[ewg] Re: iSER issues for OFED 1.3

[Vladimir Sokolovsky]

Erez Zilber wrote:
> Vladimir Sokolovsky wrote:
>>
>> > -----Original Message-----
>> > From: Erez Zilber [mailto:erezz at Voltaire.COM]
>> > Sent: Monday, October 08, 2007 5:29 PM
>> > To: Vladimir Sokolovsky
>> > Subject: Re: iSER issues for OFED 1.3
>> >
>> >
>> > >> -----Original Message-----
>> > >> From: Erez Zilber [mailto:erezz at Voltaire.COM]
>> > >> Sent: Monday, October 08, 2007 1:19 PM
>> > >> To: Vladimir Sokolovsky
>> > >> Subject: iSER issues for OFED 1.3
>> > >>
>> > >> Vlad,
>> > >>
>> > >> The are some iSER related open issues in OFED 1.3 that we need to
>> > >> complete:
>> > >>
>> > >>     * I have some fixes that I need you to merge:
>> > >>       http://lists.openfabrics.org/pipermail/ewg/2007-
>> > >> October/004638.html
>> > >>
>> > >
>> > > Working on this now.
>> > >
>> >
>> > Thanks
>> >
>>
>> kernel_patches/backport/2.6.9_U2/libiscsi_no_flush_to_2_6_9.patch is
>> broken:
>>
>> Applying patch libiscsi_no_flush_to_2_6_9.patch
>> patching file drivers/scsi/libiscsi.c
>> Hunk #1 FAILED at 1225.
>> Hunk #2 succeeded at 1640 (offset 32 lines).
>> Hunk #3 FAILED at 1784.
>> 2 out of 3 hunks FAILED -- rejects in file drivers/scsi/libiscsi.c
>> Patch libiscsi_no_flush_to_2_6_9.patch does not apply (enforce with -f)
>>
>> Please test your patches with ofed_scripts/ofed_makedist.sh
>>
> 
> I didn't know that this patch ever existed. open-iscsi is not supported 
> on RH4 up2, so why should this happen? The installation script in OFED 
> 1.2 prevented that.
> 

In any case your last commits break this patch.
Should I remove it from the backports?
>>
>> > >
>> > >>     * I sent you an e-mail yesterday (for some reason, I don't see
>> > it
>> > >> on
>> > >>       the ewg list) about separating iscsi modules from the
>> > ofa_kernel
>> > >> rpm.
>> > >>
>> > >
>> > > I suggest to take ib-bonding as an example
>> > >
>> >
>> > I spoke with MoniS about that. He says that no IB module uses the
>> > bonding module symbols. This is different from what we have in iSER:
>> > iSER uses symbols from open-iscsi. If the open-iscsi rpm looks like
>> the
>> > bonding rpm, it won't work because iSER will be later get built (in
>> the
>> > ofa_kernel rpm) against the old open-iscsi symbols (from the distro).
>> > What do you suggest to do?
>> >
>>
>> You can create Module.symvers file for the new open-iscsi symbols and
>> then copy this file into iSER subdirectory before compiling iSER.
>>
> 
> This is really problematic because it means that we need to do the 
> following:
> 
>    * Compile open-iscsi and create the binary open-iscsi rpm.
Right.

>    * Open the ofa_kernel src rpm and copy Module.symvers into it.
>    * Create the ofa_kernel src rpm again and compile it.
> 
Why is that?
open-iscsi RPM can include its Module.symvers and install it.
Then ofa_kernel rpmbuild process should copy this file into iSER subdirectory and run make.

> Unless you have another idea, I think that we should consider not 
> separating iSCSI from the ofa_kernel rpm. It will be not as clean as we 
> wanted in the first place, but the code will be simpler and easier to 
> maintain.
> 
Why do you want to separate iSCSI from ofa_kernel?
I am OK with leaving iSCSI as is.

Regards,
Vladimir