[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH] Revert "xen/arm: do not relocate Xen outside of visible RAM"



Makes sense. I will update the commit message. Yes, it is ok to push this out to Xen 4.9. Appreciate the feedback.

On 10/24/2016 5:14 PM, Stefano Stabellini wrote:
On Mon, 24 Oct 2016, Sameer Goel wrote:
This reverts commit db92b1ac55cd5e193ae22b0b6f01fb47bc9e5d2f.

There does not seem to be a restriction on non contiguous memory anymore . So,
reverting this change, so that Xen is placed at the end of the useable
system RAM even if the partitions are not contiguous.

Signed-off-by: Sameer Goel <sgoel@xxxxxxxxxxxxxx>
---
The load restriction placed in the above reverted patch resulted in a crash
when booting DOM0 on a Qualcomm platform with non contiguous system RAM.

Hi Sameer,

Thanks for the patch. It might be worth to point to
2d02b05c77fc5e7c76bf6f112db84bbaa44fdcb5 in the commit message, which I
think is the commit that made non-contiguous ram regions work properly.

This is not a fix for a currently supported platform, right? This is for
a new platform?

I am asking because if this is for a new platform, I would prefer to
consider the patch for the next development cycle, Xen 4.9. We have just
tagged 4.8-RC3, it would be best to avoid changes with potentially
harmful side effects at this stage (theoretically this change would need
to validated on every supported platforms, which at this stage is
unrealistic).



(XEN) DOM0: [ 0.000000] bootmem alloc of 64 bytes failed!
(XEN) DOM0: [ 0.000000] Kernel panic - not syncing: Out of memory
(XEN) DOM0: [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 4.7.0-rc7+ #19
(XEN) DOM0: [ 0.000000] Hardware name: (null) (DT)
(XEN) DOM0: [ 0.000000] Call trace:
(XEN) DOM0: [ 0.000000] [<ffff000008088854>] dump_backtrace+0x0/0x1e4
(XEN) DOM0: [ 0.000000] [<ffff000008088a5c>] show_stack+0x24/0x2c
(XEN) DOM0: [ 0.000000] [<ffff000008452004>] dump_stack+0x8c/0xb0
(XEN) DOM0: [ 0.000000] [<ffff00000819ee78>] panic+0x128/0x268
(XEN) DOM0: [ 0.000000] [<ffff00000902c018>] __alloc_bootmem_low+0x40/0x4c
(XEN) DOM0: [ 0.000000] [<ffff000009012adc>] setup_arch+0x2d8/0x560
(XEN) DOM0: [ 0.000000] [<ffff00000901084c>] start_kernel+0x60/0x3b4
(XEN) DOM0: [ 0.000000] [<ffff0000090101bc>] __primary_switched+0x30/0x74
(XEN) DOM0: [ 0.000000] ---[ end Kernel panic - not syncing: Out of memory

The root cause for the crash was >4GB difference between the arm grant table
(lower address) and the kernel start address. The kernel sees the grant table
region as the start of system RAM.

Since, the grant table is a reuse of the text region of Xen image this issue
would not be seen if Xen is loaded high enough in memory.

 xen/arch/arm/setup.c | 10 +---------
 1 file changed, 1 insertion(+), 9 deletions(-)

diff --git a/xen/arch/arm/setup.c b/xen/arch/arm/setup.c
index 38eb888..1678871 100644
--- a/xen/arch/arm/setup.c
+++ b/xen/arch/arm/setup.c
@@ -392,25 +392,17 @@ static paddr_t __init get_xen_paddr(void)
 {
     struct meminfo *mi = &bootinfo.mem;
     paddr_t min_size;
-    paddr_t paddr = 0, last_end;
+    paddr_t paddr = 0;
     int i;

     min_size = (_end - _start + (XEN_PADDR_ALIGN-1)) & ~(XEN_PADDR_ALIGN-1);

-    last_end = mi->bank[0].start;
-
     /* Find the highest bank with enough space. */
     for ( i = 0; i < mi->nr_banks; i++ )
     {
         const struct membank *bank = &mi->bank[i];
         paddr_t s, e;

-        /* We can only deal with contiguous memory at the moment */
-        if ( last_end != bank->start )
-            break;
-
-        last_end = bank->start + bank->size;
-
         if ( bank->size >= min_size )
         {
             e = consider_modules(bank->start, bank->start + bank->size,
--
Qualcomm Datacenter Technologies as an affiliate of Qualcomm Technologies, Inc. 
Qualcomm Technologies, Inc. is a member of the Code Aurora Forum, a Linux 
Foundation Collaborative Project.


--
Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.