Skip to contentSkip to conversation

xen/domain: Idle domain creation improvements

PATCH In review Andrew Cooper 3 participants 13 messages Jul 18, 2024 – Jul 31, 2024

lore
Share & tools
lore
#0 lore raw

cover xen/domain: Idle domain creation improvements

domain_create() has some unnecessary complexity for constructing the idle domain. Rearrange x86's needs, to simplify the common logic.

https://cirrus-ci.com/build/5234584097390592
https://gitlab.com/xen-project/people/andyhhp/xen/-/pipelines/1379531597

Andrew Cooper (4):
xen/domain: Introduce arch_init_idle_domain()
x86/domain: Implement arch_init_idle_domain()
xen/domain: Simpliy domain_create() now the idle domain is complete
earlier
arch/domain: Clean up the idle domain remnants in arch_domain_create()

 xen/arch/arm/domain.c             |   6 --
 xen/arch/x86/domain.c             |  34 +++-----
 xen/arch/x86/include/asm/domain.h |   3 +
 xen/common/domain.c               | 130 +++++++++++++++---------------
 xen/include/xen/sched.h           |   5 ++
 5 files changed, 82 insertions(+), 96 deletions(-)

base-commit: 57338346f29cea7b183403561bdc5f407163b846

Signature · 1 line
2.39.2
#1 lore raw

PATCH 1/4 xen/domain: Introduce arch_init_idle_domain()

The idle domain causes a large amount of complexity in domain_create() because of x86's need to initialise d->arch.ctxt_switch in arch_domain_create().

In order to address this, introduce an optional hook to perform extra initialisation of the idle domain.

No functional change.

---
CC: Jan Beulich <[email protected]>
CC: Roger Pau Monné <[email protected]>
CC: Stefano Stabellini <[email protected]>
CC: Julien Grall <[email protected]>
CC: Volodymyr Babchuk <[email protected]>
CC: Bertrand Marquis <[email protected]>
CC: Michal Orzel <[email protected]>
CC: Oleksii Kurochko <[email protected]>

---
 xen/common/domain.c     | 3 +++
 xen/include/xen/sched.h | 5 +++++
 2 files changed, 8 insertions(+)
xen/common/domain.c +3 -0
index fb262dcec919..e312c820d641 100644
@@ -685,6 +685,9 @@ struct domain *domain_create(domid_t domid,
 
     rangeset_domain_initialise(d);
 
+    if ( is_idle_domain(d) )
+        arch_init_idle_domain(d);
+
     /* DOMID_{XEN,IO,etc} (other than IDLE) are sufficiently constructed. */
     if ( is_system_domain(d) && !is_idle_domain(d) )
         return d;
xen/include/xen/sched.h +5 -0
index 2dcd1d1a4f8a..90666576c2f8 100644
@@ -728,6 +728,11 @@ struct domain *domain_create(domid_t domid,
                              struct xen_domctl_createdomain *config,
                              unsigned int flags);
 
+#ifndef arch_init_idle_domain
+/* Optional, if there's any construction necessary for DOMID_IDLE */
+static inline void arch_init_idle_domain(struct domain *d) {}
+#endif
+
 /*
  * rcu_lock_domain_by_id() is more efficient than get_domain_by_id().
  * This is the preferred function if the returned domain reference
Signature · 1 line
2.39.2
Reviewed-by on v1 1/4
#2 lore raw

Re: [PATCH 1/4] xen/domain: Introduce arch_init_idle_domain()

Quoted 9 lines · Andrew Cooper

The idle domain causes a large amount of complexity in domain_create() because of x86's need to initialise d->arch.ctxt_switch in arch_domain_create().

In order to address this, introduce an optional hook to perform extra initialisation of the idle domain.

No functional change.

Quoted 47 lines

---
CC: Jan Beulich <[email protected]>
CC: Roger Pau Monné <[email protected]>
CC: Stefano Stabellini <[email protected]>
CC: Julien Grall <[email protected]>
CC: Volodymyr Babchuk <[email protected]>
CC: Bertrand Marquis <[email protected]>
CC: Michal Orzel <[email protected]>
CC: Oleksii Kurochko <[email protected]>

---
 xen/common/domain.c     | 3 +++
 xen/include/xen/sched.h | 5 +++++
 2 files changed, 8 insertions(+)
xen/common/domain.c +3 -0
index fb262dcec919..e312c820d641 100644
@@ -685,6 +685,9 @@ struct domain *domain_create(domid_t domid,
 
     rangeset_domain_initialise(d);
 
+    if ( is_idle_domain(d) )
+        arch_init_idle_domain(d);
+
     /* DOMID_{XEN,IO,etc} (other than IDLE) are sufficiently constructed. */
     if ( is_system_domain(d) && !is_idle_domain(d) )
         return d;
xen/include/xen/sched.h +5 -0
index 2dcd1d1a4f8a..90666576c2f8 100644
@@ -728,6 +728,11 @@ struct domain *domain_create(domid_t domid,
                              struct xen_domctl_createdomain *config,
                              unsigned int flags);
 
+#ifndef arch_init_idle_domain
+/* Optional, if there's any construction necessary for DOMID_IDLE */
+static inline void arch_init_idle_domain(struct domain *d) {}
+#endif
+
 /*
  * rcu_lock_domain_by_id() is more efficient than get_domain_by_id().
  * This is the preferred function if the returned domain reference
Signature · 1 line
2.39.2
#3 lore raw

PATCH 2/4 x86/domain: Implement arch_init_idle_domain()

In reply to Andrew Cooper

The idle domain needs d->arch.ctxt_switch initialised on x86. Implement the new arch_init_idle_domain() in order to do this.

Right now this double-initalises the ctxt_switch pointer, but it's safe and will stop happening when domain_create() is cleaned up as a consequence.

No practical change.

---
CC: Jan Beulich <[email protected]>
CC: Roger Pau Monné <[email protected]>
CC: Stefano Stabellini <[email protected]>
CC: Julien Grall <[email protected]>
CC: Volodymyr Babchuk <[email protected]>
CC: Bertrand Marquis <[email protected]>
CC: Michal Orzel <[email protected]>
CC: Oleksii Kurochko <[email protected]>

---
 xen/arch/x86/domain.c             | 19 ++++++++++++-------
 xen/arch/x86/include/asm/domain.h |  3 +++
 2 files changed, 15 insertions(+), 7 deletions(-)
xen/arch/x86/domain.c +12 -7
index ccadfe0c9e70..eff905c6c6e5 100644
@@ -768,6 +768,17 @@ static bool emulation_flags_ok(const struct domain *d, uint32_t emflags)
     return true;
 }
 
+void __init arch_init_idle_domain(struct domain *d)
+{
+    static const struct arch_csw idle_csw = {
+        .from = paravirt_ctxt_switch_from,
+        .to   = paravirt_ctxt_switch_to,
+        .tail = idle_loop,
+    };
+
+    d->arch.ctxt_switch = &idle_csw;
+}
+
 int arch_domain_create(struct domain *d,
                        struct xen_domctl_createdomain *config,
                        unsigned int flags)
@@ -783,13 +794,7 @@ int arch_domain_create(struct domain *d,
     /* Minimal initialisation for the idle domain. */
     if ( unlikely(is_idle_domain(d)) )
     {
-        static const struct arch_csw idle_csw = {
-            .from = paravirt_ctxt_switch_from,
-            .to   = paravirt_ctxt_switch_to,
-            .tail = idle_loop,
-        };
-
-        d->arch.ctxt_switch = &idle_csw;
+        arch_init_idle_domain(d);
 
         d->arch.cpu_policy = ZERO_BLOCK_PTR; /* Catch stray misuses. */
 
xen/arch/x86/include/asm/domain.h +3 -0
index f5daeb182baa..bca3258d69ac 100644
@@ -779,6 +779,9 @@ struct arch_vcpu_io {
 /* Maxphysaddr supportable by the paging infrastructure. */
 unsigned int domain_max_paddr_bits(const struct domain *d);
 
+#define arch_init_idle_domain arch_init_idle_domain
+void arch_init_idle_domain(struct domain *d);
+
 #endif /* __ASM_DOMAIN_H__ */
 
 /*
Signature · 1 line
2.39.2
#4 lore raw

Re: [PATCH 2/4] x86/domain: Implement arch_init_idle_domain()

Quoted 47 lines · Andrew Cooper

The idle domain needs d->arch.ctxt_switch initialised on x86. Implement the new arch_init_idle_domain() in order to do this.

Right now this double-initalises the ctxt_switch pointer, but it's safe and will stop happening when domain_create() is cleaned up as a consequence.

No practical change.

---
CC: Jan Beulich <[email protected]>
CC: Roger Pau Monné <[email protected]>
CC: Stefano Stabellini <[email protected]>
CC: Julien Grall <[email protected]>
CC: Volodymyr Babchuk <[email protected]>
CC: Bertrand Marquis <[email protected]>
CC: Michal Orzel <[email protected]>
CC: Oleksii Kurochko <[email protected]>

---
 xen/arch/x86/domain.c             | 19 ++++++++++++-------
 xen/arch/x86/include/asm/domain.h |  3 +++
 2 files changed, 15 insertions(+), 7 deletions(-)
xen/arch/x86/domain.c +11 -0
index ccadfe0c9e70..eff905c6c6e5 100644
@@ -768,6 +768,17 @@ static bool emulation_flags_ok(const struct domain *d, uint32_t emflags)
     return true;
 }
 
+void __init arch_init_idle_domain(struct domain *d)
+{
+    static const struct arch_csw idle_csw = {
+        .from = paravirt_ctxt_switch_from,
+        .to   = paravirt_ctxt_switch_to,
+        .tail = idle_loop,
+    };
+
+    d->arch.ctxt_switch = &idle_csw;
+}
+
 int arch_domain_create(struct domain *d,
                        struct xen_domctl_createdomain *config,
                        unsigned int flags)
@@ -783,13 +794,7 @@ int arch_domain_create(struct domain *d,
     /* Minimal initialisation for the idle domain. */

xen/arch/x86/domain.c int arch_domain_create(struct domain *d, · line 797

Quoted 10 lines
xen/arch/x86/domain.c +1 -7
     if ( unlikely(is_idle_domain(d)) )
     {
-        static const struct arch_csw idle_csw = {
-            .from = paravirt_ctxt_switch_from,
-            .to   = paravirt_ctxt_switch_to,
-            .tail = idle_loop,
-        };
-
-        d->arch.ctxt_switch = &idle_csw;
+        arch_init_idle_domain(d);

I don't understand why you need this call to arch_init_idle_domain(d) here given the other call to arch_init_idle_domain added by patch #1.

Also I am not sure why you didn't move the line below (cpu_policy = ZERO_BLOCK_PTR) to arch_init_idle_domain as well but I admit I don't know this code

I realize you are removing all this code in patch #4... but still why the line below is not needed anymore? And why do we need to add an extra call to arch_init_idle_domain temporarily in this patch?

Quoted 19 lines
xen/arch/x86/domain.c +0 -0
         d->arch.cpu_policy = ZERO_BLOCK_PTR; /* Catch stray misuses. */
 
xen/arch/x86/include/asm/domain.h +3 -0
index f5daeb182baa..bca3258d69ac 100644
@@ -779,6 +779,9 @@ struct arch_vcpu_io {
 /* Maxphysaddr supportable by the paging infrastructure. */
 unsigned int domain_max_paddr_bits(const struct domain *d);
 
+#define arch_init_idle_domain arch_init_idle_domain
+void arch_init_idle_domain(struct domain *d);
+
 #endif /* __ASM_DOMAIN_H__ */
 
 /*
Signature · 1 line
2.39.2
#5 lore raw

Re: [PATCH 2/4] x86/domain: Implement arch_init_idle_domain()

Quoted 52 lines · Stefano Stabellini
Quoted 51 lines · Andrew Cooper

The idle domain needs d->arch.ctxt_switch initialised on x86. Implement the new arch_init_idle_domain() in order to do this.

Right now this double-initalises the ctxt_switch pointer, but it's safe and will stop happening when domain_create() is cleaned up as a consequence.

No practical change.

---
CC: Jan Beulich <[email protected]>
CC: Roger Pau Monné <[email protected]>
CC: Stefano Stabellini <[email protected]>
CC: Julien Grall <[email protected]>
CC: Volodymyr Babchuk <[email protected]>
CC: Bertrand Marquis <[email protected]>
CC: Michal Orzel <[email protected]>
CC: Oleksii Kurochko <[email protected]>

---
 xen/arch/x86/domain.c             | 19 ++++++++++++-------
 xen/arch/x86/include/asm/domain.h |  3 +++
 2 files changed, 15 insertions(+), 7 deletions(-)
xen/arch/x86/domain.c +11 -2
index ccadfe0c9e70..eff905c6c6e5 100644
@@ -768,6 +768,17 @@ static bool emulation_flags_ok(const struct domain *d, uint32_t emflags)
     return true;
 }
 
+void __init arch_init_idle_domain(struct domain *d)
+{
+    static const struct arch_csw idle_csw = {
+        .from = paravirt_ctxt_switch_from,
+        .to   = paravirt_ctxt_switch_to,
+        .tail = idle_loop,
+    };
+
+    d->arch.ctxt_switch = &idle_csw;
+}
+
 int arch_domain_create(struct domain *d,
                        struct xen_domctl_createdomain *config,
                        unsigned int flags)
@@ -783,13 +794,7 @@ int arch_domain_create(struct domain *d,
     /* Minimal initialisation for the idle domain. */
     if ( unlikely(is_idle_domain(d)) )
     {
-        static const struct arch_csw idle_csw = {
-            .from = paravirt_ctxt_switch_from,
Quoted 8 lines

xen/arch/x86/domain.c int arch_domain_create(struct domain *d, · line 797

Quoted 6 lines
xen/arch/x86/domain.c +1 -5
-            .to   = paravirt_ctxt_switch_to,
-            .tail = idle_loop,
-        };
-
-        d->arch.ctxt_switch = &idle_csw;
+        arch_init_idle_domain(d);

I don't understand why you need this call to arch_init_idle_domain(d) here given the other call to arch_init_idle_domain added by patch #1.

It's stale from an older version of the series.  I'll drop it, which will reduce the churn in this as well as patch 4.

Quoted 3 lines

Also I am not sure why you didn't move the line below (cpu_policy = ZERO_BLOCK_PTR) to arch_init_idle_domain as well but I admit I don't know this code.

See the thread on patch 4 with Jan.  It's intentional.

This was a safety check which is dubious (it's not special enough to warrant it on its own), and has never tripped.

I'll note it's intentional omission in the commit message.

~Andrew

#6 lore raw

PATCH 3/4 xen/domain: Simpliy domain_create() now the idle domain is complete earlier

In reply to Andrew Cooper

With x86 implementing arch_init_idle_domain(), there is no longer any need to call arch_domain_create() with the idle domain.

Have the idle domain exit early with all other system domains. Move the static-analysis ASSERT() earlier. Then, remove the !is_idle_domain() protections around the majority of domain_create() and remove one level of indentation.

No practical change.

---
CC: Jan Beulich <[email protected]>
CC: Roger Pau Monné <[email protected]>
CC: Stefano Stabellini <[email protected]>
CC: Julien Grall <[email protected]>
CC: Volodymyr Babchuk <[email protected]>
CC: Bertrand Marquis <[email protected]>
CC: Michal Orzel <[email protected]>
CC: Oleksii Kurochko <[email protected]>

---
 xen/common/domain.c | 127 +++++++++++++++++++++-----------------------
 1 file changed, 60 insertions(+), 67 deletions(-)
xen/common/domain.c +60 -67
index e312c820d641..8bc0bcd06ca5 100644
@@ -688,95 +688,88 @@ struct domain *domain_create(domid_t domid,
     if ( is_idle_domain(d) )
         arch_init_idle_domain(d);
 
-    /* DOMID_{XEN,IO,etc} (other than IDLE) are sufficiently constructed. */
-    if ( is_system_domain(d) && !is_idle_domain(d) )
+    /* DOMID_{XEN,IO,IDLE,etc} are sufficiently constructed. */
+    if ( is_system_domain(d) )
         return d;
 
+    /*
+     * The assertion helps static analysis tools infer that config cannot be
+     * NULL in this branch, which in turn means that it can be safely
+     * dereferenced. Therefore, this assertion is not redundant.
+     */
+    ASSERT(config);
+
 #ifdef CONFIG_HAS_PIRQ
-    if ( !is_idle_domain(d) )
-    {
-        if ( !is_hardware_domain(d) )
-            d->nr_pirqs = nr_static_irqs + extra_domU_irqs;
-        else
-            d->nr_pirqs = extra_hwdom_irqs ? nr_static_irqs + extra_hwdom_irqs
-                                           : arch_hwdom_irqs(d);
-        d->nr_pirqs = min(d->nr_pirqs, nr_irqs);
+    if ( !is_hardware_domain(d) )
+        d->nr_pirqs = nr_static_irqs + extra_domU_irqs;
+    else
+        d->nr_pirqs = extra_hwdom_irqs ? nr_static_irqs + extra_hwdom_irqs
+                                       : arch_hwdom_irqs(d);
+    d->nr_pirqs = min(d->nr_pirqs, nr_irqs);
 
-        radix_tree_init(&d->pirq_tree);
-    }
+    radix_tree_init(&d->pirq_tree);
 #endif
 
     if ( (err = arch_domain_create(d, config, flags)) != 0 )
         goto fail;
     init_status |= INIT_arch;
 
-    if ( !is_idle_domain(d) )
-    {
-        /*
-         * The assertion helps static analysis tools infer that config cannot
-         * be NULL in this branch, which in turn means that it can be safely
-         * dereferenced. Therefore, this assertion is not redundant.
-         */
-        ASSERT(config);
-
-        watchdog_domain_init(d);
-        init_status |= INIT_watchdog;
+    watchdog_domain_init(d);
+    init_status |= INIT_watchdog;
 
-        err = -ENOMEM;
-        d->iomem_caps = rangeset_new(d, "I/O Memory", RANGESETF_prettyprint_hex);
-        d->irq_caps   = rangeset_new(d, "Interrupts", 0);
-        if ( !d->iomem_caps || !d->irq_caps )
-            goto fail;
+    err = -ENOMEM;
+    d->iomem_caps = rangeset_new(d, "I/O Memory", RANGESETF_prettyprint_hex);
+    d->irq_caps   = rangeset_new(d, "Interrupts", 0);
+    if ( !d->iomem_caps || !d->irq_caps )
+        goto fail;
 
-        if ( (err = xsm_domain_create(XSM_HOOK, d, config->ssidref)) != 0 )
-            goto fail;
+    if ( (err = xsm_domain_create(XSM_HOOK, d, config->ssidref)) != 0 )
+        goto fail;
 
-        d->controller_pause_count = 1;
-        atomic_inc(&d->pause_count);
+    d->controller_pause_count = 1;
+    atomic_inc(&d->pause_count);
 
-        if ( (err = evtchn_init(d, config->max_evtchn_port)) != 0 )
-            goto fail;
-        init_status |= INIT_evtchn;
+    if ( (err = evtchn_init(d, config->max_evtchn_port)) != 0 )
+        goto fail;
+    init_status |= INIT_evtchn;
 
-        if ( (err = grant_table_init(d, config->max_grant_frames,
-                                     config->max_maptrack_frames,
-                                     config->grant_opts)) != 0 )
-            goto fail;
-        init_status |= INIT_gnttab;
+    if ( (err = grant_table_init(d, config->max_grant_frames,
+                                 config->max_maptrack_frames,
+                                 config->grant_opts)) != 0 )
+        goto fail;
+    init_status |= INIT_gnttab;
 
-        if ( (err = argo_init(d)) != 0 )
-            goto fail;
+    if ( (err = argo_init(d)) != 0 )
+        goto fail;
 
-        err = -ENOMEM;
+    err = -ENOMEM;
+    d->pbuf = xzalloc_array(char, DOMAIN_PBUF_SIZE);
+    if ( !d->pbuf )
+        goto fail;
 
-        d->pbuf = xzalloc_array(char, DOMAIN_PBUF_SIZE);
-        if ( !d->pbuf )
-            goto fail;
+    if ( (err = sched_init_domain(d, config->cpupool_id)) != 0 )
+        goto fail;
 
-        if ( (err = sched_init_domain(d, config->cpupool_id)) != 0 )
-            goto fail;
+    if ( (err = late_hwdom_init(d)) != 0 )
+        goto fail;
 
-        if ( (err = late_hwdom_init(d)) != 0 )
-            goto fail;
+    /*
+     * Must not fail beyond this point, as our caller doesn't know whether
+     * the domain has been entered into domain_list or not.
+     */
 
-        /*
-         * Must not fail beyond this point, as our caller doesn't know whether
-         * the domain has been entered into domain_list or not.
-         */
+    spin_lock(&domlist_update_lock);
+    pd = &domain_list; /* NB. domain_list maintained in order of domid. */
+    for ( pd = &domain_list; *pd != NULL; pd = &(*pd)->next_in_list )
+        if ( (*pd)->domain_id > d->domain_id )
+            break;
+    d->next_in_list = *pd;
+    d->next_in_hashbucket = domain_hash[DOMAIN_HASH(domid)];
+    rcu_assign_pointer(*pd, d);
+    rcu_assign_pointer(domain_hash[DOMAIN_HASH(domid)], d);
+    spin_unlock(&domlist_update_lock);
 
-        spin_lock(&domlist_update_lock);
-        pd = &domain_list; /* NB. domain_list maintained in order of domid. */
-        for ( pd = &domain_list; *pd != NULL; pd = &(*pd)->next_in_list )
-            if ( (*pd)->domain_id > d->domain_id )
-                break;
-        d->next_in_list = *pd;
-        d->next_in_hashbucket = domain_hash[DOMAIN_HASH(domid)];
-        rcu_assign_pointer(*pd, d);
-        rcu_assign_pointer(domain_hash[DOMAIN_HASH(domid)], d);
-        spin_unlock(&domlist_update_lock);
-
-        memcpy(d->handle, config->handle, sizeof(d->handle));
-    }
+    memcpy(d->handle, config->handle, sizeof(d->handle));
 
     return d;
 
Signature · 1 line
2.39.2
Reviewed-by on v1 3/4
#7 lore raw

Re: [PATCH 3/4] xen/domain: Simpliy domain_create() now the idle domain is complete earlier

Quoted 11 lines · Andrew Cooper

With x86 implementing arch_init_idle_domain(), there is no longer any need to call arch_domain_create() with the idle domain.

Have the idle domain exit early with all other system domains. Move the static-analysis ASSERT() earlier. Then, remove the !is_idle_domain() protections around the majority of domain_create() and remove one level of indentation.

No practical change.

#8 lore raw

PATCH 4/4 arch/domain: Clean up the idle domain remnants in arch_domain_create()

In reply to Andrew Cooper

With arch_domain_create() no longer being called with the idle domain, drop the last remaining logic.

No functional change.

---
CC: Jan Beulich <[email protected]>
CC: Roger Pau Monné <[email protected]>
CC: Stefano Stabellini <[email protected]>
CC: Julien Grall <[email protected]>
CC: Volodymyr Babchuk <[email protected]>
CC: Bertrand Marquis <[email protected]>
CC: Michal Orzel <[email protected]>
CC: Oleksii Kurochko <[email protected]>

---
 xen/arch/arm/domain.c |  6 ------
 xen/arch/x86/domain.c | 17 -----------------
 2 files changed, 23 deletions(-)
xen/arch/arm/domain.c +0 -6
index 7cfcefd27944..3ba959f86633 100644
@@ -706,12 +706,6 @@ int arch_domain_create(struct domain *d,
 
     BUILD_BUG_ON(GUEST_MAX_VCPUS < MAX_VIRT_CPUS);
 
-    /* Idle domains do not need this setup */
-    if ( is_idle_domain(d) )
-        return 0;
-
-    ASSERT(config != NULL);
-
 #ifdef CONFIG_IOREQ_SERVER
     ioreq_domain_init(d);
 #endif
xen/arch/x86/domain.c +0 -17
index eff905c6c6e5..c71b9023cb1a 100644
@@ -791,23 +791,6 @@ int arch_domain_create(struct domain *d,
 
     spin_lock_init(&d->arch.e820_lock);
 
-    /* Minimal initialisation for the idle domain. */
-    if ( unlikely(is_idle_domain(d)) )
-    {
-        arch_init_idle_domain(d);
-
-        d->arch.cpu_policy = ZERO_BLOCK_PTR; /* Catch stray misuses. */
-
-        return 0;
-    }
-
-    if ( !config )
-    {
-        /* Only IDLE is allowed with no config. */
-        ASSERT_UNREACHABLE();
-        return -EINVAL;
-    }
-
     if ( d->domain_id && cpu_has_amd_erratum(&boot_cpu_data, AMD_ERRATUM_121) )
     {
         if ( !opt_allow_unsafe )
Signature · 1 line
2.39.2
Reviewed-by on v1 4/4
#9 lore raw

Re: [PATCH 4/4] arch/domain: Clean up the idle domain remnants in arch_domain_create()

xen/arch/x86/domain.c int arch_domain_create(struct domain *d, · line 799

Quoted 12 lines · Andrew Cooper
xen/arch/x86/domain.c +0 -6
@@ -791,23 +791,6 @@ int arch_domain_create(struct domain *d,
 
     spin_lock_init(&d->arch.e820_lock);
 
-    /* Minimal initialisation for the idle domain. */
-    if ( unlikely(is_idle_domain(d)) )
-    {
-        arch_init_idle_domain(d);
-
-        d->arch.cpu_policy = ZERO_BLOCK_PTR; /* Catch stray misuses. */

This line looks to be lost in the process. Already in an earlier patch it should move to arch_init_idle_domain(), shouldn't it? With that adjustment for the entire series:

Jan

#10 lore raw

Re: [PATCH 4/4] arch/domain: Clean up the idle domain remnants in arch_domain_create()

Quoted 16 lines · Jan Beulich

xen/arch/x86/domain.c int arch_domain_create(struct domain *d, · line 799

Quoted 12 lines · Andrew Cooper
xen/arch/x86/domain.c +0 -6
@@ -791,23 +791,6 @@ int arch_domain_create(struct domain *d,
 
     spin_lock_init(&d->arch.e820_lock);
 
-    /* Minimal initialisation for the idle domain. */
-    if ( unlikely(is_idle_domain(d)) )
-    {
-        arch_init_idle_domain(d);
-
-        d->arch.cpu_policy = ZERO_BLOCK_PTR; /* Catch stray misuses. */

This line looks to be lost in the process. Already in an earlier patch it should move to arch_init_idle_domain(), shouldn't it? With that adjustment for the entire series:

Quoted 1 line

It was semi-intentional.  cpu_policy is just one of many pointers, and I don't see a good reason for it to be treated specially.

It's been around for years now, and never one triggered, not to mention the fact that there's a dwindling set of machines where a dereference of 0 isn't instantly fatal anyway.

~Andrew

#11 lore raw

Re: [PATCH 4/4] arch/domain: Clean up the idle domain remnants in arch_domain_create()

Quoted 19 lines · Andrew Cooper
Quoted 16 lines · Jan Beulich

xen/arch/x86/domain.c int arch_domain_create(struct domain *d, · line 799

Quoted 12 lines · Andrew Cooper
xen/arch/x86/domain.c +0 -6
@@ -791,23 +791,6 @@ int arch_domain_create(struct domain *d,
 
     spin_lock_init(&d->arch.e820_lock);
 
-    /* Minimal initialisation for the idle domain. */
-    if ( unlikely(is_idle_domain(d)) )
-    {
-        arch_init_idle_domain(d);
-
-        d->arch.cpu_policy = ZERO_BLOCK_PTR; /* Catch stray misuses. */

This line looks to be lost in the process. Already in an earlier patch it should move to arch_init_idle_domain(), shouldn't it? With that adjustment for the entire series:

Quoted 1 line
Quoted 6 lines

It was semi-intentional.  cpu_policy is just one of many pointers, and I don't see a good reason for it to be treated specially.

It's been around for years now, and never one triggered, not to mention the fact that there's a dwindling set of machines where a dereference of 0 isn't instantly fatal anyway.

Okay, yet would you mind mentioning this removal as intentional then in the description?

Jan

Reviewed-by on v1 4/4
#12 lore raw

Re: [PATCH 4/4] arch/domain: Clean up the idle domain remnants in arch_domain_create()

In reply to Andrew Cooper
Quoted 6 lines · Andrew Cooper

With arch_domain_create() no longer being called with the idle domain, drop the last remaining logic.

No functional change.

but one question below

Quoted 36 lines

---
CC: Jan Beulich <[email protected]>
CC: Roger Pau Monné <[email protected]>
CC: Stefano Stabellini <[email protected]>
CC: Julien Grall <[email protected]>
CC: Volodymyr Babchuk <[email protected]>
CC: Bertrand Marquis <[email protected]>
CC: Michal Orzel <[email protected]>
CC: Oleksii Kurochko <[email protected]>

---
 xen/arch/arm/domain.c |  6 ------
 xen/arch/x86/domain.c | 17 -----------------
 2 files changed, 23 deletions(-)
xen/arch/arm/domain.c +0 -6
index 7cfcefd27944..3ba959f86633 100644
@@ -706,12 +706,6 @@ int arch_domain_create(struct domain *d,
 
     BUILD_BUG_ON(GUEST_MAX_VCPUS < MAX_VIRT_CPUS);
 
-    /* Idle domains do not need this setup */
-    if ( is_idle_domain(d) )
-        return 0;
-
-    ASSERT(config != NULL);
-
 #ifdef CONFIG_IOREQ_SERVER
     ioreq_domain_init(d);
 #endif
xen/arch/x86/domain.c +0 -0
index eff905c6c6e5..c71b9023cb1a 100644

xen/arch/x86/domain.c int arch_domain_create(struct domain *d, · line 799

Quoted 9 lines
xen/arch/x86/domain.c +0 -6
@@ -791,23 +791,6 @@ int arch_domain_create(struct domain *d,
 
     spin_lock_init(&d->arch.e820_lock);
 
-    /* Minimal initialisation for the idle domain. */
-    if ( unlikely(is_idle_domain(d)) )
-    {
-        arch_init_idle_domain(d);
-
-        d->arch.cpu_policy = ZERO_BLOCK_PTR; /* Catch stray misuses. */

Are you sure you don't want to keep it?

Quoted 16 lines
xen/arch/x86/domain.c +0 -10
-        return 0;
-    }
-
-    if ( !config )
-    {
-        /* Only IDLE is allowed with no config. */
-        ASSERT_UNREACHABLE();
-        return -EINVAL;
-    }
-
     if ( d->domain_id && cpu_has_amd_erratum(&boot_cpu_data, AMD_ERRATUM_121) )
     {
         if ( !opt_allow_unsafe )
Signature · 1 line
2.39.2