[PATCH 5.15 1/2] octeontx2-pf: Avoid use of GFP_KERNEL in atomic context

[PATCH 5.15 1/2] octeontx2-pf: Avoid use of GFP_KERNEL in atomic context

[Kevin Hao]

From: Geetha sowjanya <gakula@marvell.com>

commit 87b93b678e95c7d93fe6a55b0e0fbda26d8c7760 upstream.

Using GFP_KERNEL in preemption disable context, causing below warning
when CONFIG_DEBUG_ATOMIC_SLEEP is enabled.

[   32.542271] BUG: sleeping function called from invalid context at include/linux/sched/mm.h:274
[   32.550883] in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 1, name: swapper/0
[   32.558707] preempt_count: 1, expected: 0
[   32.562710] RCU nest depth: 0, expected: 0
[   32.566800] CPU: 3 PID: 1 Comm: swapper/0 Tainted: G        W          6.2.0-rc2-00269-gae9dcb91c606 #7
[   32.576188] Hardware name: Marvell CN106XX board (DT)
[   32.581232] Call trace:
[   32.583670]  dump_backtrace.part.0+0xe0/0xf0
[   32.587937]  show_stack+0x18/0x30
[   32.591245]  dump_stack_lvl+0x68/0x84
[   32.594900]  dump_stack+0x18/0x34
[   32.598206]  __might_resched+0x12c/0x160
[   32.602122]  __might_sleep+0x48/0xa0
[   32.605689]  __kmem_cache_alloc_node+0x2b8/0x2e0
[   32.610301]  __kmalloc+0x58/0x190
[   32.613610]  otx2_sq_aura_pool_init+0x1a8/0x314
[   32.618134]  otx2_open+0x1d4/0x9d0

To avoid use of GFP_ATOMIC for memory allocation, disable preemption
after all memory allocation is done.

Fixes: 4af1b64f80fb ("octeontx2-pf: Fix lmtst ID used in aura free")
Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
[Kevin: Fix the conflict due to the context change in v5.15]
Signed-off-by: Kevin Hao <haokexin@gmail.com>
---
 drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.c b/drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.c
index f6306eedd59b..30d4c0ad712d 100644
--- a/drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.c
+++ b/drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.c
@@ -1316,7 +1316,6 @@ int otx2_sq_aura_pool_init(struct otx2_nic *pfvf)
 	if (err)
 		goto fail;
 
-	get_cpu();
 	/* Allocate pointers and free them to aura/pool */
 	for (qidx = 0; qidx < hw->tx_queues; qidx++) {
 		pool_id = otx2_get_pool_idx(pfvf, AURA_NIX_SQ, qidx);
@@ -1334,13 +1333,14 @@ int otx2_sq_aura_pool_init(struct otx2_nic *pfvf)
 			err = otx2_alloc_rbuf(pfvf, pool, &bufptr);
 			if (err)
 				goto err_mem;
+			get_cpu();
 			pfvf->hw_ops->aura_freeptr(pfvf, pool_id, bufptr);
+			put_cpu();
 			sq->sqb_ptrs[sq->sqb_count++] = (u64)bufptr;
 		}
 	}
 
 err_mem:
-	put_cpu();
 	return err ? -ENOMEM : 0;
 
 fail:
-- 
2.38.1

[PATCH 5.15 2/2] octeontx2-pf: Fix the use of GFP_KERNEL in atomic context on rt

[Kevin Hao]

commit 55ba18dc62deff5910c0fa64486dea1ff20832ff upstream.

The commit 4af1b64f80fb ("octeontx2-pf: Fix lmtst ID used in aura
free") uses the get/put_cpu() to protect the usage of percpu pointer
in ->aura_freeptr() callback, but it also unnecessarily disable the
preemption for the blockable memory allocation. The commit 87b93b678e95
("octeontx2-pf: Avoid use of GFP_KERNEL in atomic context") tried to
fix these sleep inside atomic warnings. But it only fix the one for
the non-rt kernel. For the rt kernel, we still get the similar warnings
like below.
  BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:46
  in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 1, name: swapper/0
  preempt_count: 1, expected: 0
  RCU nest depth: 0, expected: 0
  3 locks held by swapper/0/1:
   #0: ffff800009fc5fe8 (rtnl_mutex){+.+.}-{3:3}, at: rtnl_lock+0x24/0x30
   #1: ffff000100c276c0 (&mbox->lock){+.+.}-{3:3}, at: otx2_init_hw_resources+0x8c/0x3a4
   #2: ffffffbfef6537e0 (&cpu_rcache->lock){+.+.}-{2:2}, at: alloc_iova_fast+0x1ac/0x2ac
  Preemption disabled at:
  [<ffff800008b1908c>] otx2_rq_aura_pool_init+0x14c/0x284
  CPU: 20 PID: 1 Comm: swapper/0 Tainted: G        W          6.2.0-rc3-rt1-yocto-preempt-rt #1
  Hardware name: Marvell OcteonTX CN96XX board (DT)
  Call trace:
   dump_backtrace.part.0+0xe8/0xf4
   show_stack+0x20/0x30
   dump_stack_lvl+0x9c/0xd8
   dump_stack+0x18/0x34
   __might_resched+0x188/0x224
   rt_spin_lock+0x64/0x110
   alloc_iova_fast+0x1ac/0x2ac
   iommu_dma_alloc_iova+0xd4/0x110
   __iommu_dma_map+0x80/0x144
   iommu_dma_map_page+0xe8/0x260
   dma_map_page_attrs+0xb4/0xc0
   __otx2_alloc_rbuf+0x90/0x150
   otx2_rq_aura_pool_init+0x1c8/0x284
   otx2_init_hw_resources+0xe4/0x3a4
   otx2_open+0xf0/0x610
   __dev_open+0x104/0x224
   __dev_change_flags+0x1e4/0x274
   dev_change_flags+0x2c/0x7c
   ic_open_devs+0x124/0x2f8
   ip_auto_config+0x180/0x42c
   do_one_initcall+0x90/0x4dc
   do_basic_setup+0x10c/0x14c
   kernel_init_freeable+0x10c/0x13c
   kernel_init+0x2c/0x140
   ret_from_fork+0x10/0x20

Of course, we can shuffle the get/put_cpu() to only wrap the invocation
of ->aura_freeptr() as what commit 87b93b678e95 does. But there are only
two ->aura_freeptr() callbacks, otx2_aura_freeptr() and
cn10k_aura_freeptr(). There is no usage of perpcu variable in the
otx2_aura_freeptr() at all, so the get/put_cpu() seems redundant to it.
We can move the get/put_cpu() into the corresponding callback which
really has the percpu variable usage and avoid the sprinkling of
get/put_cpu() in several places.

Fixes: 4af1b64f80fb ("octeontx2-pf: Fix lmtst ID used in aura free")
Signed-off-by: Kevin Hao <haokexin@gmail.com>
Link: https://lore.kernel.org/r/20230118071300.3271125-1-haokexin@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
---
 .../net/ethernet/marvell/octeontx2/nic/otx2_common.c  | 11 ++---------
 .../net/ethernet/marvell/octeontx2/nic/otx2_common.h  |  2 ++
 2 files changed, 4 insertions(+), 9 deletions(-)

diff --git a/drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.c b/drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.c
index 30d4c0ad712d..2e225309de9c 100644
--- a/drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.c
+++ b/drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.c
@@ -962,7 +962,6 @@ static void otx2_pool_refill_task(struct work_struct *work)
 	rbpool = cq->rbpool;
 	free_ptrs = cq->pool_ptrs;
 
-	get_cpu();
 	while (cq->pool_ptrs) {
 		if (otx2_alloc_rbuf(pfvf, rbpool, &bufptr)) {
 			/* Schedule a WQ if we fails to free atleast half of the
@@ -982,7 +981,6 @@ static void otx2_pool_refill_task(struct work_struct *work)
 		pfvf->hw_ops->aura_freeptr(pfvf, qidx, bufptr + OTX2_HEAD_ROOM);
 		cq->pool_ptrs--;
 	}
-	put_cpu();
 	cq->refill_task_sched = false;
 }
 
@@ -1333,9 +1331,7 @@ int otx2_sq_aura_pool_init(struct otx2_nic *pfvf)
 			err = otx2_alloc_rbuf(pfvf, pool, &bufptr);
 			if (err)
 				goto err_mem;
-			get_cpu();
 			pfvf->hw_ops->aura_freeptr(pfvf, pool_id, bufptr);
-			put_cpu();
 			sq->sqb_ptrs[sq->sqb_count++] = (u64)bufptr;
 		}
 	}
@@ -1381,21 +1377,18 @@ int otx2_rq_aura_pool_init(struct otx2_nic *pfvf)
 	if (err)
 		goto fail;
 
-	get_cpu();
 	/* Allocate pointers and free them to aura/pool */
 	for (pool_id = 0; pool_id < hw->rqpool_cnt; pool_id++) {
 		pool = &pfvf->qset.pool[pool_id];
 		for (ptr = 0; ptr < num_ptrs; ptr++) {
 			err = otx2_alloc_rbuf(pfvf, pool, &bufptr);
 			if (err)
-				goto err_mem;
+				return -ENOMEM;
 			pfvf->hw_ops->aura_freeptr(pfvf, pool_id,
 						   bufptr + OTX2_HEAD_ROOM);
 		}
 	}
-err_mem:
-	put_cpu();
-	return err ? -ENOMEM : 0;
+	return 0;
 fail:
 	otx2_mbox_reset(&pfvf->mbox.mbox, 0);
 	otx2_aura_pool_free(pfvf);
diff --git a/drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.h b/drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.h
index 095e5de78c0b..e685628b9294 100644
--- a/drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.h
+++ b/drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.h
@@ -605,8 +605,10 @@ static inline void cn10k_aura_freeptr(void *dev, int aura, u64 buf)
 	u64 ptrs[2];
 
 	ptrs[1] = buf;
+	get_cpu();
 	/* Free only one buffer at time during init and teardown */
 	__cn10k_aura_freeptr(pfvf, aura, ptrs, 2);
+	put_cpu();
 }
 
 /* Alloc pointer from pool/aura */
-- 
2.38.1

Re: [PATCH 5.15 1/2] octeontx2-pf: Avoid use of GFP_KERNEL in atomic context

[Sasha Levin]

On Mon, Jan 23, 2023 at 06:47:34AM +0800, Kevin Hao wrote:
>From: Geetha sowjanya <gakula@marvell.com>
>
>commit 87b93b678e95c7d93fe6a55b0e0fbda26d8c7760 upstream.

Queued up, thanks!

-- 
Thanks,
Sasha