[PATCH] net: qed: reduce stack usage for TLV processing

Arnd Bergmann posted 1 patch 3 months, 2 weeks ago
drivers/net/ethernet/qlogic/qed/qed_mng_tlv.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
[PATCH] net: qed: reduce stack usage for TLV processing
Posted by Arnd Bergmann 3 months, 2 weeks ago
From: Arnd Bergmann <arnd@arndb.de>

clang gets a bit confused by the code in the qed_mfw_process_tlv_req and
ends up spilling registers to the stack hundreds of times. When sanitizers
are enabled, this can end up blowing the stack warning limit:

drivers/net/ethernet/qlogic/qed/qed_mng_tlv.c:1244:5: error: stack frame size (1824) exceeds limit (1280) in 'qed_mfw_process_tlv_req' [-Werror,-Wframe-larger-than]

Apparently the problem is the complexity of qed_mfw_update_tlvs()
after inlining, and marking the four main branches of that function
as noinline_for_stack makes this problem completely go away, the stack
usage goes down to 100 bytes.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
If anyone feels adventurous and able to figure out what exactly goes
wrong in clang, I can provide preprocessed source files for debugging.
---
 drivers/net/ethernet/qlogic/qed/qed_mng_tlv.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/qlogic/qed/qed_mng_tlv.c b/drivers/net/ethernet/qlogic/qed/qed_mng_tlv.c
index f55eed092f25..7d78f072b0a1 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_mng_tlv.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_mng_tlv.c
@@ -242,7 +242,7 @@ static int qed_mfw_get_tlv_group(u8 tlv_type, u8 *tlv_group)
 }
 
 /* Returns size of the data buffer or, -1 in case TLV data is not available. */
-static int
+static noinline_for_stack int
 qed_mfw_get_gen_tlv_value(struct qed_drv_tlv_hdr *p_tlv,
 			  struct qed_mfw_tlv_generic *p_drv_buf,
 			  struct qed_tlv_parsed_buf *p_buf)
@@ -304,7 +304,7 @@ qed_mfw_get_gen_tlv_value(struct qed_drv_tlv_hdr *p_tlv,
 	return -1;
 }
 
-static int
+static noinline_for_stack int
 qed_mfw_get_eth_tlv_value(struct qed_drv_tlv_hdr *p_tlv,
 			  struct qed_mfw_tlv_eth *p_drv_buf,
 			  struct qed_tlv_parsed_buf *p_buf)
@@ -438,7 +438,7 @@ qed_mfw_get_tlv_time_value(struct qed_mfw_tlv_time *p_time,
 	return QED_MFW_TLV_TIME_SIZE;
 }
 
-static int
+static noinline_for_stack int
 qed_mfw_get_fcoe_tlv_value(struct qed_drv_tlv_hdr *p_tlv,
 			   struct qed_mfw_tlv_fcoe *p_drv_buf,
 			   struct qed_tlv_parsed_buf *p_buf)
@@ -1073,7 +1073,7 @@ qed_mfw_get_fcoe_tlv_value(struct qed_drv_tlv_hdr *p_tlv,
 	return -1;
 }
 
-static int
+static noinline_for_stack int
 qed_mfw_get_iscsi_tlv_value(struct qed_drv_tlv_hdr *p_tlv,
 			    struct qed_mfw_tlv_iscsi *p_drv_buf,
 			    struct qed_tlv_parsed_buf *p_buf)
-- 
2.39.5
Re: [PATCH] net: qed: reduce stack usage for TLV processing
Posted by Alexander Lobakin 3 months, 2 weeks ago
From: Arnd Bergmann <arnd@kernel.org>
Date: Fri, 20 Jun 2025 15:09:53 +0200

> From: Arnd Bergmann <arnd@arndb.de>
> 
> clang gets a bit confused by the code in the qed_mfw_process_tlv_req and
> ends up spilling registers to the stack hundreds of times. When sanitizers
> are enabled, this can end up blowing the stack warning limit:
> 
> drivers/net/ethernet/qlogic/qed/qed_mng_tlv.c:1244:5: error: stack frame size (1824) exceeds limit (1280) in 'qed_mfw_process_tlv_req' [-Werror,-Wframe-larger-than]
> 
> Apparently the problem is the complexity of qed_mfw_update_tlvs()
> after inlining, and marking the four main branches of that function
> as noinline_for_stack makes this problem completely go away, the stack
> usage goes down to 100 bytes.
> 
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>

Reviewed-by: Alexander Lobakin <aleksander.lobakin@intel.com>

Thanks,
Olek