[PATCH v2 0/3] hvc_xen: fix xenboot on x86 and export to ARM

[PATCH v2 0/3] hvc_xen: fix xenboot on x86 and export to ARM

[Stefano Stabellini]

Hi all,

this series exports xenboot on ARM and ARM64 as earlycon, and gets
xenboot fully working again for PV DomUs on x86 (currently the xenboot
output only goes to the hypervisor via hypercall and not to the DomU
console).

Cheers,

Stefano


Stefano Stabellini (3):
      hvc_xen: add earlycon support
      hvc_xen: fix xenboot for DomUs
      hvc_xen: make early_printk work with HVM guests

 drivers/tty/hvc/hvc_xen.c |   75 +++++++++++++++++++++++++++++++++++----------
 1 file changed, 58 insertions(+), 17 deletions(-)

[PATCH v2 0/3] hvc_xen: fix xenboot on x86 and export to ARM

[Stefano Stabellini]

Hi all,

this series exports xenboot on ARM and ARM64 as earlycon, and gets
xenboot fully working again for PV DomUs on x86 (currently the xenboot
output only goes to the hypervisor via hypercall and not to the DomU
console).

Cheers,

Stefano


Stefano Stabellini (3):
      hvc_xen: add earlycon support
      hvc_xen: fix xenboot for DomUs
      hvc_xen: make early_printk work with HVM guests

 drivers/tty/hvc/hvc_xen.c |   75 +++++++++++++++++++++++++++++++++++----------
 1 file changed, 58 insertions(+), 17 deletions(-)

[PATCH v2 1/3] hvc_xen: add earlycon support

[Stefano Stabellini]

Introduce EARLYCON support in hvc_xen, useful for early debugging on arm
and arm64, where xen early_printk is not available.

It is different from xenboot_write_console on x86 in two ways:

- it does not return if !xen_pv_domain(), not only because ARM guests
  are xen_hvm_domain(), but also because we want to capture all the
  early boot messages, before xen support is discovered
- it does not try to print to the domU console at all, because xen
  support will only be discovered at a later point

Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>

---

Changes in v2:
- do not try to print to the DomU console
---
 drivers/tty/hvc/hvc_xen.c |   16 ++++++++++++++++
 1 file changed, 16 insertions(+)

diff --git a/drivers/tty/hvc/hvc_xen.c b/drivers/tty/hvc/hvc_xen.c
index fa816b7..68b8ec8 100644
--- a/drivers/tty/hvc/hvc_xen.c
+++ b/drivers/tty/hvc/hvc_xen.c
@@ -25,6 +25,7 @@
 #include <linux/init.h>
 #include <linux/types.h>
 #include <linux/list.h>
+#include <linux/serial_core.h>
 
 #include <asm/io.h>
 #include <asm/xen/hypervisor.h>
@@ -664,3 +665,18 @@ void xen_raw_printk(const char *fmt, ...)
 
 	xen_raw_console_write(buf);
 }
+
+static void xenboot_earlycon_write(struct console *console,
+				  const char *string,
+				  unsigned len)
+{
+	dom0_write_console(0, string, len);
+}
+
+static int __init xenboot_earlycon_setup(struct earlycon_device *device,
+					    const char *opt)
+{
+	device->con->write = xenboot_earlycon_write;
+	return 0;
+}
+EARLYCON_DECLARE(xenboot, xenboot_earlycon_setup);
-- 
1.7.10.4

[PATCH v2 1/3] hvc_xen: add earlycon support

[Stefano Stabellini]

Introduce EARLYCON support in hvc_xen, useful for early debugging on arm
and arm64, where xen early_printk is not available.

It is different from xenboot_write_console on x86 in two ways:

- it does not return if !xen_pv_domain(), not only because ARM guests
  are xen_hvm_domain(), but also because we want to capture all the
  early boot messages, before xen support is discovered
- it does not try to print to the domU console at all, because xen
  support will only be discovered at a later point

Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>

---

Changes in v2:
- do not try to print to the DomU console
---
 drivers/tty/hvc/hvc_xen.c |   16 ++++++++++++++++
 1 file changed, 16 insertions(+)

diff --git a/drivers/tty/hvc/hvc_xen.c b/drivers/tty/hvc/hvc_xen.c
index fa816b7..68b8ec8 100644
--- a/drivers/tty/hvc/hvc_xen.c
+++ b/drivers/tty/hvc/hvc_xen.c
@@ -25,6 +25,7 @@
 #include <linux/init.h>
 #include <linux/types.h>
 #include <linux/list.h>
+#include <linux/serial_core.h>
 
 #include <asm/io.h>
 #include <asm/xen/hypervisor.h>
@@ -664,3 +665,18 @@ void xen_raw_printk(const char *fmt, ...)
 
 	xen_raw_console_write(buf);
 }
+
+static void xenboot_earlycon_write(struct console *console,
+				  const char *string,
+				  unsigned len)
+{
+	dom0_write_console(0, string, len);
+}
+
+static int __init xenboot_earlycon_setup(struct earlycon_device *device,
+					    const char *opt)
+{
+	device->con->write = xenboot_earlycon_write;
+	return 0;
+}
+EARLYCON_DECLARE(xenboot, xenboot_earlycon_setup);
-- 
1.7.10.4

Re: [PATCH v2 1/3] hvc_xen: add earlycon support

[Boris Ostrovsky]

On 02/25/2016 07:10 AM, Stefano Stabellini wrote:
> Introduce EARLYCON support in hvc_xen, useful for early debugging on arm
> and arm64, where xen early_printk is not available.
>
> It is different from xenboot_write_console on x86 in two ways:
>
> - it does not return if !xen_pv_domain(), not only because ARM guests
>    are xen_hvm_domain(), but also because we want to capture all the
>    early boot messages, before xen support is discovered
> - it does not try to print to the domU console at all, because xen
>    support will only be discovered at a later point
>
> Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>

Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>

[PATCH v2 2/3] hvc_xen: fix xenboot for DomUs

[Stefano Stabellini]

The xenboot early console has been partially broken for DomU for a long
time: the output would only go to the hypervisor via hypercall
(HYPERVISOR_console_io), while it wouldn't actually go to the DomU
console. The reason is that domU_write_console would return early as no
xencons structs are configured for it.

Add an appropriate xencons struct for xenboot from the xenboot setup
callback.

Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>

---
Changes in v2:
- add return to xenboot_setup_console
---
 drivers/tty/hvc/hvc_xen.c |   32 ++++++++++++++++++++++++++------
 1 file changed, 26 insertions(+), 6 deletions(-)

diff --git a/drivers/tty/hvc/hvc_xen.c b/drivers/tty/hvc/hvc_xen.c
index 68b8ec8..bf787aa 100644
--- a/drivers/tty/hvc/hvc_xen.c
+++ b/drivers/tty/hvc/hvc_xen.c
@@ -246,6 +246,18 @@ err:
 	return -ENODEV;
 }
 
+static int xen_early_pv_console_init(struct xencons_info *info, int vtermno)
+{
+	info->evtchn = xen_start_info->console.domU.evtchn;
+	/* GFN == MFN for PV guest */
+	info->intf = gfn_to_virt(xen_start_info->console.domU.mfn);
+	info->vtermno = vtermno;
+
+	list_add_tail(&info->list, &xenconsoles);
+
+	return 0;
+}
+
 static int xen_pv_console_init(void)
 {
 	struct xencons_info *info;
@@ -265,13 +277,8 @@ static int xen_pv_console_init(void)
 		/* already configured */
 		return 0;
 	}
-	info->evtchn = xen_start_info->console.domU.evtchn;
-	/* GFN == MFN for PV guest */
-	info->intf = gfn_to_virt(xen_start_info->console.domU.mfn);
-	info->vtermno = HVC_COOKIE;
-
 	spin_lock(&xencons_lock);
-	list_add_tail(&info->list, &xenconsoles);
+	xen_early_pv_console_init(info, HVC_COOKIE);
 	spin_unlock(&xencons_lock);
 
 	return 0;
@@ -599,6 +606,18 @@ static int xen_cons_init(void)
 console_initcall(xen_cons_init);
 
 #ifdef CONFIG_EARLY_PRINTK
+static int __init xenboot_setup_console(struct console *console, char *string)
+{
+	static struct xencons_info xenboot;
+
+	if (xen_initial_domain())
+		return 0;
+	if (!xen_pv_domain())
+		return -ENODEV;
+
+	return xen_early_pv_console_init(&xenboot, 0);
+}
+
 static void xenboot_write_console(struct console *console, const char *string,
 				  unsigned len)
 {
@@ -629,6 +648,7 @@ static void xenboot_write_console(struct console *console, const char *string,
 struct console xenboot_console = {
 	.name		= "xenboot",
 	.write		= xenboot_write_console,
+	.setup		= xenboot_setup_console,
 	.flags		= CON_PRINTBUFFER | CON_BOOT | CON_ANYTIME,
 	.index		= -1,
 };
-- 
1.7.10.4

[PATCH v2 2/3] hvc_xen: fix xenboot for DomUs

[Stefano Stabellini]

The xenboot early console has been partially broken for DomU for a long
time: the output would only go to the hypervisor via hypercall
(HYPERVISOR_console_io), while it wouldn't actually go to the DomU
console. The reason is that domU_write_console would return early as no
xencons structs are configured for it.

Add an appropriate xencons struct for xenboot from the xenboot setup
callback.

Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>

---
Changes in v2:
- add return to xenboot_setup_console
---
 drivers/tty/hvc/hvc_xen.c |   32 ++++++++++++++++++++++++++------
 1 file changed, 26 insertions(+), 6 deletions(-)

diff --git a/drivers/tty/hvc/hvc_xen.c b/drivers/tty/hvc/hvc_xen.c
index 68b8ec8..bf787aa 100644
--- a/drivers/tty/hvc/hvc_xen.c
+++ b/drivers/tty/hvc/hvc_xen.c
@@ -246,6 +246,18 @@ err:
 	return -ENODEV;
 }
 
+static int xen_early_pv_console_init(struct xencons_info *info, int vtermno)
+{
+	info->evtchn = xen_start_info->console.domU.evtchn;
+	/* GFN == MFN for PV guest */
+	info->intf = gfn_to_virt(xen_start_info->console.domU.mfn);
+	info->vtermno = vtermno;
+
+	list_add_tail(&info->list, &xenconsoles);
+
+	return 0;
+}
+
 static int xen_pv_console_init(void)
 {
 	struct xencons_info *info;
@@ -265,13 +277,8 @@ static int xen_pv_console_init(void)
 		/* already configured */
 		return 0;
 	}
-	info->evtchn = xen_start_info->console.domU.evtchn;
-	/* GFN == MFN for PV guest */
-	info->intf = gfn_to_virt(xen_start_info->console.domU.mfn);
-	info->vtermno = HVC_COOKIE;
-
 	spin_lock(&xencons_lock);
-	list_add_tail(&info->list, &xenconsoles);
+	xen_early_pv_console_init(info, HVC_COOKIE);
 	spin_unlock(&xencons_lock);
 
 	return 0;
@@ -599,6 +606,18 @@ static int xen_cons_init(void)
 console_initcall(xen_cons_init);
 
 #ifdef CONFIG_EARLY_PRINTK
+static int __init xenboot_setup_console(struct console *console, char *string)
+{
+	static struct xencons_info xenboot;
+
+	if (xen_initial_domain())
+		return 0;
+	if (!xen_pv_domain())
+		return -ENODEV;
+
+	return xen_early_pv_console_init(&xenboot, 0);
+}
+
 static void xenboot_write_console(struct console *console, const char *string,
 				  unsigned len)
 {
@@ -629,6 +648,7 @@ static void xenboot_write_console(struct console *console, const char *string,
 struct console xenboot_console = {
 	.name		= "xenboot",
 	.write		= xenboot_write_console,
+	.setup		= xenboot_setup_console,
 	.flags		= CON_PRINTBUFFER | CON_BOOT | CON_ANYTIME,
 	.index		= -1,
 };
-- 
1.7.10.4

Re: [PATCH v2 2/3] hvc_xen: fix xenboot for DomUs

[Boris Ostrovsky]

On 02/25/2016 07:10 AM, Stefano Stabellini wrote:
> The xenboot early console has been partially broken for DomU for a long
> time: the output would only go to the hypervisor via hypercall
> (HYPERVISOR_console_io), while it wouldn't actually go to the DomU
> console. The reason is that domU_write_console would return early as no
> xencons structs are configured for it.
>
> Add an appropriate xencons struct for xenboot from the xenboot setup
> callback.
>
> Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
>
> ---
> Changes in v2:
> - add return to xenboot_setup_console
> ---
>   drivers/tty/hvc/hvc_xen.c |   32 ++++++++++++++++++++++++++------
>   1 file changed, 26 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/tty/hvc/hvc_xen.c b/drivers/tty/hvc/hvc_xen.c
> index 68b8ec8..bf787aa 100644
> --- a/drivers/tty/hvc/hvc_xen.c
> +++ b/drivers/tty/hvc/hvc_xen.c
> @@ -246,6 +246,18 @@ err:
>   	return -ENODEV;
>   }
>   
> +static int xen_early_pv_console_init(struct xencons_info *info, int vtermno)

Nit: not sure whether "early" is the right word to use here: when called 
from xen_pv_console_init() it's not really on the early path.

Other than that:
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>

> +{
> +	info->evtchn = xen_start_info->console.domU.evtchn;
> +	/* GFN == MFN for PV guest */
> +	info->intf = gfn_to_virt(xen_start_info->console.domU.mfn);
> +	info->vtermno = vtermno;
> +
> +	list_add_tail(&info->list, &xenconsoles);
> +
> +	return 0;
> +}
> +
>   static int xen_pv_console_init(void)
>   {
>   	struct xencons_info *info;
> @@ -265,13 +277,8 @@ static int xen_pv_console_init(void)
>   		/* already configured */
>   		return 0;
>   	}
> -	info->evtchn = xen_start_info->console.domU.evtchn;
> -	/* GFN == MFN for PV guest */
> -	info->intf = gfn_to_virt(xen_start_info->console.domU.mfn);
> -	info->vtermno = HVC_COOKIE;
> -
>   	spin_lock(&xencons_lock);
> -	list_add_tail(&info->list, &xenconsoles);
> +	xen_early_pv_console_init(info, HVC_COOKIE);
>   	spin_unlock(&xencons_lock);
>   
>   	return 0;
> @@ -599,6 +606,18 @@ static int xen_cons_init(void)
>   console_initcall(xen_cons_init);
>   
>   #ifdef CONFIG_EARLY_PRINTK
> +static int __init xenboot_setup_console(struct console *console, char *string)
> +{
> +	static struct xencons_info xenboot;
> +
> +	if (xen_initial_domain())
> +		return 0;
> +	if (!xen_pv_domain())
> +		return -ENODEV;
> +
> +	return xen_early_pv_console_init(&xenboot, 0);
> +}
> +
>   static void xenboot_write_console(struct console *console, const char *string,
>   				  unsigned len)
>   {
> @@ -629,6 +648,7 @@ static void xenboot_write_console(struct console *console, const char *string,
>   struct console xenboot_console = {
>   	.name		= "xenboot",
>   	.write		= xenboot_write_console,
> +	.setup		= xenboot_setup_console,
>   	.flags		= CON_PRINTBUFFER | CON_BOOT | CON_ANYTIME,
>   	.index		= -1,
>   };

Re: [PATCH v2 2/3] hvc_xen: fix xenboot for DomUs

[Stefano Stabellini]

On Thu, 25 Feb 2016, Boris Ostrovsky wrote:
> On 02/25/2016 07:10 AM, Stefano Stabellini wrote:
> > The xenboot early console has been partially broken for DomU for a long
> > time: the output would only go to the hypervisor via hypercall
> > (HYPERVISOR_console_io), while it wouldn't actually go to the DomU
> > console. The reason is that domU_write_console would return early as no
> > xencons structs are configured for it.
> > 
> > Add an appropriate xencons struct for xenboot from the xenboot setup
> > callback.
> > 
> > Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
> > 
> > ---
> > Changes in v2:
> > - add return to xenboot_setup_console
> > ---
> >   drivers/tty/hvc/hvc_xen.c |   32 ++++++++++++++++++++++++++------
> >   1 file changed, 26 insertions(+), 6 deletions(-)
> > 
> > diff --git a/drivers/tty/hvc/hvc_xen.c b/drivers/tty/hvc/hvc_xen.c
> > index 68b8ec8..bf787aa 100644
> > --- a/drivers/tty/hvc/hvc_xen.c
> > +++ b/drivers/tty/hvc/hvc_xen.c
> > @@ -246,6 +246,18 @@ err:
> >   	return -ENODEV;
> >   }
> >   +static int xen_early_pv_console_init(struct xencons_info *info, int
> > vtermno)
> 
> Nit: not sure whether "early" is the right word to use here: when called from
> xen_pv_console_init() it's not really on the early path.

What about xencons_info_pv_init?


> Other than that:
> Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>

Thanks!


> > +{
> > +	info->evtchn = xen_start_info->console.domU.evtchn;
> > +	/* GFN == MFN for PV guest */
> > +	info->intf = gfn_to_virt(xen_start_info->console.domU.mfn);
> > +	info->vtermno = vtermno;
> > +
> > +	list_add_tail(&info->list, &xenconsoles);
> > +
> > +	return 0;
> > +}
> > +
> >   static int xen_pv_console_init(void)
> >   {
> >   	struct xencons_info *info;
> > @@ -265,13 +277,8 @@ static int xen_pv_console_init(void)
> >   		/* already configured */
> >   		return 0;
> >   	}
> > -	info->evtchn = xen_start_info->console.domU.evtchn;
> > -	/* GFN == MFN for PV guest */
> > -	info->intf = gfn_to_virt(xen_start_info->console.domU.mfn);
> > -	info->vtermno = HVC_COOKIE;
> > -
> >   	spin_lock(&xencons_lock);
> > -	list_add_tail(&info->list, &xenconsoles);
> > +	xen_early_pv_console_init(info, HVC_COOKIE);
> >   	spin_unlock(&xencons_lock);
> >     	return 0;
> > @@ -599,6 +606,18 @@ static int xen_cons_init(void)
> >   console_initcall(xen_cons_init);
> >     #ifdef CONFIG_EARLY_PRINTK
> > +static int __init xenboot_setup_console(struct console *console, char
> > *string)
> > +{
> > +	static struct xencons_info xenboot;
> > +
> > +	if (xen_initial_domain())
> > +		return 0;
> > +	if (!xen_pv_domain())
> > +		return -ENODEV;
> > +
> > +	return xen_early_pv_console_init(&xenboot, 0);
> > +}
> > +
> >   static void xenboot_write_console(struct console *console, const char
> > *string,
> >   				  unsigned len)
> >   {
> > @@ -629,6 +648,7 @@ static void xenboot_write_console(struct console
> > *console, const char *string,
> >   struct console xenboot_console = {
> >   	.name		= "xenboot",
> >   	.write		= xenboot_write_console,
> > +	.setup		= xenboot_setup_console,
> >   	.flags		= CON_PRINTBUFFER | CON_BOOT | CON_ANYTIME,
> >   	.index		= -1,
> >   };
> 

Re: [PATCH v2 2/3] hvc_xen: fix xenboot for DomUs

[Stefano Stabellini]

On Thu, 25 Feb 2016, Boris Ostrovsky wrote:
> On 02/25/2016 07:10 AM, Stefano Stabellini wrote:
> > The xenboot early console has been partially broken for DomU for a long
> > time: the output would only go to the hypervisor via hypercall
> > (HYPERVISOR_console_io), while it wouldn't actually go to the DomU
> > console. The reason is that domU_write_console would return early as no
> > xencons structs are configured for it.
> > 
> > Add an appropriate xencons struct for xenboot from the xenboot setup
> > callback.
> > 
> > Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
> > 
> > ---
> > Changes in v2:
> > - add return to xenboot_setup_console
> > ---
> >   drivers/tty/hvc/hvc_xen.c |   32 ++++++++++++++++++++++++++------
> >   1 file changed, 26 insertions(+), 6 deletions(-)
> > 
> > diff --git a/drivers/tty/hvc/hvc_xen.c b/drivers/tty/hvc/hvc_xen.c
> > index 68b8ec8..bf787aa 100644
> > --- a/drivers/tty/hvc/hvc_xen.c
> > +++ b/drivers/tty/hvc/hvc_xen.c
> > @@ -246,6 +246,18 @@ err:
> >   	return -ENODEV;
> >   }
> >   +static int xen_early_pv_console_init(struct xencons_info *info, int
> > vtermno)
> 
> Nit: not sure whether "early" is the right word to use here: when called from
> xen_pv_console_init() it's not really on the early path.

What about xencons_info_pv_init?


> Other than that:
> Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>

Thanks!


> > +{
> > +	info->evtchn = xen_start_info->console.domU.evtchn;
> > +	/* GFN == MFN for PV guest */
> > +	info->intf = gfn_to_virt(xen_start_info->console.domU.mfn);
> > +	info->vtermno = vtermno;
> > +
> > +	list_add_tail(&info->list, &xenconsoles);
> > +
> > +	return 0;
> > +}
> > +
> >   static int xen_pv_console_init(void)
> >   {
> >   	struct xencons_info *info;
> > @@ -265,13 +277,8 @@ static int xen_pv_console_init(void)
> >   		/* already configured */
> >   		return 0;
> >   	}
> > -	info->evtchn = xen_start_info->console.domU.evtchn;
> > -	/* GFN == MFN for PV guest */
> > -	info->intf = gfn_to_virt(xen_start_info->console.domU.mfn);
> > -	info->vtermno = HVC_COOKIE;
> > -
> >   	spin_lock(&xencons_lock);
> > -	list_add_tail(&info->list, &xenconsoles);
> > +	xen_early_pv_console_init(info, HVC_COOKIE);
> >   	spin_unlock(&xencons_lock);
> >     	return 0;
> > @@ -599,6 +606,18 @@ static int xen_cons_init(void)
> >   console_initcall(xen_cons_init);
> >     #ifdef CONFIG_EARLY_PRINTK
> > +static int __init xenboot_setup_console(struct console *console, char
> > *string)
> > +{
> > +	static struct xencons_info xenboot;
> > +
> > +	if (xen_initial_domain())
> > +		return 0;
> > +	if (!xen_pv_domain())
> > +		return -ENODEV;
> > +
> > +	return xen_early_pv_console_init(&xenboot, 0);
> > +}
> > +
> >   static void xenboot_write_console(struct console *console, const char
> > *string,
> >   				  unsigned len)
> >   {
> > @@ -629,6 +648,7 @@ static void xenboot_write_console(struct console
> > *console, const char *string,
> >   struct console xenboot_console = {
> >   	.name		= "xenboot",
> >   	.write		= xenboot_write_console,
> > +	.setup		= xenboot_setup_console,
> >   	.flags		= CON_PRINTBUFFER | CON_BOOT | CON_ANYTIME,
> >   	.index		= -1,
> >   };
> 

Re: [PATCH v2 2/3] hvc_xen: fix xenboot for DomUs

[Boris Ostrovsky]

On 02/26/2016 10:39 AM, Stefano Stabellini wrote:
> On Thu, 25 Feb 2016, Boris Ostrovsky wrote:
>> On 02/25/2016 07:10 AM, Stefano Stabellini wrote:
>>> The xenboot early console has been partially broken for DomU for a long
>>> time: the output would only go to the hypervisor via hypercall
>>> (HYPERVISOR_console_io), while it wouldn't actually go to the DomU
>>> console. The reason is that domU_write_console would return early as no
>>> xencons structs are configured for it.
>>>
>>> Add an appropriate xencons struct for xenboot from the xenboot setup
>>> callback.
>>>
>>> Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
>>>
>>> ---
>>> Changes in v2:
>>> - add return to xenboot_setup_console
>>> ---
>>>    drivers/tty/hvc/hvc_xen.c |   32 ++++++++++++++++++++++++++------
>>>    1 file changed, 26 insertions(+), 6 deletions(-)
>>>
>>> diff --git a/drivers/tty/hvc/hvc_xen.c b/drivers/tty/hvc/hvc_xen.c
>>> index 68b8ec8..bf787aa 100644
>>> --- a/drivers/tty/hvc/hvc_xen.c
>>> +++ b/drivers/tty/hvc/hvc_xen.c
>>> @@ -246,6 +246,18 @@ err:
>>>    	return -ENODEV;
>>>    }
>>>    +static int xen_early_pv_console_init(struct xencons_info *info, int
>>> vtermno)
>> Nit: not sure whether "early" is the right word to use here: when called from
>> xen_pv_console_init() it's not really on the early path.
> What about xencons_info_pv_init?

Sure, that sounds good.

(BTW, I didn't mean that my R-b was conditional on this)


-boris

>
>
>> Other than that:
>> Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
> Thanks!
>
>
>>> +{
>>> +	info->evtchn = xen_start_info->console.domU.evtchn;
>>> +	/* GFN == MFN for PV guest */
>>> +	info->intf = gfn_to_virt(xen_start_info->console.domU.mfn);
>>> +	info->vtermno = vtermno;
>>> +
>>> +	list_add_tail(&info->list, &xenconsoles);
>>> +
>>> +	return 0;
>>> +}
>>> +
>>>    static int xen_pv_console_init(void)
>>>    {
>>>    	struct xencons_info *info;
>>> @@ -265,13 +277,8 @@ static int xen_pv_console_init(void)
>>>    		/* already configured */
>>>    		return 0;
>>>    	}
>>> -	info->evtchn = xen_start_info->console.domU.evtchn;
>>> -	/* GFN == MFN for PV guest */
>>> -	info->intf = gfn_to_virt(xen_start_info->console.domU.mfn);
>>> -	info->vtermno = HVC_COOKIE;
>>> -
>>>    	spin_lock(&xencons_lock);
>>> -	list_add_tail(&info->list, &xenconsoles);
>>> +	xen_early_pv_console_init(info, HVC_COOKIE);
>>>    	spin_unlock(&xencons_lock);
>>>      	return 0;
>>> @@ -599,6 +606,18 @@ static int xen_cons_init(void)
>>>    console_initcall(xen_cons_init);
>>>      #ifdef CONFIG_EARLY_PRINTK
>>> +static int __init xenboot_setup_console(struct console *console, char
>>> *string)
>>> +{
>>> +	static struct xencons_info xenboot;
>>> +
>>> +	if (xen_initial_domain())
>>> +		return 0;
>>> +	if (!xen_pv_domain())
>>> +		return -ENODEV;
>>> +
>>> +	return xen_early_pv_console_init(&xenboot, 0);
>>> +}
>>> +
>>>    static void xenboot_write_console(struct console *console, const char
>>> *string,
>>>    				  unsigned len)
>>>    {
>>> @@ -629,6 +648,7 @@ static void xenboot_write_console(struct console
>>> *console, const char *string,
>>>    struct console xenboot_console = {
>>>    	.name		= "xenboot",
>>>    	.write		= xenboot_write_console,
>>> +	.setup		= xenboot_setup_console,
>>>    	.flags		= CON_PRINTBUFFER | CON_BOOT | CON_ANYTIME,
>>>    	.index		= -1,
>>>    };

Re: [PATCH v2 2/3] hvc_xen: fix xenboot for DomUs

[Stefano Stabellini]

On Fri, 26 Feb 2016, Boris Ostrovsky wrote:
> On 02/26/2016 10:39 AM, Stefano Stabellini wrote:
> > On Thu, 25 Feb 2016, Boris Ostrovsky wrote:
> > > On 02/25/2016 07:10 AM, Stefano Stabellini wrote:
> > > > The xenboot early console has been partially broken for DomU for a long
> > > > time: the output would only go to the hypervisor via hypercall
> > > > (HYPERVISOR_console_io), while it wouldn't actually go to the DomU
> > > > console. The reason is that domU_write_console would return early as no
> > > > xencons structs are configured for it.
> > > > 
> > > > Add an appropriate xencons struct for xenboot from the xenboot setup
> > > > callback.
> > > > 
> > > > Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
> > > > 
> > > > ---
> > > > Changes in v2:
> > > > - add return to xenboot_setup_console
> > > > ---
> > > >    drivers/tty/hvc/hvc_xen.c |   32 ++++++++++++++++++++++++++------
> > > >    1 file changed, 26 insertions(+), 6 deletions(-)
> > > > 
> > > > diff --git a/drivers/tty/hvc/hvc_xen.c b/drivers/tty/hvc/hvc_xen.c
> > > > index 68b8ec8..bf787aa 100644
> > > > --- a/drivers/tty/hvc/hvc_xen.c
> > > > +++ b/drivers/tty/hvc/hvc_xen.c
> > > > @@ -246,6 +246,18 @@ err:
> > > >    	return -ENODEV;
> > > >    }
> > > >    +static int xen_early_pv_console_init(struct xencons_info *info, int
> > > > vtermno)
> > > Nit: not sure whether "early" is the right word to use here: when called
> > > from
> > > xen_pv_console_init() it's not really on the early path.
> > What about xencons_info_pv_init?
> 
> Sure, that sounds good.
> 
> (BTW, I didn't mean that my R-b was conditional on this)

No problem. I made this change and committed the three patches to
for-linus-4.6, based on v4.5-rc5.

Re: [PATCH v2 2/3] hvc_xen: fix xenboot for DomUs

[Stefano Stabellini]

On Fri, 26 Feb 2016, Boris Ostrovsky wrote:
> On 02/26/2016 10:39 AM, Stefano Stabellini wrote:
> > On Thu, 25 Feb 2016, Boris Ostrovsky wrote:
> > > On 02/25/2016 07:10 AM, Stefano Stabellini wrote:
> > > > The xenboot early console has been partially broken for DomU for a long
> > > > time: the output would only go to the hypervisor via hypercall
> > > > (HYPERVISOR_console_io), while it wouldn't actually go to the DomU
> > > > console. The reason is that domU_write_console would return early as no
> > > > xencons structs are configured for it.
> > > > 
> > > > Add an appropriate xencons struct for xenboot from the xenboot setup
> > > > callback.
> > > > 
> > > > Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
> > > > 
> > > > ---
> > > > Changes in v2:
> > > > - add return to xenboot_setup_console
> > > > ---
> > > >    drivers/tty/hvc/hvc_xen.c |   32 ++++++++++++++++++++++++++------
> > > >    1 file changed, 26 insertions(+), 6 deletions(-)
> > > > 
> > > > diff --git a/drivers/tty/hvc/hvc_xen.c b/drivers/tty/hvc/hvc_xen.c
> > > > index 68b8ec8..bf787aa 100644
> > > > --- a/drivers/tty/hvc/hvc_xen.c
> > > > +++ b/drivers/tty/hvc/hvc_xen.c
> > > > @@ -246,6 +246,18 @@ err:
> > > >    	return -ENODEV;
> > > >    }
> > > >    +static int xen_early_pv_console_init(struct xencons_info *info, int
> > > > vtermno)
> > > Nit: not sure whether "early" is the right word to use here: when called
> > > from
> > > xen_pv_console_init() it's not really on the early path.
> > What about xencons_info_pv_init?
> 
> Sure, that sounds good.
> 
> (BTW, I didn't mean that my R-b was conditional on this)

No problem. I made this change and committed the three patches to
for-linus-4.6, based on v4.5-rc5.

[PATCH v2 3/3] hvc_xen: make early_printk work with HVM guests

[Stefano Stabellini]

Refactor the existing code in xen_raw_console_write to get the generic
early_printk console work with HVM guests.

Take the opportunity to replace the outb loop with a single outsb call
to reduce the number of vmexit.

Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
---
 drivers/tty/hvc/hvc_xen.c |   27 ++++++++++++++++-----------
 1 file changed, 16 insertions(+), 11 deletions(-)

diff --git a/drivers/tty/hvc/hvc_xen.c b/drivers/tty/hvc/hvc_xen.c
index bf787aa..a1c3746 100644
--- a/drivers/tty/hvc/hvc_xen.c
+++ b/drivers/tty/hvc/hvc_xen.c
@@ -605,6 +605,16 @@ static int xen_cons_init(void)
 }
 console_initcall(xen_cons_init);
 
+#ifdef CONFIG_X86
+static void xen_hvm_early_write(uint32_t vtermno, const char *str, int len)
+{
+	if (xen_cpuid_base())
+		outsb(0xe9, str, len);
+}
+#else
+static void xen_hvm_early_write(uint32_t vtermno, const char *str, int len) { }
+#endif
+
 #ifdef CONFIG_EARLY_PRINTK
 static int __init xenboot_setup_console(struct console *console, char *string)
 {
@@ -624,8 +634,10 @@ static void xenboot_write_console(struct console *console, const char *string,
 	unsigned int linelen, off = 0;
 	const char *pos;
 
-	if (!xen_pv_domain())
+	if (!xen_pv_domain()) {
+		xen_hvm_early_write(0, string, len);
 		return;
+	}
 
 	dom0_write_console(0, string, len);
 
@@ -661,17 +673,10 @@ void xen_raw_console_write(const char *str)
 
 	if (xen_domain()) {
 		rc = dom0_write_console(0, str, len);
-#ifdef CONFIG_X86
-		if (rc == -ENOSYS && xen_hvm_domain())
-			goto outb_print;
-
-	} else if (xen_cpuid_base()) {
-		int i;
-outb_print:
-		for (i = 0; i < len; i++)
-			outb(str[i], 0xe9);
-#endif
+		if (rc != -ENOSYS || !xen_hvm_domain())
+			return;
 	}
+	xen_hvm_early_write(0, str, len);
 }
 
 void xen_raw_printk(const char *fmt, ...)
-- 
1.7.10.4

[PATCH v2 3/3] hvc_xen: make early_printk work with HVM guests

[Stefano Stabellini]

Refactor the existing code in xen_raw_console_write to get the generic
early_printk console work with HVM guests.

Take the opportunity to replace the outb loop with a single outsb call
to reduce the number of vmexit.

Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
---
 drivers/tty/hvc/hvc_xen.c |   27 ++++++++++++++++-----------
 1 file changed, 16 insertions(+), 11 deletions(-)

diff --git a/drivers/tty/hvc/hvc_xen.c b/drivers/tty/hvc/hvc_xen.c
index bf787aa..a1c3746 100644
--- a/drivers/tty/hvc/hvc_xen.c
+++ b/drivers/tty/hvc/hvc_xen.c
@@ -605,6 +605,16 @@ static int xen_cons_init(void)
 }
 console_initcall(xen_cons_init);
 
+#ifdef CONFIG_X86
+static void xen_hvm_early_write(uint32_t vtermno, const char *str, int len)
+{
+	if (xen_cpuid_base())
+		outsb(0xe9, str, len);
+}
+#else
+static void xen_hvm_early_write(uint32_t vtermno, const char *str, int len) { }
+#endif
+
 #ifdef CONFIG_EARLY_PRINTK
 static int __init xenboot_setup_console(struct console *console, char *string)
 {
@@ -624,8 +634,10 @@ static void xenboot_write_console(struct console *console, const char *string,
 	unsigned int linelen, off = 0;
 	const char *pos;
 
-	if (!xen_pv_domain())
+	if (!xen_pv_domain()) {
+		xen_hvm_early_write(0, string, len);
 		return;
+	}
 
 	dom0_write_console(0, string, len);
 
@@ -661,17 +673,10 @@ void xen_raw_console_write(const char *str)
 
 	if (xen_domain()) {
 		rc = dom0_write_console(0, str, len);
-#ifdef CONFIG_X86
-		if (rc == -ENOSYS && xen_hvm_domain())
-			goto outb_print;
-
-	} else if (xen_cpuid_base()) {
-		int i;
-outb_print:
-		for (i = 0; i < len; i++)
-			outb(str[i], 0xe9);
-#endif
+		if (rc != -ENOSYS || !xen_hvm_domain())
+			return;
 	}
+	xen_hvm_early_write(0, str, len);
 }
 
 void xen_raw_printk(const char *fmt, ...)
-- 
1.7.10.4

Re: [PATCH v2 3/3] hvc_xen: make early_printk work with HVM guests

[Boris Ostrovsky]

On 02/25/2016 07:10 AM, Stefano Stabellini wrote:
> Refactor the existing code in xen_raw_console_write to get the generic
> early_printk console work with HVM guests.
>
> Take the opportunity to replace the outb loop with a single outsb call
> to reduce the number of vmexit.
>
> Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>

Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>

Re: [PATCH v2 0/3] hvc_xen: fix xenboot on x86 and export to ARM

[Stefano Stabellini]

On Thu, 25 Feb 2016, Stefano Stabellini wrote:
> Hi all,
> 
> this series exports xenboot on ARM and ARM64 as earlycon, and gets
> xenboot fully working again for PV DomUs on x86 (currently the xenboot
> output only goes to the hypervisor via hypercall and not to the DomU
> console).
> 
> Cheers,
> 
> Stefano
> 
> 
> Stefano Stabellini (3):
>       hvc_xen: add earlycon support
>       hvc_xen: fix xenboot for DomUs
>       hvc_xen: make early_printk work with HVM guests

Boris has reviewed all the patches in the series. Are you all happy if I
go ahead and  commit the whole series to xentip/for-linus-4.6 ?

Re: [PATCH v2 0/3] hvc_xen: fix xenboot on x86 and export to ARM

[Stefano Stabellini]

On Thu, 25 Feb 2016, Stefano Stabellini wrote:
> Hi all,
> 
> this series exports xenboot on ARM and ARM64 as earlycon, and gets
> xenboot fully working again for PV DomUs on x86 (currently the xenboot
> output only goes to the hypervisor via hypercall and not to the DomU
> console).
> 
> Cheers,
> 
> Stefano
> 
> 
> Stefano Stabellini (3):
>       hvc_xen: add earlycon support
>       hvc_xen: fix xenboot for DomUs
>       hvc_xen: make early_printk work with HVM guests

Boris has reviewed all the patches in the series. Are you all happy if I
go ahead and  commit the whole series to xentip/for-linus-4.6 ?

Re: [PATCH v2 0/3] hvc_xen: fix xenboot on x86 and export to ARM

[David Vrabel]

On 26/02/16 14:28, Stefano Stabellini wrote:
> On Thu, 25 Feb 2016, Stefano Stabellini wrote:
>> Hi all,
>>
>> this series exports xenboot on ARM and ARM64 as earlycon, and gets
>> xenboot fully working again for PV DomUs on x86 (currently the xenboot
>> output only goes to the hypervisor via hypercall and not to the DomU
>> console).
>>
>> Cheers,
>>
>> Stefano
>>
>>
>> Stefano Stabellini (3):
>>       hvc_xen: add earlycon support
>>       hvc_xen: fix xenboot for DomUs
>>       hvc_xen: make early_printk work with HVM guests
> 
> Boris has reviewed all the patches in the series. Are you all happy if I
> go ahead and  commit the whole series to xentip/for-linus-4.6 ?

Yes.  I don't think that branch exists yet so you'll need to create it.
 I usually use the latest tagged rc as a base.

David

Re: [PATCH v2 0/3] hvc_xen: fix xenboot on x86 and export to ARM

[David Vrabel]

On 26/02/16 14:28, Stefano Stabellini wrote:
> On Thu, 25 Feb 2016, Stefano Stabellini wrote:
>> Hi all,
>>
>> this series exports xenboot on ARM and ARM64 as earlycon, and gets
>> xenboot fully working again for PV DomUs on x86 (currently the xenboot
>> output only goes to the hypervisor via hypercall and not to the DomU
>> console).
>>
>> Cheers,
>>
>> Stefano
>>
>>
>> Stefano Stabellini (3):
>>       hvc_xen: add earlycon support
>>       hvc_xen: fix xenboot for DomUs
>>       hvc_xen: make early_printk work with HVM guests
> 
> Boris has reviewed all the patches in the series. Are you all happy if I
> go ahead and  commit the whole series to xentip/for-linus-4.6 ?

Yes.  I don't think that branch exists yet so you'll need to create it.
 I usually use the latest tagged rc as a base.

David