NetXen driver name

NetXen driver name

[Amit Kale]

Hi,

NetXen ethernet driver present in 2.6.20 has name "netxen_nic". The product 
itself is refered to as "nx_nic". To make it inline with the product, would 
it be a good idea to change the name of the driver to "nx_nic"?

Thanks.
-Amit

Re: NetXen driver name

[Andi Kleen]

Amit Kale <amitkale@netxen.com> writes:

> Hi,
> 
> NetXen ethernet driver present in 2.6.20 has name "netxen_nic". The product 
> itself is refered to as "nx_nic". To make it inline with the product, would 
> it be a good idea to change the name of the driver to "nx_nic"?

When it's already in a released kernel it's a little late to rename the module.
You would cause upgrade pain to users who have hardcoded the module name.

-Andi

Re: NetXen driver name

[Amit Kale]

The already released kernel contains a broken driver.  It broke due to some 
code rearrangement changes someone submitted to fix sparse warnings.

-Amit

On Thursday 08 February 2007 23:45, Andi Kleen wrote:
> Amit Kale <amitkale@netxen.com> writes:
> > Hi,
> >
> > NetXen ethernet driver present in 2.6.20 has name "netxen_nic". The
> > product itself is refered to as "nx_nic". To make it inline with the
> > product, would it be a good idea to change the name of the driver to
> > "nx_nic"?
>
> When it's already in a released kernel it's a little late to rename the
> module. You would cause upgrade pain to users who have hardcoded the module
> name.
>
> -Andi

Re: NetXen driver name

[Andi Kleen]

On Monday 12 February 2007 09:03, Amit Kale wrote:
> The already released kernel contains a broken driver.  It broke due to some 
> code rearrangement changes someone submitted to fix sparse warnings.


You mean it doesn't compile? 

-Andi

Re: NetXen driver name

[Amit Kale]

On Monday 12 February 2007 13:42, Andi Kleen wrote:
> On Monday 12 February 2007 09:03, Amit Kale wrote:
> > The already released kernel contains a broken driver.  It broke due to
> > some code rearrangement changes someone submitted to fix sparse warnings.
>
> You mean it doesn't compile?

It compiles, but doesn't work if tried on real hardware.
We'll send a fix for the same in a few hours.
-Amit

>
> -Andi

Re: NetXen driver name

[Al Viro]

On Mon, Feb 12, 2007 at 09:12:56AM +0100, Andi Kleen wrote:
> On Monday 12 February 2007 09:03, Amit Kale wrote:
> > The already released kernel contains a broken driver.  It broke due to some 
> > code rearrangement changes someone submitted to fix sparse warnings.
 
s/sparse warnings/breakage on big-endian boxen/, actually.  This stuff
was sent to netdev tree and sat there for a while.  Pity if some breakage
occured and went unnoticed; could you give more details?

> You mean it doesn't compile? 

Re: NetXen driver name

[Amit Kale]

Sending it again from correct smtp server.

On Monday 12 February 2007 13:42, Andi Kleen wrote:
> On Monday 12 February 2007 09:03, Amit Kale wrote:
> > The already released kernel contains a broken driver.  It broke due to
> > some code rearrangement changes someone submitted to fix sparse warnings.
>
> You mean it doesn't compile?

It compiles, but doesn't work if tried on real hardware.
We'll send a fix for the same in a few hours.
-Amit

>
> -Andi

Re: NetXen driver name

[Andi Kleen]

On Monday 12 February 2007 09:28, Amit Kale wrote:
> On Monday 12 February 2007 13:42, Andi Kleen wrote:
> > On Monday 12 February 2007 09:03, Amit Kale wrote:
> > > The already released kernel contains a broken driver.  It broke due to
> > > some code rearrangement changes someone submitted to fix sparse warnings.
> >
> > You mean it doesn't compile?
> 
> It compiles, but doesn't work if tried on real hardware.
> We'll send a fix for the same in a few hours.

But didn't the driver make it into distributions with the old name? 
That could be another compat issue (such things easily make unhappy
users and you probably wouldn't like that)

But ok then a rename is probably still possible. You have to ask Jeff of course.

-Andi

Re: NetXen driver name

[Amit Kale]

On Monday 12 February 2007 13:59, Al Viro wrote:
> On Mon, Feb 12, 2007 at 09:12:56AM +0100, Andi Kleen wrote:
> > On Monday 12 February 2007 09:03, Amit Kale wrote:
> > > The already released kernel contains a broken driver.  It broke due to
> > > some code rearrangement changes someone submitted to fix sparse
> > > warnings.
>
> s/sparse warnings/breakage on big-endian boxen/, actually.  This stuff
> was sent to netdev tree and sat there for a while.  Pity if some breakage
> occured and went unnoticed; could you give more details?

We were in middle of other changes last week. 

Here it is.  This is FYI only. We'll send a set of patches with this change as 
well as some other in a short time.

diff -Narup not_working/netxen/netxen_nic.h got_working/netxen/netxen_nic.h
--- not_working/netxen/netxen_nic.h	2007-02-08 23:20:34.468721000 -0800
+++ got_working/netxen/netxen_nic.h	2007-02-08 23:19:44.526769000 -0800
@@ -252,7 +252,7 @@ typedef u32 netxen_ctx_msg;
 #define netxen_set_msg_ctxid(config_word, val)	\
 	((config_word) &= ~(0x3ff<<18), (config_word) |= (val & 0x3ff) << 18)
 #define netxen_set_msg_opcode(config_word, val)	\
-	((config_word) &= ~(0xf<<24), (config_word) |= (val & 0xf) << 24)
+	((config_word) &= ~(0xf<<28), (config_word) |= (val & 0xf) << 28)
 
 struct netxen_rcv_context {
 	__le64 rcv_ring_addr;
@@ -309,8 +309,8 @@ struct netxen_ring_ctx {
 	((cmd_desc)->num_of_buffers_total_length &= ~cpu_to_le32(0xff), \
 	(cmd_desc)->num_of_buffers_total_length |= cpu_to_le32((val) & 0xff))
 #define netxen_set_cmd_desc_totallength(cmd_desc, val)	\
-	((cmd_desc)->num_of_buffers_total_length &= cpu_to_le32(0xff), \
-	(cmd_desc)->num_of_buffers_total_length |= cpu_to_le32(val << 24))
+	((cmd_desc)->num_of_buffers_total_length &= cpu_to_le32(0xffffff), \
+	(cmd_desc)->num_of_buffers_total_length |= cpu_to_le32(val << 8))
 
 #define netxen_get_cmd_desc_opcode(cmd_desc)	\
 	((le16_to_cpu((cmd_desc)->flags_opcode) >> 7) & 0x003F)