The kernel build regressions/improvements email contained a couple of
issues with old compilers (in fact all the reports were on different
architectures, but all gcc 5.5) and the FIELD_PREP() and FIELD_GET()
conversions. They're all because an integer #define that should have
been declared as unsigned, was shifted to the point that it could set
the sign bit.
The fix just involves making sure the defines use the "U" identifier on
the constants to make sure they're unsigned. Should make the checkers
happier.
Confirmed with objdump before/after that there is no change to the
binaries.
Issues were reported as follows:
/drivers/net/ethernet/intel/ice/ice_base.c:238:7: note: in expansion of macro 'FIELD_GET'
(FIELD_GET(GLINT_CTL_ITR_GRAN_25_M, regval) == ICE_ITR_GRAN_US))
^
/include/linux/compiler_types.h:435:38: error: call to '__compiletime_assert_1093' declared with attribute error: FIELD_GET: mask is not constant
drivers/net/ethernet/intel/ice/ice_nvm.c:709:16: note: in expansion of macro ‘FIELD_GET’
orom->major = FIELD_GET(ICE_OROM_VER_MASK, combo_ver);
^
/include/linux/compiler_types.h:435:38: error: call to ‘__compiletime_assert_796’ declared with attribute error: FIELD_GET: mask is not constant
drivers/net/ethernet/intel/ice/ice_common.c:945:18: note: in expansion of macro ‘FIELD_GET’
u8 max_agg_bw = FIELD_GET(GL_PWR_MODE_CTL_CAR_MAX_BW_M,
^
/include/linux/compiler_types.h:435:38: error: call to ‘__compiletime_assert_420’ declared with attribute error: FIELD_GET: mask is not constant
drivers/net/ethernet/intel/i40e/i40e_dcb.c:458:8: note: in expansion of macro ‘FIELD_GET’
oui = FIELD_GET(I40E_LLDP_TLV_OUI_MASK, ouisubtype);
^
Reported-by: Geert Uytterhoeven <[email protected]>
Closes: https://lore.kernel.org/lkml/[email protected]/ #i40e #ice
Fixes: 62589808d73b ("i40e: field get conversion")
Fixes: 5a259f8e0baf ("ice: field get conversion")
Signed-off-by: Jesse Brandeburg <[email protected]>
---
drivers/net/ethernet/intel/i40e/i40e_dcb.h | 2 +-
drivers/net/ethernet/intel/ice/ice_osdep.h | 2 +-
drivers/net/ethernet/intel/ice/ice_type.h | 2 +-
3 files changed, 3 insertions(+), 3 deletions(-)
diff --git a/drivers/net/ethernet/intel/i40e/i40e_dcb.h b/drivers/net/ethernet/intel/i40e/i40e_dcb.h
index 6b60dc9b7736..d76497566e40 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_dcb.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_dcb.h
@@ -43,7 +43,7 @@
#define I40E_LLDP_TLV_SUBTYPE_SHIFT 0
#define I40E_LLDP_TLV_SUBTYPE_MASK (0xFF << I40E_LLDP_TLV_SUBTYPE_SHIFT)
#define I40E_LLDP_TLV_OUI_SHIFT 8
-#define I40E_LLDP_TLV_OUI_MASK (0xFFFFFF << I40E_LLDP_TLV_OUI_SHIFT)
+#define I40E_LLDP_TLV_OUI_MASK (0xFFFFFFU << I40E_LLDP_TLV_OUI_SHIFT)
/* Defines for IEEE ETS TLV */
#define I40E_IEEE_ETS_MAXTC_SHIFT 0
diff --git a/drivers/net/ethernet/intel/ice/ice_osdep.h b/drivers/net/ethernet/intel/ice/ice_osdep.h
index 82bc54fec7f3..a2562f04267f 100644
--- a/drivers/net/ethernet/intel/ice/ice_osdep.h
+++ b/drivers/net/ethernet/intel/ice/ice_osdep.h
@@ -24,7 +24,7 @@
#define rd64(a, reg) readq((a)->hw_addr + (reg))
#define ice_flush(a) rd32((a), GLGEN_STAT)
-#define ICE_M(m, s) ((m) << (s))
+#define ICE_M(m, s) ((m ## U) << (s))
struct ice_dma_mem {
void *va;
diff --git a/drivers/net/ethernet/intel/ice/ice_type.h b/drivers/net/ethernet/intel/ice/ice_type.h
index 41ab6d7bbd9e..a508e917ce5f 100644
--- a/drivers/net/ethernet/intel/ice/ice_type.h
+++ b/drivers/net/ethernet/intel/ice/ice_type.h
@@ -1072,7 +1072,7 @@ struct ice_aq_get_set_rss_lut_params {
#define ICE_OROM_VER_BUILD_SHIFT 8
#define ICE_OROM_VER_BUILD_MASK (0xffff << ICE_OROM_VER_BUILD_SHIFT)
#define ICE_OROM_VER_SHIFT 24
-#define ICE_OROM_VER_MASK (0xff << ICE_OROM_VER_SHIFT)
+#define ICE_OROM_VER_MASK (0xffU << ICE_OROM_VER_SHIFT)
#define ICE_SR_PFA_PTR 0x40
#define ICE_SR_1ST_NVM_BANK_PTR 0x42
#define ICE_SR_NVM_BANK_SIZE 0x43
--
2.39.3
Hello:
This patch was applied to netdev/net.git (main)
by Jakub Kicinski <[email protected]>:
On Mon, 5 Feb 2024 18:29:06 -0800 you wrote:
> The kernel build regressions/improvements email contained a couple of
> issues with old compilers (in fact all the reports were on different
> architectures, but all gcc 5.5) and the FIELD_PREP() and FIELD_GET()
> conversions. They're all because an integer #define that should have
> been declared as unsigned, was shifted to the point that it could set
> the sign bit.
>
> [...]
Here is the summary with links:
- [net,v1] net: intel: fix old compiler regressions
https://git.kernel.org/netdev/net/c/75428f537d7c
You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html