[PATCH] infiniband: change some kmalloc to kvmalloc to support CONFIG_PROVE_LOCKING=y

[PATCH] infiniband: change some kmalloc to kvmalloc to support CONFIG_PROVE_LOCKING=y

[wangyugui]

When CONFIG_PROVE_LOCKING=y, one kmalloc of infiniband hit the max alloc size limitation.

WARNING: CPU: 36 PID: 8 at mm/page_alloc.c:5350 __alloc_pages+0x27e/0x3e0
 Call Trace:
  kmalloc_order+0x2a/0xb0
  kmalloc_order_trace+0x19/0xf0
  __kmalloc+0x231/0x270
  ib_setup_port_attrs+0xd8/0x870 [ib_core]
  ib_register_device+0x419/0x4e0 [ib_core]
  bnxt_re_task+0x208/0x2d0 [bnxt_re]

change this kmalloc to kvmalloc to support CONFIG_PROVE_LOCKING=y

Signed-off-by: wangyugui <wangyugui@e16-tech.com>
---
 drivers/infiniband/core/sysfs.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/infiniband/core/sysfs.c b/drivers/infiniband/core/sysfs.c
index 6146c3c1cbe5..8d709986b88c 100644
--- a/drivers/infiniband/core/sysfs.c
+++ b/drivers/infiniband/core/sysfs.c
@@ -757,7 +757,7 @@ static void ib_port_release(struct kobject *kobj)
 	if (port->hw_stats_data)
 		kfree(port->hw_stats_data->stats);
 	kfree(port->hw_stats_data);
-	kfree(port);
+	kvfree(port);
 }
 
 static void ib_port_gid_attr_release(struct kobject *kobj)
@@ -1189,7 +1189,7 @@ static struct ib_port *setup_port(struct ib_core_device *coredev, int port_num,
 	struct ib_port *p;
 	int ret;
 
-	p = kzalloc(struct_size(p, attrs_list,
+	p = kvzalloc(struct_size(p, attrs_list,
 				attr->gid_tbl_len + attr->pkey_tbl_len),
 		    GFP_KERNEL);
 	if (!p)
-- 
2.32.0

Re: [PATCH] infiniband: change some kmalloc to kvmalloc to support CONFIG_PROVE_LOCKING=y

[Jason Gunthorpe]

On Tue, Oct 19, 2021 at 08:26:56AM +0800, wangyugui wrote:
> When CONFIG_PROVE_LOCKING=y, one kmalloc of infiniband hit the max alloc size limitation.
> 
> WARNING: CPU: 36 PID: 8 at mm/page_alloc.c:5350 __alloc_pages+0x27e/0x3e0
>  Call Trace:
>   kmalloc_order+0x2a/0xb0
>   kmalloc_order_trace+0x19/0xf0
>   __kmalloc+0x231/0x270
>   ib_setup_port_attrs+0xd8/0x870 [ib_core]
>   ib_register_device+0x419/0x4e0 [ib_core]
>   bnxt_re_task+0x208/0x2d0 [bnxt_re]
> 
> change this kmalloc to kvmalloc to support CONFIG_PROVE_LOCKING=y
> 
> Signed-off-by: wangyugui <wangyugui@e16-tech.com>
> ---
>  drivers/infiniband/core/sysfs.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)

Huh? what causes ib_port to get larger than MAX_ORDER?

The only array is attrs_list and I don't see something that scales
with

Jason

Re: [PATCH] infiniband: change some kmalloc to kvmalloc to support CONFIG_PROVE_LOCKING=y

[Wang Yugui]

Hi,

> On Tue, Oct 19, 2021 at 08:26:56AM +0800, wangyugui wrote:
> > When CONFIG_PROVE_LOCKING=y, one kmalloc of infiniband hit the max alloc size limitation.
> > 
> > WARNING: CPU: 36 PID: 8 at mm/page_alloc.c:5350 __alloc_pages+0x27e/0x3e0
> >  Call Trace:
> >   kmalloc_order+0x2a/0xb0
> >   kmalloc_order_trace+0x19/0xf0
> >   __kmalloc+0x231/0x270
> >   ib_setup_port_attrs+0xd8/0x870 [ib_core]
> >   ib_register_device+0x419/0x4e0 [ib_core]
> >   bnxt_re_task+0x208/0x2d0 [bnxt_re]
> > 
> > change this kmalloc to kvmalloc to support CONFIG_PROVE_LOCKING=y
> > 
> > Signed-off-by: wangyugui <wangyugui@e16-tech.com>
> > ---
> >  drivers/infiniband/core/sysfs.c | 4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> Huh? what causes ib_port to get larger than MAX_ORDER?
> 
> The only array is attrs_list and I don't see something that scales
> with

The array size is not fixed. so it maybe a little big in some case.

struct ib_port {
	struct kobject kobj;
	struct ib_device *ibdev;
	struct gid_attr_group *gid_attr_group;
	struct hw_stats_port_data *hw_stats_data;

	struct attribute_group groups[3];
	const struct attribute_group *groups_list[5];
	u32 port_num;
	struct port_table_attribute attrs_list[];
};

=>
struct port_table_attribute {
	struct ib_port_attribute attr;

=>
struct ib_port_attribute {
	struct attribute attr;

=>
struct attribute {
	const char		*name;
	umode_t			mode;
#ifdef CONFIG_DEBUG_LOCK_ALLOC
	bool			ignore_lockdep:1;
	struct lock_class_key	*key;
	struct lock_class_key	skey;
#endif
};

When CONFIG_PROVE_LOCKING=y, we need CONFIG_DEBUG_LOCK_ALLOC=y too.

This problem happen when CONFIG_PROVE_LOCKING=y(CONFIG_DEBUG_LOCK_ALLOC=y),
but not happen when CONFIG_PROVE_LOCKING=n(CONFIG_DEBUG_LOCK_ALLOC=n).

Best Regards
Wang Yugui (wangyugui@e16-tech.com)
2021/10/19

Re: [PATCH] infiniband: change some kmalloc to kvmalloc to support CONFIG_PROVE_LOCKING=y

[Jason Gunthorpe]

On Tue, Oct 19, 2021 at 08:26:56AM +0800, wangyugui wrote:
> When CONFIG_PROVE_LOCKING=y, one kmalloc of infiniband hit the max alloc size limitation.
> 
> WARNING: CPU: 36 PID: 8 at mm/page_alloc.c:5350 __alloc_pages+0x27e/0x3e0
>  Call Trace:
>   kmalloc_order+0x2a/0xb0
>   kmalloc_order_trace+0x19/0xf0
>   __kmalloc+0x231/0x270
>   ib_setup_port_attrs+0xd8/0x870 [ib_core]
>   ib_register_device+0x419/0x4e0 [ib_core]
>   bnxt_re_task+0x208/0x2d0 [bnxt_re]
> 
> change this kmalloc to kvmalloc to support CONFIG_PROVE_LOCKING=y
> 
> Signed-off-by: wangyugui <wangyugui@e16-tech.com>
> ---
>  drivers/infiniband/core/sysfs.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)

Applied to for-next, thanks

Jason