| Message ID | 20231216-new-gemini-ethernet-regression-v2-2-64c269413dfa@linaro.org (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | Fix a regression in the Gemini ethernet controller. | expand |
On Sat, Dec 16, 2023 at 8:36 PM Linus Walleij <linus.walleij@linaro.org> wrote: > > We had workarounds were the ethernet checksumming engine would be bypassed > for larger frames, this fixed devices using DSA, but regressed devices > where the ethernet was connected directly to a PHY. > > The devices with a PHY connected directly can't handle large frames > either way, with or without bypass. Looking at the size of the frame > is probably just wrong. > > Rework the workaround such that we just bypass the checksumming engine if > the ethertype inside the actual frame is something else than 0x0800 > (IPv4) or 0x86dd (IPv6). These are the only frames the checksumming engine > can actually handle. VLAN framing (0x8100) also works fine. > > We can't inspect skb->protocol because DSA frames will sometimes have a > custom ethertype despite skb->protocol is e.g. 0x0800. > > After this both devices with direct ethernet attached such as D-Link > DNS-313 and devices with a DSA switch with a custom ethertype such as > D-Link DIR-685 work fine. > > Fixes: d4d0c5b4d279 ("net: ethernet: cortina: Handle large frames") > Signed-off-by: Linus Walleij <linus.walleij@linaro.org> > --- > drivers/net/ethernet/cortina/gemini.c | 21 ++++++++++++++++++++- > 1 file changed, 20 insertions(+), 1 deletion(-) > > diff --git a/drivers/net/ethernet/cortina/gemini.c b/drivers/net/ethernet/cortina/gemini.c > index 6a7ea051391a..1400f19bf05b 100644 > --- a/drivers/net/ethernet/cortina/gemini.c > +++ b/drivers/net/ethernet/cortina/gemini.c > @@ -1143,7 +1143,9 @@ static int gmac_map_tx_bufs(struct net_device *netdev, struct sk_buff *skb, > skb_frag_t *skb_frag; > dma_addr_t mapping; > unsigned short mtu; > + u16 ethertype; > void *buffer; > + __be16 *p; > > mtu = ETH_HLEN; > mtu += netdev->mtu; > @@ -1158,7 +1160,24 @@ static int gmac_map_tx_bufs(struct net_device *netdev, struct sk_buff *skb, > word3 |= mtu; > } > > - if (skb->ip_summed == CHECKSUM_PARTIAL) { > + /* Dig out the the ethertype actually in the buffer and not what the > + * protocol claims to be. This is the raw data that the checksumming > + * offload engine will have to deal with. > + */ > + p = (__be16 *)(skb->data + 2 * ETH_ALEN); > + ethertype = ntohs(*p); > + if (ethertype == ETH_P_8021Q) { > + p += 2; /* +2 sizeof(__be16) */ > + ethertype = ntohs(*p); > + } Presumably all you need is to call vlan_get_protocol() ? > + > + if (ethertype != ETH_P_IP && ethertype != ETH_P_IPV6) { > + /* Hardware offloaded checksumming isn't working on non-IP frames. > + * This happens for example on some DSA switches using a custom > + * ethertype. Just bypass the engine for those. > + */ > + word1 |= TSS_BYPASS_BIT; > + } else if (skb->ip_summed == CHECKSUM_PARTIAL) { > int tcp = 0; > > /* We do not switch off the checksumming on non TCP/UDP > > -- > 2.34.1 >
On Mon, Dec 18, 2023 at 3:50 PM Eric Dumazet <edumazet@google.com> wrote: > On Sat, Dec 16, 2023 at 8:36 PM Linus Walleij <linus.walleij@linaro.org> wrote: > > + /* Dig out the the ethertype actually in the buffer and not what the > > + * protocol claims to be. This is the raw data that the checksumming > > + * offload engine will have to deal with. > > + */ > > + p = (__be16 *)(skb->data + 2 * ETH_ALEN); > > + ethertype = ntohs(*p); > > + if (ethertype == ETH_P_8021Q) { > > + p += 2; /* +2 sizeof(__be16) */ > > + ethertype = ntohs(*p); > > + } > > Presumably all you need is to call vlan_get_protocol() ? Sadly no. As the comment says: we want the ethertype that is actually in the skb, not what is in skb->protocol, and the code in vlan_get_protocol() just trusts skb->protocol to be the ethertype in the frame, especially if vlan is not used. This is often what we want: DSA switches will "wash" custom ethertypes before they go out, but in this case the custom ethertype upsets the ethernet checksum engine used as conduit interface toward the DSA switch. Yours, Linus Walleij
On Tue, Dec 19, 2023 at 12:42 AM Linus Walleij <linus.walleij@linaro.org> wrote: > > On Mon, Dec 18, 2023 at 3:50 PM Eric Dumazet <edumazet@google.com> wrote: > > On Sat, Dec 16, 2023 at 8:36 PM Linus Walleij <linus.walleij@linaro.org> wrote: > > > > + /* Dig out the the ethertype actually in the buffer and not what the > > > + * protocol claims to be. This is the raw data that the checksumming > > > + * offload engine will have to deal with. > > > + */ > > > + p = (__be16 *)(skb->data + 2 * ETH_ALEN); > > > + ethertype = ntohs(*p); > > > + if (ethertype == ETH_P_8021Q) { > > > + p += 2; /* +2 sizeof(__be16) */ > > > + ethertype = ntohs(*p); > > > + } > > > > Presumably all you need is to call vlan_get_protocol() ? > > Sadly no. As the comment says: we want the ethertype that is actually in the > skb, not what is in skb->protocol, and the code in vlan_get_protocol() just > trusts skb->protocol to be the ethertype in the frame, especially if vlan > is not used. > > This is often what we want: DSA switches will "wash" custom ethertypes > before they go out, but in this case the custom ethertype upsets the > ethernet checksum engine used as conduit interface toward the DSA > switch. Problem is that your code misses skb_header_pointer() or pskb_may_pull() call... Second "ethertype = ntohs(*p);" might access uninitialized data. If this is a common operation, perhaps use a common helper from all drivers, this would help code review a bit...
On Tue, Dec 19, 2023 at 10:15 AM Eric Dumazet <edumazet@google.com> wrote: > On Tue, Dec 19, 2023 at 12:42 AM Linus Walleij <linus.walleij@linaro.org> wrote: > > This is often what we want: DSA switches will "wash" custom ethertypes > > before they go out, but in this case the custom ethertype upsets the > > ethernet checksum engine used as conduit interface toward the DSA > > switch. > > Problem is that your code misses skb_header_pointer() or > pskb_may_pull() call... > Second "ethertype = ntohs(*p);" might access uninitialized data. Yeah, needs to be done properly and look at skb->len etc. > If this is a common operation, perhaps use a common helper from all drivers, > this would help code review a bit... You are right, Maxime opened a discussion on it in a parallel, I'll cook something up! Yours, Linus Walleij
diff --git a/drivers/net/ethernet/cortina/gemini.c b/drivers/net/ethernet/cortina/gemini.c index 6a7ea051391a..1400f19bf05b 100644 --- a/drivers/net/ethernet/cortina/gemini.c +++ b/drivers/net/ethernet/cortina/gemini.c @@ -1143,7 +1143,9 @@ static int gmac_map_tx_bufs(struct net_device *netdev, struct sk_buff *skb, skb_frag_t *skb_frag; dma_addr_t mapping; unsigned short mtu; + u16 ethertype; void *buffer; + __be16 *p; mtu = ETH_HLEN; mtu += netdev->mtu; @@ -1158,7 +1160,24 @@ static int gmac_map_tx_bufs(struct net_device *netdev, struct sk_buff *skb, word3 |= mtu; } - if (skb->ip_summed == CHECKSUM_PARTIAL) { + /* Dig out the the ethertype actually in the buffer and not what the + * protocol claims to be. This is the raw data that the checksumming + * offload engine will have to deal with. + */ + p = (__be16 *)(skb->data + 2 * ETH_ALEN); + ethertype = ntohs(*p); + if (ethertype == ETH_P_8021Q) { + p += 2; /* +2 sizeof(__be16) */ + ethertype = ntohs(*p); + } + + if (ethertype != ETH_P_IP && ethertype != ETH_P_IPV6) { + /* Hardware offloaded checksumming isn't working on non-IP frames. + * This happens for example on some DSA switches using a custom + * ethertype. Just bypass the engine for those. + */ + word1 |= TSS_BYPASS_BIT; + } else if (skb->ip_summed == CHECKSUM_PARTIAL) { int tcp = 0; /* We do not switch off the checksumming on non TCP/UDP
We had workarounds were the ethernet checksumming engine would be bypassed for larger frames, this fixed devices using DSA, but regressed devices where the ethernet was connected directly to a PHY. The devices with a PHY connected directly can't handle large frames either way, with or without bypass. Looking at the size of the frame is probably just wrong. Rework the workaround such that we just bypass the checksumming engine if the ethertype inside the actual frame is something else than 0x0800 (IPv4) or 0x86dd (IPv6). These are the only frames the checksumming engine can actually handle. VLAN framing (0x8100) also works fine. We can't inspect skb->protocol because DSA frames will sometimes have a custom ethertype despite skb->protocol is e.g. 0x0800. After this both devices with direct ethernet attached such as D-Link DNS-313 and devices with a DSA switch with a custom ethertype such as D-Link DIR-685 work fine. Fixes: d4d0c5b4d279 ("net: ethernet: cortina: Handle large frames") Signed-off-by: Linus Walleij <linus.walleij@linaro.org> --- drivers/net/ethernet/cortina/gemini.c | 21 ++++++++++++++++++++- 1 file changed, 20 insertions(+), 1 deletion(-)