[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <4557d3ad541b4272bc1286480af5e562@baidu.com>
Date: Thu, 27 Aug 2020 08:53:09 +0000
From: "Li,Rongqing" <lirongqing@...du.com>
To: Eric Dumazet <eric.dumazet@...il.com>,
"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
"intel-wired-lan@...ts.osuosl.org" <intel-wired-lan@...ts.osuosl.org>
Subject: RE: [PATCH] iavf: use kvzalloc instead of kzalloc for rx/tx_bi buffer
> -----Original Message-----
> From: Eric Dumazet [mailto:eric.dumazet@...il.com]
> Sent: Thursday, August 27, 2020 4:26 PM
> To: Li,Rongqing <lirongqing@...du.com>; netdev@...r.kernel.org;
> intel-wired-lan@...ts.osuosl.org
> Subject: Re: [PATCH] iavf: use kvzalloc instead of kzalloc for rx/tx_bi buffer
>
>
>
> On 8/27/20 12:53 AM, Li RongQing wrote:
> > when changes the rx/tx ring to 4096, kzalloc may fail due to a
> > temporary shortage on slab entries.
> >
> > kvmalloc is used to allocate this memory as there is no need to have
> > this memory area physical continuously.
> >
> > Signed-off-by: Li RongQing <lirongqing@...du.com>
> > ---
>
>
> Well, fallback to vmalloc() overhead because order-1 pages are not readily
> available when the NIC is setup (usually one time per boot) is adding TLB cost
> at run time, for billions of packets to come, maybe for months.
>
> Surely trying a bit harder to get order-1 pages is desirable.
>
> __GFP_RETRY_MAYFAIL is supposed to help here.
Could we add __GFP_RETRY_MAYFAIL to kvmalloc, to ensure the allocation success ?
>
I see that lots of drivers are using vmalloc for this buffer, should we change it kmalloc?
grep "buffer_info =" drivers/net/ethernet/intel/ -rI|grep alloc
drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c: tx_ring->tx_buffer_info = vmalloc(size);
drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c: rx_ring->rx_buffer_info = vmalloc(size);
drivers/net/ethernet/intel/ixgbe/ixgbe_main.c: tx_ring->tx_buffer_info = vmalloc_node(size, ring_node);
drivers/net/ethernet/intel/ixgbe/ixgbe_main.c: tx_ring->tx_buffer_info = vmalloc(size);
drivers/net/ethernet/intel/ixgbe/ixgbe_main.c: rx_ring->rx_buffer_info = vmalloc_node(size, ring_node);
drivers/net/ethernet/intel/ixgbe/ixgbe_main.c: rx_ring->rx_buffer_info = vmalloc(size);
drivers/net/ethernet/intel/ixgb/ixgb_main.c: txdr->buffer_info = vzalloc(size);
drivers/net/ethernet/intel/ixgb/ixgb_main.c: rxdr->buffer_info = vzalloc(size);
drivers/net/ethernet/intel/e1000e/ethtool.c: tx_ring->buffer_info = kcalloc(tx_ring->count,
drivers/net/ethernet/intel/e1000e/ethtool.c: rx_ring->buffer_info = kcalloc(rx_ring->count,
drivers/net/ethernet/intel/e1000e/netdev.c: tx_ring->buffer_info = vzalloc(size);
drivers/net/ethernet/intel/e1000e/netdev.c: rx_ring->buffer_info = vzalloc(size);
drivers/net/ethernet/intel/igb/igb_main.c: tx_ring->tx_buffer_info = vmalloc(size);
drivers/net/ethernet/intel/igb/igb_main.c: rx_ring->rx_buffer_info = vmalloc(size);
drivers/net/ethernet/intel/e1000/e1000_ethtool.c: txdr->buffer_info = kcalloc(txdr->count, sizeof(struct e1000_tx_buffer),
drivers/net/ethernet/intel/e1000/e1000_ethtool.c: rxdr->buffer_info = kcalloc(rxdr->count, sizeof(struct e1000_rx_buffer),
drivers/net/ethernet/intel/e1000/e1000_main.c: txdr->buffer_info = vzalloc(size);
drivers/net/ethernet/intel/e1000/e1000_main.c: rxdr->buffer_info = vzalloc(size);
drivers/net/ethernet/intel/igc/igc_main.c: tx_ring->tx_buffer_info = vzalloc(size);
drivers/net/ethernet/intel/igc/igc_main.c: rx_ring->rx_buffer_info = vzalloc(size);
drivers/net/ethernet/intel/igbvf/netdev.c: tx_ring->buffer_info = vzalloc(size);
drivers/net/ethernet/intel/igbvf/netdev.c: rx_ring->buffer_info = vzalloc(size);
-Li
Powered by blists - more mailing lists