From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Michael S. Tsirkin" Subject: Re: [PATCH net-next] net: allow large number of tx queues Date: Thu, 20 Jun 2013 11:35:30 +0300 Message-ID: <20130620083530.GA23909@redhat.com> References: <1371620452-49349-1-git-send-email-jasowang@redhat.com> <1371620452-49349-2-git-send-email-jasowang@redhat.com> <1371623518.3252.267.camel@edumazet-glaptop> <20130619091132.GA2816@redhat.com> <1371635763.3252.289.camel@edumazet-glaptop> <20130619154059.GA13735@redhat.com> <1371657511.3252.324.camel@edumazet-glaptop> <20130619180709.GA15017@redhat.com> <1371716151.3252.390.camel@edumazet-glaptop> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Jason Wang , davem@davemloft.net, hkchu@google.com, netdev , dwmw2@infradead.org To: Eric Dumazet Return-path: Received: from mx1.redhat.com ([209.132.183.28]:25131 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754012Ab3FTIfN (ORCPT ); Thu, 20 Jun 2013 04:35:13 -0400 Content-Disposition: inline In-Reply-To: <1371716151.3252.390.camel@edumazet-glaptop> Sender: netdev-owner@vger.kernel.org List-ID: On Thu, Jun 20, 2013 at 01:15:51AM -0700, Eric Dumazet wrote: > From: Eric Dumazet > > On Wed, 2013-06-19 at 21:07 +0300, Michael S. Tsirkin wrote: > > > As we have explicit code to recover, maybe set __GFP_NOWARN? > > If we are out of memory, vzalloc will warn too, we don't > > need two warnings. > > Yes, that's absolutely the right thing to do. > > Its yet not clear this patch is really needed, but here it is. > > Thanks > > [PATCH net-next] net: allow large number of tx queues > > netif_alloc_netdev_queues() uses kcalloc() to allocate memory > for the "struct netdev_queue *_tx" array. > > For large number of tx queues, kcalloc() might fail, so this > patch does a fallback to vzalloc(). > > As vmalloc() adds overhead on a critical network path, add __GFP_REPEAT > to kzalloc() flags to do this fallback only when really needed. > > Signed-off-by: Eric Dumazet FWIW Acked-by: Michael S. Tsirkin This makes it possible to go up to 1K queues which should be enough, so we can revert the first chunk in edfb6a148ce62e5e19354a1dcd9a34e00815c2a1. > --- > net/core/dev.c | 26 +++++++++++++++++++------- > 1 file changed, 19 insertions(+), 7 deletions(-) > > diff --git a/net/core/dev.c b/net/core/dev.c > index fa007db..722f633 100644 > --- a/net/core/dev.c > +++ b/net/core/dev.c > @@ -130,6 +130,7 @@ > #include > #include > #include > +#include > > #include "net-sysfs.h" > > @@ -5253,17 +5254,28 @@ static void netdev_init_one_queue(struct net_device *dev, > #endif > } > > +static void netif_free_tx_queues(struct net_device *dev) > +{ > + if (is_vmalloc_addr(dev->_tx)) > + vfree(dev->_tx); > + else > + kfree(dev->_tx); > +} > + > static int netif_alloc_netdev_queues(struct net_device *dev) > { > unsigned int count = dev->num_tx_queues; > struct netdev_queue *tx; > + size_t sz = count * sizeof(*tx); > > - BUG_ON(count < 1); > - > - tx = kcalloc(count, sizeof(struct netdev_queue), GFP_KERNEL); > - if (!tx) > - return -ENOMEM; > + BUG_ON(count < 1 || count > 0xffff); > > + tx = kzalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT); > + if (!tx) { > + tx = vzalloc(sz); > + if (!tx) > + return -ENOMEM; > + } > dev->_tx = tx; > > netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL); > @@ -5811,7 +5823,7 @@ free_all: > > free_pcpu: > free_percpu(dev->pcpu_refcnt); > - kfree(dev->_tx); > + netif_free_tx_queues(dev); > #ifdef CONFIG_RPS > kfree(dev->_rx); > #endif > @@ -5836,7 +5848,7 @@ void free_netdev(struct net_device *dev) > > release_net(dev_net(dev)); > > - kfree(dev->_tx); > + netif_free_tx_queues(dev); > #ifdef CONFIG_RPS > kfree(dev->_rx); > #endif >