| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <OS0PR01MB5922165EFE14E02388B34F4086B29@OS0PR01MB5922.jpnprd01.prod.outlook.com>
Date: Fri, 8 Oct 2021 06:46:32 +0000
From: Biju Das <biju.das.jz@...renesas.com>
To: Biju Das <biju.das.jz@...renesas.com>,
Sergey Shtylyov <s.shtylyov@....ru>,
"David S. Miller" <davem@...emloft.net>,
Jakub Kicinski <kuba@...nel.org>
CC: Sergei Shtylyov <sergei.shtylyov@...il.com>,
Geert Uytterhoeven <geert+renesas@...der.be>,
Sergey Shtylyov <s.shtylyov@...russia.ru>,
Adam Ford <aford173@...il.com>, Andrew Lunn <andrew@...n.ch>,
Yuusuke Ashizuka <ashiduka@...itsu.com>,
Yoshihiro Shimoda <yoshihiro.shimoda.uh@...esas.com>,
"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
"linux-renesas-soc@...r.kernel.org"
<linux-renesas-soc@...r.kernel.org>,
Chris Paterson <Chris.Paterson2@...esas.com>,
Biju Das <biju.das@...renesas.com>,
Prabhakar Mahadev Lad <prabhakar.mahadev-lad.rj@...renesas.com>
Subject: RE: [RFC 07/12] ravb: Fillup ravb_rx_gbeth() stub
Hi Sergey,
> Subject: RE: [RFC 07/12] ravb: Fillup ravb_rx_gbeth() stub
>
> Hi Sergey,
>
>
> Thanks for the feedback.
>
> > Subject: Re: [RFC 07/12] ravb: Fillup ravb_rx_gbeth() stub
> >
> > On 10/7/21 8:49 AM, Biju Das wrote:
> >
> > [...]
> > >>>>> Fillup ravb_rx_gbeth() function to support RZ/G2L.
> > >>>>>
> > >>>>> This patch also renames ravb_rcar_rx to ravb_rx_rcar to be
> > >>>>> consistent with the naming convention used in sh_eth driver.
> > >>>>>
> > >>>>> Signed-off-by: Biju Das <biju.das.jz@...renesas.com>
> > >>>>> Reviewed-by: Lad Prabhakar
> > >>>>> <prabhakar.mahadev-lad.rj@...renesas.com>[...]
> > >>>>> diff --git a/drivers/net/ethernet/renesas/ravb_main.c
> > >>>>> b/drivers/net/ethernet/renesas/ravb_main.c
> > >>>>> index 37164a983156..42573eac82b9 100644
> > >>>>> --- a/drivers/net/ethernet/renesas/ravb_main.c
> > >>>>> +++ b/drivers/net/ethernet/renesas/ravb_main.c
> > >>>>> @@ -720,6 +720,23 @@ static void ravb_get_tx_tstamp(struct
> > >>>>> net_device
> > >>>> *ndev)
> > >>>>> }
> > >>>>> }
> > >>>>>
> > >>>>> +static void ravb_rx_csum_gbeth(struct sk_buff *skb) {
> > >>>>> + u8 *hw_csum;
> > >>>>> +
> > >>>>> + /* The hardware checksum is contained in sizeof(__sum16) (2)
> > bytes
> > >>>>> + * appended to packet data
> > >>>>> + */
> > >>>>> + if (unlikely(skb->len < sizeof(__sum16)))
> > >>>>> + return;
> > >>>>> + hw_csum = skb_tail_pointer(skb) - sizeof(__sum16);
> > >>>>
> > >>>> Not 32-bit? The manual says the IP checksum is stored in the
> > >>>> first
> > >>>> 2 bytes.
> > >>>
> > >>> It is 16 bit. It is on last 2 bytes.
> >
> > The IP checksum is at the 1st 2 bytes of the overall 4-byte
> > checksum (coming after the packet payload), no?
>
> Sorry, I got confused with your question earlier. Now it is clear for me.
>
> I agree the checksum part is stored in last 4bytes. Of this, the first 2
> bytes IPV4 checksum and last 2 bytes TCP/UDP/ICMP checksum.
>
> >
> > >> So you're saying the manual is wrong?
> > >
> > > I am not sure which manual you are referring here.
> > >
> > > I am referring to Rev.1.00 Sep, 2021 of RZ/G2L hardware manual and
> >
> > Same here.
> >
> > [...]
> >
> > > Please check the section 30.5.6.1 checksum calculation handling> And
> > > figure 30.25 the field of checksum attaching field
> >
> > I have.
> >
> > > Also see Table 30.17 for checksum values for non-error conditions.
> >
> > > TCP/UDP/ICPM checksum is at last 2bytes.
> >
> > What are you arguing with then? :-)
> > My point was that your code fetched the TCP/UDP/ICMP checksum ISO
> > the IP checksum because it subtracts sizeof(__sum16), while should
> > probably subtract sizeof(__wsum)
>
> Agreed. My code missed IP4 checksum result. May be we need to extract 2
> checksum info from last 4 bytes. First checksum(2bytes) is IP4 header
> checksum and next checksum(2 bytes) for TCP/UDP/ICMP and use this info
> finding the non error case mentioned in Table 30.17.
>
> For eg:-
> IPV6 non error-condition --> "0xFFFF"-->IPV4HeaderCSum value and "0x0000"
> TCP/UDP/ICMP CSUM value
>
> IPV4 non error-condition --> "0x0000"-->IPV4HeaderCSum value and "0x0000"
> TCP/UDP/ICMP CSUM value
>
> Do you agree?
What I meant here is some thing like below, please let me know if you have any issues with
this, otherwise I would like to send the patch with below changes.
Further improvements can happen later.
Please let me know.
+/* Hardware checksum status */
+#define IPV4_RX_CSUM_OK 0x00000000
+#define IPV6_RX_CSUM_OK 0xFFFF0000
+
enum ravb_reg {
/* AVB-DMAC registers */
CCC = 0x0000,
diff --git a/drivers/net/ethernet/renesas/ravb_main.c b/drivers/net/ethernet/renesas/ravb_main.c
index bbb42e5328e4..d9201fbbd472 100644
--- a/drivers/net/ethernet/renesas/ravb_main.c
+++ b/drivers/net/ethernet/renesas/ravb_main.c
@@ -722,16 +722,18 @@ static void ravb_get_tx_tstamp(struct net_device *ndev)
static void ravb_rx_csum_gbeth(struct sk_buff *skb)
{
- u16 *hw_csum;
+ u32 csum_result;
+ u8 *hw_csum;
/* The hardware checksum is contained in sizeof(__sum16) (2) bytes
* appended to packet data
*/
- if (unlikely(skb->len < sizeof(__sum16)))
+ if (unlikely(skb->len < sizeof(__wsum)))
return;
- hw_csum = (u16*)(skb_tail_pointer(skb) - sizeof(__sum16));
+ hw_csum = skb_tail_pointer(skb) - sizeof(__wsum);
+ csum_result = get_unaligned_le32(hw_csum);
- if (*hw_csum == 0)
+ if (csum_result == IPV4_RX_CSUM_OK || csum_result == IPV6_RX_CSUM_OK)
Regards,
Biju
>
> Regards,
> Biju
> >
> > >>>>> +
> > >>>>> + if (*hw_csum == 0)
> > >>>>
> > >>>> You only check the 1st byte, not the full checksum!
> > >>>
> > >>> As I said earlier, "0" value on last 16 bit, means no checksum
> error.
> > >>
> > >> How's that? 'hw_csum' is declared as 'u8 *'!
> > >
> > > It is my mistake, which will be taken care in the next patch by
> > > using
> > u16 *.
> >
> > Note that this 'u16' halfword can be unaligned, that's why the
> > current code uses get_unaligned_le16().
> >
> > >>>>> + skb->ip_summed = CHECKSUM_UNNECESSARY;
> > >>>>> + else
> > >>>>> + skb->ip_summed = CHECKSUM_NONE;
> > >>>>
> > >>>> So the TCP/UDP/ICMP checksums are not dealt with? Why enable
> > >>>> them
> > >> then?
> > >>>
> > >>> If last 2bytes is zero, means there is no checksum error w.r.to
> > >> TCP/UDP/ICMP checksums.
> > >>
> > >> Why checksum them independently then?
> > >
> > > It is a hardware feature.
> >
> > Switchable, isn't it?
>
> >
> > > Regards,
> > > Biju
> >
> > [...]
> >
> > MBR, Sergey
Powered by blists - more mailing lists