[kernel-xen] Kernel Archive

Gordan Bobic gordan at bobich.net
Tue Jul 15 17:31:14 EST 2014


I have only used ZoL with 3.14 briefly, until I found the Nvidia drivers recent enough to build against it are broken.

I've been using ZoL in dom0 with zvols for VM backing for a couple of years now, but as mentioned, I'm still on kernel 3.9.9 due to Nvidia driver issue.

And yes, that FTBFS bug got me, too.


Christiaan den Besten <chris at prolocation.net> wrote:

>Hi !
>
>> Is there an archive of older kernels provided in the repository? There
>> are many reasons why forcing the latest and only the latest version on
>> people simply isn't useful.
>> 
>> For example, I have an Nvidia GPU on my dom0 and use ZFS. Until
>> recently, ZoL didn't support kernels as recent as 3.14, and Nvidia
>> drivers newer than 319.x are broken for Xen dom0 use. Running with a
>> 3.14 kernel requires a newer Nvidia driver than 319.x series.
>> 
>> Consequently I am still running 3.9.9, and now that new ZoL was released
>> that supports newer kernels I wanted to upgrade as far as Nvidia driver
>> would allow, but I can't because the only upgrade option is to 3.14,
>> which requires a newer Nvidia driver, which is broken.
>
>I am kinda curious how many people are using this kernel && ZoL. I just recently (last weekend) build a Dom0 host with this combination using zvol's as backing storage for the VMs.
>
>Noticed ZFS required a lot more CPU power on the DomU (and of course more memory for ARC).
>
>But when the drives get pushed to their max, the load on the DomU sky rockets with all 'zvol kernel threads' hitting D-state.
>
>Anyone else using this latest kernel + ZoL and has some experience to share?
>
>bye,
>Chris
>
>PS : For latest kernel to work I had to patch spl a bit by reverting one GIT update :
>
>[root at xen399.prolocation ~]# cat /etc/modprobe.d/zfs.conf
># 4Gb ( 4096 * 1024 * 1024 )
>options zfs zfs_arc_max=4294967296
>
>[root at xen399.prolocation ~]# diff -ruN /usr/src/spl-0.6.3/module/spl/spl-proc.c.org /usr/src/spl-0.6.3/module/spl/spl-proc.c
>--- /usr/src/spl-0.6.3/module/spl/spl-proc.c.org        2014-07-12 19:56:16.510863289 +0200
>+++ /usr/src/spl-0.6.3/module/spl/spl-proc.c    2014-07-12 19:58:26.665595197 +0200
>@@ -29,7 +29,6 @@
> #include <linux/kmod.h>
> #include <linux/seq_file.h>
> #include <linux/proc_compat.h>
>-#include <linux/version.h>
> #include <spl-debug.h>
>
> #ifdef SS_DEBUG_SUBSYS
>@@ -38,12 +37,6 @@
>
> #define SS_DEBUG_SUBSYS SS_PROC
>
>-#if defined(CONSTIFY_PLUGIN) && LINUX_VERSION_CODE >= KERNEL_VERSION(3,8,0)
>-typedef struct ctl_table __no_const spl_ctl_table;
>-#else
>-typedef struct ctl_table spl_ctl_table;
>-#endif
>-
> #ifdef DEBUG_KMEM
> static unsigned long table_min = 0;
> static unsigned long table_max = ~0;
>@@ -328,7 +321,7 @@
> SPL_PROC_HANDLER(proc_console_max_delay_cs)
> {
>         int rc, max_delay_cs;
>-        spl_ctl_table dummy = *table;
>+       struct ctl_table dummy = *table;
>         long d;
>        SENTRY;
>
>@@ -360,7 +353,7 @@
> SPL_PROC_HANDLER(proc_console_min_delay_cs)
> {
>         int rc, min_delay_cs;
>-        spl_ctl_table dummy = *table;
>+       struct ctl_table dummy = *table;
>         long d;
>        SENTRY;
>
>@@ -392,7 +385,7 @@
> SPL_PROC_HANDLER(proc_console_backoff)
> {
>         int rc, backoff;
>-        spl_ctl_table dummy = *table;
>+       struct ctl_table dummy = *table;
>        SENTRY;
>
>         dummy.data = &backoff;
>@@ -422,7 +415,7 @@
> {
>         int rc = 0;
>         unsigned long min = 0, max = ~0, val;
>-        spl_ctl_table dummy = *table;
>+       struct ctl_table dummy = *table;
>        SENTRY;
>
>         dummy.data = &val;
>@@ -449,7 +442,7 @@
> {
>         int rc = 0;
>         unsigned long min = 0, max = ~0, val = 0, mask;
>-        spl_ctl_table dummy = *table;
>+       struct ctl_table dummy = *table;
>         spl_kmem_cache_t *skc;
>         SENTRY;
>_______________________________________________
>kernel-xen mailing list
>kernel-xen at lists.wireless.org.au
>https://lists.wireless.org.au/mailman/listinfo/kernel-xen


More information about the kernel-xen mailing list