From: Arnd Bergmann arnd@arndb.de
[ Upstream commit 24171a5a4a952c26568ff0d2a0bc8c4708a95e1d ]
With large values of CONFIG_NR_CPUS, three Intel ethernet drivers fail to compile like:
In function ‘i40e_free_q_vector’, inlined from ‘i40e_vsi_alloc_q_vectors’ at drivers/net/ethernet/intel/i40e/i40e_main.c:12112:3: 571 | _compiletime_assert(condition, msg, __compiletime_assert_, __COUNTER__) include/linux/rcupdate.h:1084:17: note: in expansion of macro ‘BUILD_BUG_ON’ 1084 | BUILD_BUG_ON(offsetof(typeof(*(ptr)), rhf) >= 4096); \ drivers/net/ethernet/intel/i40e/i40e_main.c:5113:9: note: in expansion of macro ‘kfree_rcu’ 5113 | kfree_rcu(q_vector, rcu); | ^~~~~~~~~
The problem is that the 'rcu' member in 'q_vector' is too far from the start of the structure. Move this member before the CPU mask instead, in all three drivers.
Signed-off-by: Arnd Bergmann arnd@arndb.de Acked-by: David S. Miller davem@davemloft.net Reviewed-by: Aleksandr Loktionov aleksandr.loktionov@intel.com Reviewed-by: Alexander Lobakin aleksander.lobakin@intel.com Tested-by: Sunitha Mekala sunithax.d.mekala@intel.com (A Contingent worker at Intel) Signed-off-by: Tony Nguyen anthony.l.nguyen@intel.com Signed-off-by: Sasha Levin sashal@kernel.org ---
LLM Generated explanations, may be completely bogus:
**YES**
This commit should be backported to stable kernel trees for the following reasons:
1. **Compilation Failure Fix**: This fixes a build-time compilation failure that occurs with large values of CONFIG_NR_CPUS. The error prevents the kernel from building successfully, which is a significant issue for systems configured with high CPU counts.
2. **Root Cause**: The issue stems from the `kfree_rcu()` macro in `include/linux/rcupdate.h:1084` which includes a `BUILD_BUG_ON(offsetof(typeof(*(ptr)), rhf) >= 4096)` check. This compile-time assertion ensures that the RCU head field is within the first 4096 bytes of the structure.
3. **Simple and Safe Fix**: The fix is straightforward - it moves the `struct rcu_head rcu` member earlier in the structure, placing it before the `cpumask_t affinity_mask` member. This is a safe change because: - It only reorders structure members - The RCU head doesn't need to be at any specific location functionally - Moving it earlier reduces the offset from the structure start - No functional code changes are required
4. **Affects Multiple Drivers**: The issue affects three Intel ethernet drivers (i40e, ixgbe, fm10k), all of which are fixed identically by moving the rcu member earlier in their respective q_vector structures.
5. **Configuration-Dependent Bug**: This bug only manifests with large CONFIG_NR_CPUS values (likely >= 512 or higher), where the `cpumask_t` type becomes large enough to push the rcu member beyond the 4096-byte offset limit. Systems with high core counts are becoming more common in production environments.
6. **No Risk of Regression**: The change is minimal and doesn't alter any functionality. It's purely a structural reordering that maintains all existing behavior while fixing the compilation issue.
7. **Meets Stable Criteria**: This fix clearly meets the stable kernel criteria as it: - Fixes a real bug (compilation failure) - Is minimal and contained - Has no risk of introducing new issues - Affects users with legitimate configurations
The commit is an excellent candidate for stable backporting as it fixes a real compilation issue that prevents kernel builds on systems with large CPU counts, and the fix is trivial with no risk of regression.
drivers/net/ethernet/intel/fm10k/fm10k.h | 3 ++- drivers/net/ethernet/intel/i40e/i40e.h | 2 +- drivers/net/ethernet/intel/ixgbe/ixgbe.h | 3 ++- 3 files changed, 5 insertions(+), 3 deletions(-)
diff --git a/drivers/net/ethernet/intel/fm10k/fm10k.h b/drivers/net/ethernet/intel/fm10k/fm10k.h index b14441944b4b2..db72e5e4acd36 100644 --- a/drivers/net/ethernet/intel/fm10k/fm10k.h +++ b/drivers/net/ethernet/intel/fm10k/fm10k.h @@ -189,13 +189,14 @@ struct fm10k_q_vector { struct fm10k_ring_container rx, tx;
struct napi_struct napi; + struct rcu_head rcu; /* to avoid race with update stats on free */ + cpumask_t affinity_mask; char name[IFNAMSIZ + 9];
#ifdef CONFIG_DEBUG_FS struct dentry *dbg_q_vector; #endif /* CONFIG_DEBUG_FS */ - struct rcu_head rcu; /* to avoid race with update stats on free */
/* for dynamic allocation of rings associated with this q_vector */ struct fm10k_ring ring[0] ____cacheline_internodealigned_in_smp; diff --git a/drivers/net/ethernet/intel/i40e/i40e.h b/drivers/net/ethernet/intel/i40e/i40e.h index 4c8c31692e9e0..edb5e408c980f 100644 --- a/drivers/net/ethernet/intel/i40e/i40e.h +++ b/drivers/net/ethernet/intel/i40e/i40e.h @@ -864,6 +864,7 @@ struct i40e_q_vector { u16 reg_idx; /* register index of the interrupt */
struct napi_struct napi; + struct rcu_head rcu; /* to avoid race with update stats on free */
struct i40e_ring_container rx; struct i40e_ring_container tx; @@ -874,7 +875,6 @@ struct i40e_q_vector { cpumask_t affinity_mask; struct irq_affinity_notify affinity_notify;
- struct rcu_head rcu; /* to avoid race with update stats on free */ char name[I40E_INT_NAME_STR_LEN]; bool arm_wb_state; } ____cacheline_internodealigned_in_smp; diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe.h b/drivers/net/ethernet/intel/ixgbe/ixgbe.h index 0142ca226bf50..0f9a1fce0dee9 100644 --- a/drivers/net/ethernet/intel/ixgbe/ixgbe.h +++ b/drivers/net/ethernet/intel/ixgbe/ixgbe.h @@ -458,9 +458,10 @@ struct ixgbe_q_vector { struct ixgbe_ring_container rx, tx;
struct napi_struct napi; + struct rcu_head rcu; /* to avoid race with update stats on free */ + cpumask_t affinity_mask; int numa_node; - struct rcu_head rcu; /* to avoid race with update stats on free */ char name[IFNAMSIZ + 9];
/* for dynamic allocation of rings associated with this q_vector */