Re: [PATCH v4] net/ipv4/inet_fragment: Batch fqdir destroy works

2020-12-12 Thread Jakub Kicinski
On Fri, 11 Dec 2020 15:36:53 +0100 Eric Dumazet wrote: > On Fri, Dec 11, 2020 at 12:24 PM SeongJae Park wrote: > > From: SeongJae Park > > > > On a few of our systems, I found frequent 'unshare(CLONE_NEWNET)' calls > > make the number of active slab objects including 'sock_inode_cache' type > > r

Re: [PATCH v4] net/ipv4/inet_fragment: Batch fqdir destroy works

2020-12-11 Thread SeongJae Park
On Fri, 11 Dec 2020 15:36:53 +0100 Eric Dumazet wrote: > On Fri, Dec 11, 2020 at 12:24 PM SeongJae Park wrote: > > > > From: SeongJae Park > > > > On a few of our systems, I found frequent 'unshare(CLONE_NEWNET)' calls > > make the number of active slab objects including 'sock_inode_cache' type

Re: [PATCH v4] net/ipv4/inet_fragment: Batch fqdir destroy works

2020-12-11 Thread Eric Dumazet
On Fri, Dec 11, 2020 at 12:24 PM SeongJae Park wrote: > > From: SeongJae Park > > On a few of our systems, I found frequent 'unshare(CLONE_NEWNET)' calls > make the number of active slab objects including 'sock_inode_cache' type > rapidly and continuously increase. As a result, memory pressure o

[PATCH v4] net/ipv4/inet_fragment: Batch fqdir destroy works

2020-12-11 Thread SeongJae Park
From: SeongJae Park On a few of our systems, I found frequent 'unshare(CLONE_NEWNET)' calls make the number of active slab objects including 'sock_inode_cache' type rapidly and continuously increase. As a result, memory pressure occurs. In more detail, I made an artificial reproducer that resem