Skip to contentSkip to conversation

Minor improvement to early boot code

PATCH Superseded Wei Liu 3 participants 15 messages Sep 28, 2018

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Wei Liu

author
#0 lore raw

cover Minor improvement to early boot code

Wei Liu (2):
x86: make sure module array is large enough in pvh-boot.c
xen: make opt_xen_console true before console initialisation

 xen/arch/x86/guest/pvh-boot.c | 9 ++++++++-
 xen/drivers/char/console.c    | 7 ++++++-
 2 files changed, 14 insertions(+), 2 deletions(-)
Signature · 7 lines
2.11.0


_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
author
#1 lore raw

PATCH 1/2 x86: make sure module array is large enough in pvh-boot.c

The relocation code in __start_xen requires one extra element in the module array. By the looks of it the temporary array is already large enough. Add a BUG_ON to catch any issue in the future.

While at it, turn another ASSERT to BUG_ON as well.

---
 xen/arch/x86/guest/pvh-boot.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)
xen/arch/x86/guest/pvh-boot.c +8 -1
index 0e9e5bfdf6..2e103105fa 100644
@@ -42,7 +42,14 @@ static void __init convert_pvh_info(void)
     module_t *mod;
     unsigned int i;
 
-    ASSERT(pvh_info->magic == XEN_HVM_START_MAGIC_VALUE);
+    BUG_ON(pvh_info->magic != XEN_HVM_START_MAGIC_VALUE);
+
+    /*
+     * Temporary module array needs to be at least one element bigger than
+     * required. The extra element is used to aid relocation. See
+     * arch/x86/setup.c:__start_xen().
+     */
+    BUG_ON(ARRAY_SIZE(pvh_mbi_mods) <= pvh_info->nr_modules);
 
     /*
      * Turn hvm_start_info into mbi. Luckily all modules are placed under 4GB
Signature · 7 lines
2.11.0


_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
#2 lore raw

Re: [PATCH 1/2] x86: make sure module array is large enough in pvh-boot.c

Quoted 6 lines
Quoted 1 line
Quoted 1 line · On 28.09.18 at 10:24, <[email protected]> wrote:

The relocation code in __start_xen requires one extra element in the module array. By the looks of it the temporary array is already large enough. Add a BUG_ON to catch any issue in the future.

While at it, turn another ASSERT to BUG_ON as well.

Hmm, a little strange - I thought we had agreed on panic() before. The extra output BUG_ON() produces is unlikely to be helpful here. With this switched to panic() it certainly can have my ack.

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

author
#3 lore raw

Re: [PATCH 1/2] x86: make sure module array is large enough in pvh-boot.c

Quoted 10 lines · Jan Beulich
Quoted 6 lines
Quoted 1 line
Quoted 1 line · On 28.09.18 at 10:24, <[email protected]> wrote:

The relocation code in __start_xen requires one extra element in the module array. By the looks of it the temporary array is already large enough. Add a BUG_ON to catch any issue in the future.

While at it, turn another ASSERT to BUG_ON as well.

Hmm, a little strange - I thought we had agreed on panic() before. The extra output BUG_ON() produces is unlikely to be helpful here. With this switched to panic() it certainly can have my ack.

Oops, I have no idea why I agreed to use panic but wrote BUG_ON instead. My memory has failed me.

I will resend this patch.

Wei.

Quoted 4 lines

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

author
#4 lore raw

PATCH 2/2 xen: make opt_xen_console true before console initialisation

In reply to Wei Liu

This helps capture issues before console is initialised. The impact is minimal because it only affects Xen running in as a guest.

---
 xen/drivers/char/console.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
xen/drivers/char/console.c +6 -1
index e48039dd82..bbec98304e 100644
@@ -91,7 +91,8 @@ static uint32_t conringc, conringp;
 static int __read_mostly sercon_handle = -1;
 
 #ifdef CONFIG_X86
-static bool __read_mostly opt_console_xen; /* console=xen */
+/* Set to true at start of day to catch early boot issues */
+static bool __read_mostly opt_console_xen = true; /* console=xen */
 #endif
 
 static DEFINE_SPINLOCK(console_lock);
@@ -821,6 +822,10 @@ void __init console_init_preirq(void)
 
     serial_init_preirq();
 
+#ifdef CONFIG_X86
+    opt_console_xen = false;
+#endif
+
     /* Where should console output go? */
     for ( p = opt_console; p != NULL; p = strchr(p, ',') )
     {
Signature · 7 lines
2.11.0


_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
#5 lore raw

Re: [PATCH 2/2] xen: make opt_xen_console true before console initialisation

xen/drivers/char/console.c static uint32_t conringc, conringp; · line 95

Quoted 10 lines
Quoted 1 line
Quoted 1 line · On 28.09.18 at 10:24, <[email protected]> wrote:
xen/drivers/char/console.c +2 -1
@@ -91,7 +91,8 @@ static uint32_t conringc, conringp;
 static int __read_mostly sercon_handle = -1;
 
 #ifdef CONFIG_X86
-static bool __read_mostly opt_console_xen; /* console=xen */
+/* Set to true at start of day to catch early boot issues */
+static bool __read_mostly opt_console_xen = true; /* console=xen */

When Andrew suggested this, iirc he said to make the variable tristate. Otherwise ...

xen/drivers/char/console.c void __init console_init_preirq(void) · line 827

Quoted 7 lines
xen/drivers/char/console.c +3 -0
@@ -821,6 +822,10 @@ void __init console_init_preirq(void)
 
     serial_init_preirq();
 
+#ifdef CONFIG_X86
+    opt_console_xen = false;
+#endif

... you possibly override a user specified "true" here.

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

author
#6 lore raw

Re: [PATCH 2/2] xen: make opt_xen_console true before console initialisation

Quoted 6 lines · Jan Beulich
Quoted 6 lines
Quoted 1 line
Quoted 1 line · On 28.09.18 at 10:24, <[email protected]> wrote:
xen/drivers/char/console.c +0 -0
@@ -91,7 +91,8 @@ static uint32_t conringc, conringp;
 static int __read_mostly sercon_handle = -1;
 
Quoted 7 lines

xen/drivers/char/console.c static uint32_t conringc, conringp; · line 95

Quoted 4 lines
xen/drivers/char/console.c +2 -1
 #ifdef CONFIG_X86
-static bool __read_mostly opt_console_xen; /* console=xen */
+/* Set to true at start of day to catch early boot issues */
+static bool __read_mostly opt_console_xen = true; /* console=xen */

When Andrew suggested this, iirc he said to make the variable tristate. Otherwise ...

I figure it doesn't need to be tristate.

Quoted 10 lines

xen/drivers/char/console.c void __init console_init_preirq(void) · line 827

Quoted 8 lines
xen/drivers/char/console.c +3 -0
@@ -821,6 +822,10 @@ void __init console_init_preirq(void)
 
     serial_init_preirq();
 
+#ifdef CONFIG_X86
+    opt_console_xen = false;
+#endif

... you possibly override a user specified "true" here.

Do I? This is right before option parsing, during which opt_console_xen is set.

Wei.

Quoted 4 lines

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#7 lore raw

Re: [PATCH 2/2] xen: make opt_xen_console true before console initialisation

Quoted 17 lines · Wei Liu
Quoted 6 lines · Jan Beulich
Quoted 6 lines
Quoted 1 line
Quoted 1 line · On 28.09.18 at 10:24, <[email protected]> wrote:
xen/drivers/char/console.c +0 -0
@@ -91,7 +91,8 @@ static uint32_t conringc, conringp;
 static int __read_mostly sercon_handle = -1;
 
Quoted 6 lines

xen/drivers/char/console.c static uint32_t conringc, conringp; · line 95

Quoted 4 lines
xen/drivers/char/console.c +2 -1
 #ifdef CONFIG_X86
-static bool __read_mostly opt_console_xen; /* console=xen */
+/* Set to true at start of day to catch early boot issues */
+static bool __read_mostly opt_console_xen = true; /* console=xen */

When Andrew suggested this, iirc he said to make the variable tristate. Otherwise ...

I figure it doesn't need to be tristate.

Quoted 2 lines
Quoted 2 lines
xen/drivers/char/console.c +0 -0
@@ -821,6 +822,10 @@ void __init console_init_preirq(void)
 
Quoted 8 lines
Quoted 6 lines

xen/drivers/char/console.c void __init console_init_preirq(void) · line 827

Quoted 5 lines
xen/drivers/char/console.c +3 -0
     serial_init_preirq();
 
+#ifdef CONFIG_X86
+    opt_console_xen = false;
+#endif

... you possibly override a user specified "true" here.

Do I? This is right before option parsing, during which opt_console_xen is set.

This is fine in principle, but you can't initialise it to true at compile time, or native boots will try using hypercalls / port 0xe9.

The best bet is to set it to true in the PVH entry asm.

~Andrew

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

author
#8 lore raw

Re: [PATCH 2/2] xen: make opt_xen_console true before console initialisation

Quoted 24 lines · Andrew Cooper
Quoted 23 lines · Wei Liu
Quoted 6 lines · Jan Beulich
Quoted 6 lines
Quoted 1 line
Quoted 1 line · On 28.09.18 at 10:24, <[email protected]> wrote:
xen/drivers/char/console.c +0 -0
@@ -91,7 +91,8 @@ static uint32_t conringc, conringp;
 static int __read_mostly sercon_handle = -1;
 
Quoted 6 lines

xen/drivers/char/console.c static uint32_t conringc, conringp; · line 95

Quoted 4 lines
xen/drivers/char/console.c +2 -1
 #ifdef CONFIG_X86
-static bool __read_mostly opt_console_xen; /* console=xen */
+/* Set to true at start of day to catch early boot issues */
+static bool __read_mostly opt_console_xen = true; /* console=xen */

When Andrew suggested this, iirc he said to make the variable tristate. Otherwise ...

I figure it doesn't need to be tristate.

Quoted 8 lines

xen/drivers/char/console.c void __init console_init_preirq(void) · line 827

Quoted 7 lines
xen/drivers/char/console.c +3 -0
@@ -821,6 +822,10 @@ void __init console_init_preirq(void)
 
     serial_init_preirq();
 
+#ifdef CONFIG_X86
+    opt_console_xen = false;
+#endif

... you possibly override a user specified "true" here.

Quoted 7 lines
Quoted 2 lines

Do I? This is right before option parsing, during which opt_console_xen is set.

This is fine in principle, but you can't initialise it to true at compile time, or native boots will try using hypercalls / port 0xe9.

The best bet is to set it to true in the PVH entry asm.

Right. I will set it in the PVH entry instead.

Wei.

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#9 lore raw

Re: [PATCH 2/2] xen: make opt_xen_console true before console initialisation

In reply to Wei Liu
Quoted 23 lines
Quoted 1 line
Quoted 1 line · On 28.09.18 at 17:12, <[email protected]> wrote:
Quoted 6 lines · Jan Beulich
Quoted 6 lines
Quoted 1 line
Quoted 1 line · On 28.09.18 at 10:24, <[email protected]> wrote:
xen/drivers/char/console.c +0 -0
@@ -91,7 +91,8 @@ static uint32_t conringc, conringp;
 static int __read_mostly sercon_handle = -1;
 
Quoted 7 lines

xen/drivers/char/console.c static uint32_t conringc, conringp; · line 95

Quoted 4 lines
xen/drivers/char/console.c +2 -1
 #ifdef CONFIG_X86
-static bool __read_mostly opt_console_xen; /* console=xen */
+/* Set to true at start of day to catch early boot issues */
+static bool __read_mostly opt_console_xen = true; /* console=xen */

When Andrew suggested this, iirc he said to make the variable tristate. Otherwise ...

I figure it doesn't need to be tristate.

Quoted 5 lines
Quoted 5 lines
xen/drivers/char/console.c +0 -0
@@ -821,6 +822,10 @@ void __init console_init_preirq(void)
 
     serial_init_preirq();
 
Quoted 8 lines
Quoted 5 lines

xen/drivers/char/console.c void __init console_init_preirq(void) · line 827

Quoted 3 lines
xen/drivers/char/console.c +3 -0
+#ifdef CONFIG_X86
+    opt_console_xen = false;
+#endif

... you possibly override a user specified "true" here.

Do I? This is right before option parsing, during which opt_console_xen is set.

Hmm, I see - I didn't recall we still have this ad hoc parsing in place, instead of using custom_param(). But isn't this too early then, as messages issued from the parsing code then would not be visible?

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

author
#10 lore raw

Re: [PATCH 2/2] xen: make opt_xen_console true before console initialisation

Quoted 29 lines · Jan Beulich
Quoted 29 lines
Quoted 1 line
Quoted 1 line · On 28.09.18 at 17:12, <[email protected]> wrote:
Quoted 6 lines · Jan Beulich
Quoted 6 lines
Quoted 1 line
Quoted 1 line · On 28.09.18 at 10:24, <[email protected]> wrote:
xen/drivers/char/console.c +0 -0
@@ -91,7 +91,8 @@ static uint32_t conringc, conringp;
 static int __read_mostly sercon_handle = -1;
 
Quoted 7 lines

xen/drivers/char/console.c static uint32_t conringc, conringp; · line 95

Quoted 4 lines
xen/drivers/char/console.c +2 -1
 #ifdef CONFIG_X86
-static bool __read_mostly opt_console_xen; /* console=xen */
+/* Set to true at start of day to catch early boot issues */
+static bool __read_mostly opt_console_xen = true; /* console=xen */

When Andrew suggested this, iirc he said to make the variable tristate. Otherwise ...

I figure it doesn't need to be tristate.

Quoted 10 lines

xen/drivers/char/console.c void __init console_init_preirq(void) · line 827

Quoted 8 lines
xen/drivers/char/console.c +3 -0
@@ -821,6 +822,10 @@ void __init console_init_preirq(void)
 
     serial_init_preirq();
 
+#ifdef CONFIG_X86
+    opt_console_xen = false;
+#endif

... you possibly override a user specified "true" here.

Quoted 6 lines
Quoted 2 lines

Do I? This is right before option parsing, during which opt_console_xen is set.

Hmm, I see - I didn't recall we still have this ad hoc parsing in place, instead of using custom_param(). But isn't this too early then, as messages issued from the parsing code then would not be visible?

To me it doesn't make anything worse than before. But I see your point. I will see if there is an easy way to fix this.

Wei.

Quoted 4 lines

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

author
#11 lore raw

Re: [PATCH 2/2] xen: make opt_xen_console true before console initialisation

Quoted 33 lines · Wei Liu
Quoted 32 lines · Jan Beulich
Quoted 31 lines
Quoted 1 line
Quoted 1 line · On 28.09.18 at 17:12, <[email protected]> wrote:
Quoted 6 lines · Jan Beulich
Quoted 6 lines
Quoted 1 line
Quoted 1 line · On 28.09.18 at 10:24, <[email protected]> wrote:
xen/drivers/char/console.c +0 -0
@@ -91,7 +91,8 @@ static uint32_t conringc, conringp;
 static int __read_mostly sercon_handle = -1;
 
Quoted 7 lines

xen/drivers/char/console.c static uint32_t conringc, conringp; · line 95

Quoted 4 lines
xen/drivers/char/console.c +2 -1
 #ifdef CONFIG_X86
-static bool __read_mostly opt_console_xen; /* console=xen */
+/* Set to true at start of day to catch early boot issues */
+static bool __read_mostly opt_console_xen = true; /* console=xen */

When Andrew suggested this, iirc he said to make the variable tristate. Otherwise ...

I figure it doesn't need to be tristate.

Quoted 10 lines

xen/drivers/char/console.c void __init console_init_preirq(void) · line 827

Quoted 8 lines
xen/drivers/char/console.c +3 -0
@@ -821,6 +822,10 @@ void __init console_init_preirq(void)
 
     serial_init_preirq();
 
+#ifdef CONFIG_X86
+    opt_console_xen = false;
+#endif

... you possibly override a user specified "true" here.

Do I? This is right before option parsing, during which opt_console_xen is set.

Quoted 6 lines
Quoted 3 lines

Hmm, I see - I didn't recall we still have this ad hoc parsing in place, instead of using custom_param(). But isn't this too early then, as messages issued from the parsing code then would not be visible?

To me it doesn't make anything worse than before. But I see your point. I will see if there is an easy way to fix this.

At the moment the ordering of the options matter, which means if "xen" is not the first option we can lose output.

I think one easy way of improving is to parse opt_console twice. The first time solely looks for "xen" and the second time parses the rest options.

Wei.

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#12 lore raw

Re: [PATCH 2/2] xen: make opt_xen_console true before console initialisation

Quoted 7 lines
Quoted 1 line
Quoted 1 line · On 28.09.18 at 17:53, <[email protected]> wrote:

At the moment the ordering of the options matter, which means if "xen" is not the first option we can lose output.

I think one easy way of improving is to parse opt_console twice. The first time solely looks for "xen" and the second time parses the rest options.

What's wrong with making the variable a tristate, and move your set-to-false after the parsing loop? If the user said "false" for the option, then they apparently don't care much about the output.

But yes, what you suggest is an option as well.

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

author
#13 lore raw

Re: [PATCH 2/2] xen: make opt_xen_console true before console initialisation

Quoted 11 lines · Jan Beulich
Quoted 7 lines
Quoted 1 line
Quoted 1 line · On 28.09.18 at 17:53, <[email protected]> wrote:

At the moment the ordering of the options matter, which means if "xen" is not the first option we can lose output.

I think one easy way of improving is to parse opt_console twice. The first time solely looks for "xen" and the second time parses the rest options.

What's wrong with making the variable a tristate, and move your set-to-false after the parsing loop? If the user said "false" for the option, then they apparently don't care much about the output.

Oh, there is nothing wrong with that. It just I didn't think about that solution while thinking about this problem.

Quoted 2 lines

But yes, what you suggest is an option as well.

Yeah, I already got code for it.

Wei.

Quoted 4 lines

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

author
#14 lore raw

Re: [PATCH 2/2] xen: make opt_xen_console true before console initialisation

Quoted 13 lines · Wei Liu
Quoted 11 lines · Jan Beulich
Quoted 7 lines
Quoted 1 line
Quoted 1 line · On 28.09.18 at 17:53, <[email protected]> wrote:

At the moment the ordering of the options matter, which means if "xen" is not the first option we can lose output.

I think one easy way of improving is to parse opt_console twice. The first time solely looks for "xen" and the second time parses the rest options.

What's wrong with making the variable a tristate, and move your set-to-false after the parsing loop? If the user said "false" for the option, then they apparently don't care much about the output.

Quoted 7 lines

Oh, there is nothing wrong with that. It just I didn't think about that solution while thinking about this problem.

Quoted 2 lines

But yes, what you suggest is an option as well.

Yeah, I already got code for it.

After fixing the bugs in my code it has become too long for my liking so I will turn to use tristate instead. :p

Wei.

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel