2018-09-11 19:56:09

by Boris Ostrovsky

Subject: [PATCH] x86/EISA: Don't probe EISA bus for Xen PV guests

For unprivileged Xen PV guests this is normal memory and ioremap will
not be able to properly map it.

While at it, since ioremap may return NULL, add a test for pointer's
validity.

Reported-by: Andy Smith <[email protected]>
Signed-off-by: Boris Ostrovsky <[email protected]>
---
arch/x86/kernel/eisa.c | 10 ++++++++--
1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kernel/eisa.c b/arch/x86/kernel/eisa.c
index f260e452e4f8..fcf3de1712c8 100644
--- a/arch/x86/kernel/eisa.c
+++ b/arch/x86/kernel/eisa.c
@@ -7,11 +7,17 @@
#include <linux/eisa.h>
#include <linux/io.h>

+#include <xen/xen.h>
+
static __init int eisa_bus_probe(void)
{
- void __iomem *p = ioremap(0x0FFFD9, 4);
+ void __iomem *p;
+
+ if (xen_pv_domain() && !xen_initial_domain())
+ return 0;

- if (readl(p) == 'E' + ('I'<<8) + ('S'<<16) + ('A'<<24))
+ p = ioremap(0x0FFFD9, 4);
+ if (p && readl(p) == 'E' + ('I'<<8) + ('S'<<16) + ('A'<<24))
EISA_bus = 1;
iounmap(p);
return 0;
--
2.17.1

2018-09-11 20:22:18

by Thomas Gleixner

Subject: Re: [PATCH] x86/EISA: Don't probe EISA bus for Xen PV guests

On Tue, 11 Sep 2018, Boris Ostrovsky wrote:

> For unprivileged Xen PV guests this is normal memory and ioremap will
> not be able to properly map it.
>
> While at it, since ioremap may return NULL, add a test for pointer's
> validity.

I assume this goes back to very dead kernels, so that should go with a Cc
stable, right?

Thanks,

tglx

2018-09-11 20:29:46

by Boris Ostrovsky

Subject: Re: [PATCH] x86/EISA: Don't probe EISA bus for Xen PV guests

On 9/11/18 4:20 PM, Thomas Gleixner wrote:
> On Tue, 11 Sep 2018, Boris Ostrovsky wrote:
>
>> For unprivileged Xen PV guests this is normal memory and ioremap will
>> not be able to properly map it.
>>
>> While at it, since ioremap may return NULL, add a test for pointer's
>> validity.
> I assume this goes back to very dead kernels, so that should go with a Cc
> stable, right?
>


Yes, I forgot to add those (and now I did), thanks for pointing this out.

It has to go at least all the way back to 4.13 since before
f7eaf6e00fd58 ("x86/boot: Move EISA setup to a separate file"), just by
luck, early_ioremap() worked for the guests.


-boris

2018-09-11 21:40:45

by tip-bot for Vasyl Gomonovych

Subject: [tip:x86/urgent] x86/EISA: Don't probe EISA bus for Xen PV guests

Commit-ID: 6a92b11169a65b3f8cc512c75a252cbd0d096ba0
Gitweb: https://git.kernel.org/tip/6a92b11169a65b3f8cc512c75a252cbd0d096ba0
Author: Boris Ostrovsky <[email protected]>
AuthorDate: Tue, 11 Sep 2018 15:55:38 -0400
Committer: Thomas Gleixner <[email protected]>
CommitDate: Tue, 11 Sep 2018 23:36:50 +0200

x86/EISA: Don't probe EISA bus for Xen PV guests

For unprivileged Xen PV guests this is normal memory and ioremap will
not be able to properly map it.

While at it, since ioremap may return NULL, add a test for pointer's
validity.

Reported-by: Andy Smith <[email protected]>
Signed-off-by: Boris Ostrovsky <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Link: https://lkml.kernel.org/r/[email protected]

---
arch/x86/kernel/eisa.c | 10 ++++++++--
1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kernel/eisa.c b/arch/x86/kernel/eisa.c
index f260e452e4f8..e8c8c5d78dbd 100644
--- a/arch/x86/kernel/eisa.c
+++ b/arch/x86/kernel/eisa.c
@@ -7,11 +7,17 @@
#include <linux/eisa.h>
#include <linux/io.h>

+#include <xen/xen.h>
+
static __init int eisa_bus_probe(void)
{
- void __iomem *p = ioremap(0x0FFFD9, 4);
+ void __iomem *p;
+
+ if (xen_pv_domain() && !xen_initial_domain())
+ return 0;

- if (readl(p) == 'E' + ('I'<<8) + ('S'<<16) + ('A'<<24))
+ p = ioremap(0x0FFFD9, 4);
+ if (p && readl(p) == 'E' + ('I' << 8) + ('S' << 16) + ('A' << 24))
EISA_bus = 1;
iounmap(p);
return 0;