[PATCH 0/3] cxgb3 driver update

[PATCH 0/3] cxgb3 driver update

[Divy Le Ray]

Hi Jeff,

I'm submitting three more patches for inclusion in netdev#upstream.
These patches are built over the series I resent yesterday night.
The patch numbering reflects the stacking.

Here is a brief description:
-   avoid false positives in the xgmac hang workaround
-   Properly set the CQ_ERR bit in RDMA CQ contexts.
-   Update CQ context operations time out values

Cheers,
Divy

Re: [PATCH 0/3] cxgb3 driver update

[Al Viro]

On Wed, Aug 22, 2007 at 11:35:20PM -0700, Divy Le Ray wrote:
> Hi Jeff,
> 
> I'm submitting three more patches for inclusion in netdev#upstream.
> These patches are built over the series I resent yesterday night.
> The patch numbering reflects the stacking.
> 
> Here is a brief description:
> -   avoid false positives in the xgmac hang workaround
> -   Properly set the CQ_ERR bit in RDMA CQ contexts.
> -   Update CQ context operations time out values

Speaking of cxgb3, could you explain what the hell is
static int do_term(struct t3cdev *dev, struct sk_buff *skb)
{
        unsigned int hwtid = ntohl(skb->priority) >> 8 & 0xfffff;
doing?  AFAIK, skb->priority is not net-endian...

Another odd place is
int t3_seeprom_write(struct adapter *adapter, u32 addr, u32 data)
{   
        u16 val;
        int attempts = EEPROM_MAX_POLL;
        unsigned int base = adapter->params.pci.vpd_cap_addr;

        if ((addr >= EEPROMSIZE && addr != EEPROM_STAT_ADDR) || (addr & 3))   
                return -EINVAL;

        pci_write_config_dword(adapter->pdev, base + PCI_VPD_DATA,
                               cpu_to_le32(data));
with callers like
int t3_seeprom_wp(struct adapter *adapter, int enable)
{
        return t3_seeprom_write(adapter, EEPROM_STAT_ADDR, enable ? 0xc : 0);

IOW, you really get little-endian values passed to pci_write_config_dword()
and it expects a host-endian as the last argument...

Re: [PATCH 0/3] cxgb3 driver update

[Divy Le Ray]

Hi Al,

> Speaking of cxgb3, could you explain what the hell is
> static int do_term(struct t3cdev *dev, struct sk_buff *skb)
> {
>         unsigned int hwtid = ntohl(skb->priority) >> 8 & 0xfffff;
> doing?  AFAIK, skb->priority is not net-endian...
>

the RDMA connection id is saved in the skb's priority field for TERM 
messages
because it is not in the CPL message that comes up from the hardware.
Yet the RDMA driver needs it, so sge.c::process_responses() overloads
the skb's priority and csum with these values.


>
> Another odd place is
> int t3_seeprom_write(struct adapter *adapter, u32 addr, u32 data)
> {  
>         u16 val;
>         int attempts = EEPROM_MAX_POLL;
>         unsigned int base = adapter->params.pci.vpd_cap_addr;
>
>         if ((addr >= EEPROMSIZE && addr != EEPROM_STAT_ADDR) || (addr 
> & 3))  
>                 return -EINVAL;
>
>         pci_write_config_dword(adapter->pdev, base + PCI_VPD_DATA,
>                                cpu_to_le32(data));
> with callers like
> int t3_seeprom_wp(struct adapter *adapter, int enable)
> {
>         return t3_seeprom_write(adapter, EEPROM_STAT_ADDR, enable ? 
> 0xc : 0);
>
> IOW, you really get little-endian values passed to 
> pci_write_config_dword()
> and it expects a host-endian as the last argument...
>

It looks like a bug. Thanks for spotting this.

Cheers,
Divy