Re: Compilation problem

[PUCCETTI Armand <armand.puccetti@cea.fr>]

Steven Hand a écrit :
>> When compiling XEN 3.0.3 (KERNELS = linux-2.6-xen0 linux-2.6-xenU) with=
>>
>> Linux kernel 2.6.16.29 or 2.6.26.35
>> and with gcc 4.1, I stumble over the following compilation error:
>>
>> make[3]: entrant dans le repertoire
>> .../xen-3.0.3_0-src/linux-2.6.16.35-xen0
>> CHK include/linux/version.h
>> CC [M] net/ipv4/netfilter/ip_conntrack_proto_tcp.o
>> In file included from net/ipv4/netfilter/ip_conntrack_proto_tcp.c:33:
>> include/net/tcp.h: In function tcp_skb_pcount:
>> include/net/tcp.h:555: error: struct skb_shared_info has no member
>> named tso_segs
>> include/net/tcp.h: In function tcp_skb_mss:
>> include/net/tcp.h:561: error: struct skb_shared_info has no member
>> named tso_size
>> make[6]: *** [net/ipv4/netfilter/ip_conntrack_proto_tcp.o] Erreur 1
>> make[5]: *** [net/ipv4/netfilter] Erreur 2
>> make[4]: *** [net/ipv4] Erreur 2
>> make[3]: *** [net] Erreur 2
>>     
>
> The build normally patches all of the references to tso_segs and tso_size
> with references to gso_segs and gso_size, see 
>
>  patches/linux-2.6.16.29/net-gso-0-base.patch 
>
> This will only happen for linux kernel versions that have appropriate 
> subdirectories in the patches directory. Since you're building a random
> other kernel, the patching will not occur and you'll get your error. 
>
> Are you sure you see this issue with 2.6.16.29? (which is the kernel
> shipped with 3.0.3-0)? 
>   
You're right, this does not happen with kernel 2.6.15.29 but 2.6.15.28. 
For  the former
the compilation generates vmlinuz.img.2.6.16.29-xen0 and -xenU files
but which I cannot start correctly (on this problem I sent a message 
that is in
the digest vol22 issue 29). That was why I tried to compile the code 
with another
(earlier or later) kernel.

For kernel 2.6.16.29 I could not (yet) catch the bootup messages, but 
Mats guessed
that it might be due to the absence of the SATA driver (the HDD is SATA 
in my config)
that I checked on my host machine.

On that one (P4HT with Ubuntu 6.10, kernel 2.6.17.10), the SATA drivers
are located in /lib/modules/2.6.17-10-generic/kernel/drivers/scsi and 
there are

3w-9xxx.ko      fdomain.ko      pdc_adma.ko      sata_uli.ko
3w-xxxx.ko      gdth.ko         ppa.ko           sata_via.ko
aacraid         ide-scsi.ko     qla1280.ko       sata_vsc.ko
ahci.ko         imm.ko          qlogicfas408.ko  scsi_debug.ko
aic7xxx         initio.ko       raid_class.ko    scsi_mod.ko
aic7xxx_old.ko  ipr.ko          sas              scsi_transport_fc.ko
aic94xx         ips.ko          sata_mv.ko       scsi_transport_iscsi.ko
arcmsr          iscsi_tcp.ko    sata_nv.ko       scsi_transport_sas.ko
ata_piix.ko     libata.ko       sata_promise.ko  scsi_transport_spi.ko
atp870u.ko      lpfc            sata_qstor.ko    sd_mod.ko
BusLogic.ko     megaraid        sata_sil24.ko    sg.ko
ch.ko           megaraid.ko     sata_sil.ko      sr_mod.ko
dc395x.ko       osst.ko         sata_sis.ko      st.ko
dmx3191d.ko     pata_pcmcia.ko  sata_svw.ko      sym53c8xx_2
eata.ko         pcmcia          sata_sx4.ko      tmscsim.ko


An "lsmod |grep ata" gives me:

ata_piix     13828 6
libata         88984 1     ata_piix
scsi_mod     181424 3 sg,sd_mod,libata

so they're apparently loaded.

However, in /lib/modules/2.6.16.29-xen0/kernel/drivers, there is no scsi 
directory and in
/lib/modules/2.6.16.29-xenU/kernel/drivers/scsi there are

osst.ko      scsi_transport_fc.ko   sd_mod.ko  sr_mod.ko
scsi_mod.ko  scsi_transport_spi.ko  sg.ko      st.ko

Do I miss something?
Might that be the solution, and how can I generate the drivers 
apparently missing (ata_piix.ko and libata)?

thanks
Armand
>   
>> It appears to be the same error with kernel 2.6.17.10.
>>     
>
> Indeed. 
>
>   
>> Any advice?
>>     
>  
> Use 2.6.16.29, or create new patches against the later kernels; you 
> can probably get some help from looking at the -unstable tree.
>
>
>
> cheers,
>
> S.
>
>
>