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

Re: [PATCH v2 13/40] xen/mpu: introduce unified function setup_early_uart to map early UART



Hi Penny,

On 31/01/2023 05:38, Penny Zheng wrote:
-----Original Message-----
From: Julien Grall <julien@xxxxxxx>
Sent: Monday, January 30, 2023 6:00 PM
To: Penny Zheng <Penny.Zheng@xxxxxxx>; xen-devel@xxxxxxxxxxxxxxxxxxxx
Cc: Wei Chen <Wei.Chen@xxxxxxx>; Stefano Stabellini
<sstabellini@xxxxxxxxxx>; Bertrand Marquis <Bertrand.Marquis@xxxxxxx>;
Volodymyr Babchuk <Volodymyr_Babchuk@xxxxxxxx>
Subject: Re: [PATCH v2 13/40] xen/mpu: introduce unified function
setup_early_uart to map early UART



On 30/01/2023 06:24, Penny Zheng wrote:
Hi, Julien

Hi Penny,

-----Original Message-----
From: Julien Grall <julien@xxxxxxx>
Sent: Sunday, January 29, 2023 3:43 PM
To: Penny Zheng <Penny.Zheng@xxxxxxx>; xen-
devel@xxxxxxxxxxxxxxxxxxxx
Cc: Wei Chen <Wei.Chen@xxxxxxx>; Stefano Stabellini
<sstabellini@xxxxxxxxxx>; Bertrand Marquis
<Bertrand.Marquis@xxxxxxx>; Volodymyr Babchuk
<Volodymyr_Babchuk@xxxxxxxx>
Subject: Re: [PATCH v2 13/40] xen/mpu: introduce unified function
setup_early_uart to map early UART

Hi Penny,

On 29/01/2023 06:17, Penny Zheng wrote:
-----Original Message-----
From: Julien Grall <julien@xxxxxxx>
Sent: Wednesday, January 25, 2023 3:09 AM
To: Penny Zheng <Penny.Zheng@xxxxxxx>; xen-
devel@xxxxxxxxxxxxxxxxxxxx
Cc: Wei Chen <Wei.Chen@xxxxxxx>; Stefano Stabellini
<sstabellini@xxxxxxxxxx>; Bertrand Marquis
<Bertrand.Marquis@xxxxxxx>; Volodymyr Babchuk
<Volodymyr_Babchuk@xxxxxxxx>
Subject: Re: [PATCH v2 13/40] xen/mpu: introduce unified function
setup_early_uart to map early UART

Hi Peny,

Hi Julien,


On 13/01/2023 05:28, Penny Zheng wrote:
In MMU system, we map the UART in the fixmap (when earlyprintk is
used).
However in MPU system, we map the UART with a transient MPU
memory
region.

So we introduce a new unified function setup_early_uart to replace
the previous setup_fixmap.

Signed-off-by: Penny Zheng <penny.zheng@xxxxxxx>
Signed-off-by: Wei Chen <wei.chen@xxxxxxx>
---
     xen/arch/arm/arm64/head.S               |  2 +-
     xen/arch/arm/arm64/head_mmu.S           |  4 +-
     xen/arch/arm/arm64/head_mpu.S           | 52
+++++++++++++++++++++++++
     xen/arch/arm/include/asm/early_printk.h |  1 +
     4 files changed, 56 insertions(+), 3 deletions(-)

diff --git a/xen/arch/arm/arm64/head.S
b/xen/arch/arm/arm64/head.S
index 7f3f973468..a92883319d 100644
--- a/xen/arch/arm/arm64/head.S
+++ b/xen/arch/arm/arm64/head.S
@@ -272,7 +272,7 @@ primary_switched:
              * afterwards.
              */
             bl    remove_identity_mapping
-        bl    setup_fixmap
+        bl    setup_early_uart
     #ifdef CONFIG_EARLY_PRINTK
             /* Use a virtual address to access the UART. */
             ldr   x23, =EARLY_UART_VIRTUAL_ADDRESS
diff --git a/xen/arch/arm/arm64/head_mmu.S
b/xen/arch/arm/arm64/head_mmu.S index b59c40495f..a19b7c873d
100644
--- a/xen/arch/arm/arm64/head_mmu.S
+++ b/xen/arch/arm/arm64/head_mmu.S
@@ -312,7 +312,7 @@ ENDPROC(remove_identity_mapping)
      *
      * Clobbers x0 - x3
      */
-ENTRY(setup_fixmap)
+ENTRY(setup_early_uart)

This function is doing more than enable the early UART. It also
setups the fixmap even earlyprintk is not configured.

True, true.
I've thoroughly read the MMU implementation of setup_fixmap, and
I'll try to split it up.


I am not entirely sure what could be the name. Maybe this needs to
be split further.

     #ifdef CONFIG_EARLY_PRINTK
             /* Add UART to the fixmap table */
             ldr   x0, =EARLY_UART_VIRTUAL_ADDRESS
@@ -325,7 +325,7 @@ ENTRY(setup_fixmap)
             dsb   nshst

             ret
-ENDPROC(setup_fixmap)
+ENDPROC(setup_early_uart)

     /* Fail-stop */
     fail:   PRINT("- Boot failed -\r\n")
diff --git a/xen/arch/arm/arm64/head_mpu.S
b/xen/arch/arm/arm64/head_mpu.S index e2ac69b0cc..72d1e0863d
100644
--- a/xen/arch/arm/arm64/head_mpu.S
+++ b/xen/arch/arm/arm64/head_mpu.S
@@ -18,8 +18,10 @@
     #define REGION_TEXT_PRBAR       0x38    /* SH=11 AP=10 XN=00 */
     #define REGION_RO_PRBAR         0x3A    /* SH=11 AP=10 XN=10 */
     #define REGION_DATA_PRBAR       0x32    /* SH=11 AP=00 XN=10 */
+#define REGION_DEVICE_PRBAR     0x22    /* SH=10 AP=00 XN=10 */

     #define REGION_NORMAL_PRLAR     0x0f    /* NS=0 ATTR=111 EN=1
*/
+#define REGION_DEVICE_PRLAR     0x09    /* NS=0 ATTR=100 EN=1 */

     /*
      * Macro to round up the section address to be PAGE_SIZE
aligned @@
-334,6 +336,56 @@ ENTRY(enable_mm)
         ret
     ENDPROC(enable_mm)

+/*
+ * Map the early UART with a new transient MPU memory region.
+ *

Missing "Inputs: "

+ * x27: region selector
+ * x28: prbar
+ * x29: prlar
+ *
+ * Clobbers x0 - x4
+ *
+ */
+ENTRY(setup_early_uart)
+#ifdef CONFIG_EARLY_PRINTK
+    /* stack LR as write_pr will be called later like nested function */
+    mov   x3, lr
+
+    /*
+     * MPU region for early UART is a transient region, since it will be
+     * replaced by specific device memory layout when FDT gets
parsed.

I would rather not mention "FDT" here because this code is
independent to the firmware table used.

However, any reason to use a transient region rather than the one
that will be used for the UART driver?


We don’t want to define a MPU region for each device driver. It will
exhaust MPU regions very quickly.
What the usual size of an MPU?

However, even if you don't want to define one for every device, it
still seem to be sensible to define a fixed temporary one for the
early UART as this would simplify the assembly code.


We will add fixed MPU regions for Xen static heap in function setup_mm.
If we put early uart region in front(fixed region place), it will
leave holes later after removing it.

Why? The entry could be re-used to map the devices entry.



In commit " [PATCH v2 28/40] xen/mpu: map boot module section in
MPU
system",

Did you mean patch #27?

A new FDT property `mpu,device-memory-section` will be introduced
for users to statically configure the whole system device memory
with the
least number of memory regions in Device Tree.
This section shall cover all devices that will be used in Xen, like
`UART`,
`GIC`, etc.
For FVP_BaseR_AEMv8R, we have the following definition:
```
mpu,device-memory-section = <0x0 0x80000000 0x0 0x7ffff000>; ```

I am a bit worry this will be a recipe for mistake. Do you have an
example where the MPU will be exhausted if we reserve some entries
for each device (or some)?


Yes, we have internal platform where MPU regions are only 16.

Internal is in silicon (e.g. real) or virtual platform?


Sorry, we met this kind of type platform is all I'm allowed to say.
Due to NDA, I couldn’t tell more.

  It will almost eat up
all MPU regions based on current implementation, when launching two
guests in platform.

Let's calculate the most simple scenario:
The following is MPU-related static configuration in device tree:
```
          mpu,boot-module-section = <0x0 0x10000000 0x0 0x10000000>;
          mpu,guest-memory-section = <0x0 0x20000000 0x0 0x40000000>;
          mpu,device-memory-section = <0x0 0x80000000 0x0 0x7ffff000>;
          mpu,shared-memory-section = <0x0 0x7a000000 0x0 0x02000000>;

          xen,static-heap = <0x0 0x60000000 0x0 0x1a000000>; ``` At the
end of the boot, before reshuffling, the MPU region usage will be as
follows:
7 (defined in assembly) + FDT(early_fdt_map) + 5 (at least one region for
each "mpu,xxx-memory-section").

Can you list the 7 sections? Is it including the init section?


Yes, I'll draw the layout for you:

Thanks!

'''
  Xen MPU Map before reorg:

xen_mpumap[0] : Xen text
xen_mpumap[1] : Xen read-only data
xen_mpumap[2] : Xen read-only after init data
xen_mpumap[3] : Xen read-write data
xen_mpumap[4] : Xen BSS
xen_mpumap[5] : Xen static heap
......
xen_mpumap[max_xen_mpumap - 7]: Static shared memory section
xen_mpumap[max_xen_mpumap - 6]: Boot Module memory section(kernel, initramfs, 
etc)
xen_mpumap[max_xen_mpumap - 5]: Device memory section
xen_mpumap[max_xen_mpumap - 4]: Guest memory section
xen_mpumap[max_xen_mpumap - 3]: Early FDT
xen_mpumap[max_xen_mpumap - 2]: Xen init data
xen_mpumap[max_xen_mpumap - 1]: Xen init text

In the end of boot, function init_done will do the reorg and boot-only region 
clean-up:

Xen MPU Map after reorg(idle vcpu):

xen_mpumap[0] : Xen text
xen_mpumap[1] : Xen read-only data
xen_mpumap[2] : Xen read-only after init data

In theory 1 and 2 could be merged after boot. But I guess it might be complicated?

xen_mpumap[3] : Xen read-write data
xen_mpumap[4] : Xen BSS
xen_mpumap[5] : Xen static heap
xen_mpumap[6] : Guest memory section

Why do you need to map the "Guest memory section" for the idle vCPU?

xen_mpumap[7] : Device memory section

I might be missing some context here. But why this section is not also mapped in the context of the guest vCPU?

For instance, how would you write to the serial console when the context is the guest vCPU?

xen_mpumap[6] : Static shared memory section

Xen MPU Map on runtime(guest vcpu):

xen_mpumap[0] : Xen text
xen_mpumap[1] : Xen read-only data
xen_mpumap[2] : Xen read-only after init data
xen_mpumap[3] : Xen read-write data
xen_mpumap[4] : Xen BSS
xen_mpumap[5] : Xen static heap
xen_mpumap[6] : Guest memory
xen_mpumap[7] : vGIC map
xen_mpumap[8] : vPL011 map

I was expected the PL011 to be fully emulated. So why is this necessary?

xen_mpumap[9] : Passthrough device map(UART, etc)
xen_mpumap[10] : Static shared memory section


That will be already at least 13 MPU regions ;\.

The section I am the most concern of is mpu,device-memory-section
because it would likely mean that all the devices will be mapped in Xen.
Is there any risk that the guest may use different memory attribute?


Yes, on current implementation, per-domain vgic, vpl011, and passthrough device 
map
will be individually added into per-domain P2M mapping, then when switching 
into guest
vcpu from xen idle vcpu, device memory section will be replaced by vgic, 
vpl011, passthrough
device map.

Per above, I am not entirely sure how you could remove the device memory section when using the guest vCPU.

Now about the layout between init and runtime. From previous discussion, you said you didn't want to have init section to be fixed because of the section "Xen static heap".

Furthermore, you also mention that you didn't want a bitmap. So how about the following for the assembly part:

xen_mpumap[0] : Xen text
xen_mpumap[1] : Xen read-only data
xen_mpumap[2] : Xen read-only after init data
xen_mpumap[3] : Xen read-write data
xen_mpumap[4] : Xen BSS
xen_mpumap[5]: Early FDT
xen_mpumap[6]: Xen init data
xen_mpumap[7]: Xen init text
xen_mpumap[8]: Early UART (optional)

Then when you switch to C, you could have:

xen_mpumap[0] : Xen text
xen_mpumap[1] : Xen read-only data
xen_mpumap[2] : Xen read-only after init data
xen_mpumap[3] : Xen read-write data
xen_mpumap[4] : Xen BSS
xen_mpumap[5]: Early FDT
xen_mpumap[6]: Xen init data
xen_mpumap[7]: Xen init text

xen_mpumap[max_xen_mpumap - 4]: Device memory section
xen_mpumap[max_xen_mpumap - 3]: Guest memory section
xen_mpumap[max_xen_mpumap - 2]: Static shared memory section
xen_mpumap[max_xen_mpumap - 1] : Xen static heap

And at runtime, you would keep the "Xen static heap" right at the end of the MPU and keep the middle entries as the switchable one.

There would be not bitmap with this solution and all the entries for the assembly code would be fixed.

Cheers,

--
Julien Grall



 


Rackspace

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