|prev| prev + 1)
+}
+
+/// Invalidates the cache state and only keeps the shared generation.
+fn invalidate_cache_state(cache: &mut ApiTokenSecretCache) {
+    cache.secrets.clear();
+}
\ No newline at end of file
-- 
2.47.3



_______________________________________________
pbs-devel mailing list
pbs-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pbs-devel