Skip to contentSkip to conversation

tcp: configurable source port perturb table size

PATCH Applied Gleb Mazovetskiy 3 participants 3 messages Nov 14, 2022 – Nov 16, 2022

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tcp: configurable source port perturb table size

On embedded systems with little memory and no relevant security concerns, it is beneficial to reduce the size of the table.

Reducing the size from 2^16 to 2^8 saves 255 KiB
of kernel RAM.

Makes the table size configurable as an expert option.

The size was previously increased from 2^8 to 2^16
in commit 4c2c8f03a5ab ("tcp: increase source port perturb table to 2^16").

---
 net/ipv4/Kconfig           | 10 ++++++++++
 net/ipv4/inet_hashtables.c | 10 +++++-----
 2 files changed, 15 insertions(+), 5 deletions(-)
net/ipv4/Kconfig +10 -0
index e983bb0c5012..2dfb12230f08 100644
@@ -402,6 +402,16 @@ config INET_IPCOMP
 
 	  If unsure, say Y.
 
+config INET_TABLE_PERTURB_ORDER
+	int "INET: Source port perturbation table size (as power of 2)" if EXPERT
+	default 16
+	help
+	  Source port perturbation table size (as power of 2) for
+	  RFC 6056 3.3.4.  Algorithm 4: Double-Hash Port Selection Algorithm.
+
+	  The default is almost always what you want.
+	  Only change this if you know what you are doing.
+
 config INET_XFRM_TUNNEL
 	tristate
 	select INET_TUNNEL
net/ipv4/inet_hashtables.c +5 -5
index d3dc28156622..033bf3c2538f 100644
@@ -906,13 +906,13 @@ EXPORT_SYMBOL_GPL(inet_bhash2_update_saddr);
  * Note that we use 32bit integers (vs RFC 'short integers')
  * because 2^16 is not a multiple of num_ephemeral and this
  * property might be used by clever attacker.
+ *
  * RFC claims using TABLE_LENGTH=10 buckets gives an improvement, though
- * attacks were since demonstrated, thus we use 65536 instead to really
- * give more isolation and privacy, at the expense of 256kB of kernel
- * memory.
+ * attacks were since demonstrated, thus we use 65536 by default instead
+ * to really give more isolation and privacy, at the expense of 256kB
+ * of kernel memory.
  */
-#define INET_TABLE_PERTURB_SHIFT 16
-#define INET_TABLE_PERTURB_SIZE (1 << INET_TABLE_PERTURB_SHIFT)
+#define INET_TABLE_PERTURB_SIZE (1 << CONFIG_INET_TABLE_PERTURB_ORDER)
 static u32 *table_perturb;
 
 int __inet_hash_connect(struct inet_timewait_death_row *death_row,
Signature · 1 line
2.37.2
Reviewed-by on v1 1/1
#1 lore raw
From:   Gleb Mazovetskiy <[email protected]>
Date:   Mon, 14 Nov 2022 22:56:16 +0000
Quoted 7 lines

On embedded systems with little memory and no relevant security concerns, it is beneficial to reduce the size of the table.

Reducing the size from 2^16 to 2^8 saves 255 KiB
of kernel RAM.

Quoted 7 lines

Makes the table size configurable as an expert option.

The size was previously increased from 2^8 to 2^16
in commit 4c2c8f03a5ab ("tcp: increase source port perturb table to 2^16").

Looks good.

Quoted 51 lines
---
 net/ipv4/Kconfig           | 10 ++++++++++
 net/ipv4/inet_hashtables.c | 10 +++++-----
 2 files changed, 15 insertions(+), 5 deletions(-)
net/ipv4/Kconfig +10 -0
index e983bb0c5012..2dfb12230f08 100644
@@ -402,6 +402,16 @@ config INET_IPCOMP
 
 	  If unsure, say Y.
 
+config INET_TABLE_PERTURB_ORDER
+	int "INET: Source port perturbation table size (as power of 2)" if EXPERT
+	default 16
+	help
+	  Source port perturbation table size (as power of 2) for
+	  RFC 6056 3.3.4.  Algorithm 4: Double-Hash Port Selection Algorithm.
+
+	  The default is almost always what you want.
+	  Only change this if you know what you are doing.
+
 config INET_XFRM_TUNNEL
 	tristate
 	select INET_TUNNEL
net/ipv4/inet_hashtables.c +5 -5
index d3dc28156622..033bf3c2538f 100644
@@ -906,13 +906,13 @@ EXPORT_SYMBOL_GPL(inet_bhash2_update_saddr);
  * Note that we use 32bit integers (vs RFC 'short integers')
  * because 2^16 is not a multiple of num_ephemeral and this
  * property might be used by clever attacker.
+ *
  * RFC claims using TABLE_LENGTH=10 buckets gives an improvement, though
- * attacks were since demonstrated, thus we use 65536 instead to really
- * give more isolation and privacy, at the expense of 256kB of kernel
- * memory.
+ * attacks were since demonstrated, thus we use 65536 by default instead
+ * to really give more isolation and privacy, at the expense of 256kB
+ * of kernel memory.
  */
-#define INET_TABLE_PERTURB_SHIFT 16
-#define INET_TABLE_PERTURB_SIZE (1 << INET_TABLE_PERTURB_SHIFT)
+#define INET_TABLE_PERTURB_SIZE (1 << CONFIG_INET_TABLE_PERTURB_ORDER)
 static u32 *table_perturb;
 
 int __inet_hash_connect(struct inet_timewait_death_row *death_row,
Signature · 1 line
2.37.2
#2 lore raw
In reply to Gleb Mazovetskiy
This patch was applied to netdev/net.git (master) by David S. Miller <[email protected]>:

Hello:

This patch was applied to netdev/net.git (master)
by David S. Miller <[email protected]>:

Quoted 8 lines · On Mon, 14 Nov 2022 22:56:16 +0000 you wrote:

On embedded systems with little memory and no relevant security concerns, it is beneficial to reduce the size of the table.

Reducing the size from 2^16 to 2^8 saves 255 KiB
of kernel RAM.

[...]

Here is the summary with links:
- [1/1] tcp: configurable source port perturb table size

    https://git.kernel.org/netdev/net/c/aeac4ec8f46d

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html