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

[Xen-changelog] [xen staging] xen/arm32: Don't blindly unmask interrupts on trap without a change of level



commit 61b683571f0abd12395b1454cd055f2ad9bb3a37
Author:     Julien Grall <julien.grall@xxxxxxx>
AuthorDate: Fri Oct 11 17:49:28 2019 +0100
Commit:     Jan Beulich <jbeulich@xxxxxxxx>
CommitDate: Thu Oct 31 16:22:34 2019 +0100

    xen/arm32: Don't blindly unmask interrupts on trap without a change of level
    
    Exception vectors will unmask interrupts regardless the state of them in
    the interrupted context.
    
    One of the consequences is IRQ will be unmasked when receiving an
    undefined instruction exception (used by WARN*) from the hypervisor.
    This could result to unexpected behavior such as deadlock (if a lock was
    shared with interrupts).
    
    In a nutshell, interrupts should only be unmasked when it is safe to do.
    Xen only unmask IRQ and Abort interrupts, so the logic can stay simple.
    
    As vectors exceptions may be shared between guest and hypervisor, we now
    need to have a different policy for the interrupts.
    
    On exception from hypervisor, each vector will select the list of
    interrupts to inherit from the interrupted context. Any interrupts not
    listed will be kept masked.
    
    On exception from the guest, the Abort and IRQ will be unmasked
    depending on the exact vector.
    
    The interrupts will be kept unmasked when the vector cannot used by
    either guest or hypervisor.
    
    Note that each vector is not anymore preceded by ALIGN. This is fine
    because the alignment is already bigger than what we need.
    
    This is part of XSA-303.
    
    Reported-by: Julien Grall <Julien.Grall@xxxxxxx>
    Signed-off-by: Julien Grall <julien.grall@xxxxxxx>
    Reviewed-by: Stefano Stabellini <sstabellini@xxxxxxxxxx>
    Reviewed-by: Andre Przywara <andre.przywara@xxxxxxx>
---
 xen/arch/arm/arm32/entry.S | 138 +++++++++++++++++++++++++++++++++++----------
 1 file changed, 109 insertions(+), 29 deletions(-)

diff --git a/xen/arch/arm/arm32/entry.S b/xen/arch/arm/arm32/entry.S
index 150cbc0b4b..ec90cca093 100644
--- a/xen/arch/arm/arm32/entry.S
+++ b/xen/arch/arm/arm32/entry.S
@@ -4,6 +4,17 @@
 #include <asm/alternative.h>
 #include <public/xen.h>
 
+/*
+ * Short-hands to defined the interrupts (A, I, F)
+ *
+ * _ means the interrupt state will not change
+ * X means the state of interrupt X will change
+ *
+ * To be used with msr cpsr_* only
+ */
+#define IFLAGS_AIF      PSR_ABT_MASK | PSR_IRQ_MASK | PSR_FIQ_MASK
+#define IFLAGS_A_F      PSR_ABT_MASK | PSR_FIQ_MASK
+
 #define SAVE_ONE_BANKED(reg)    mrs r11, reg; str r11, [sp, #UREGS_##reg]
 #define RESTORE_ONE_BANKED(reg) ldr r11, [sp, #UREGS_##reg]; msr reg, r11
 
@@ -106,10 +117,18 @@ skip_check:
         mov pc, lr
 
         /*
-         * Macro to define trap entry. The iflags corresponds to the list of
-         * interrupts (Asynchronous Abort, IRQ, FIQ) to unmask.
+         * Macro to define a trap entry.
+         *
+         *  @guest_iflags: Optional list of interrupts to unmask when
+         *      entering from guest context. As this is used with cpsie,
+         *      the letter (a, i, f) should be used.
+         *
+         *  @hyp_iflags: Optional list of interrupts to inherit when
+         *      entering from hypervisor context. Any interrupts not
+         *      listed will be kept unchanged. As this is used with cpsr_*,
+         *      IFLAGS_* short-hands should be used.
          */
-        .macro vector trap, iflags
+        .macro vector trap, guest_iflags=n, hyp_iflags=0
         /* Save registers in the stack */
         sub     sp, #(UREGS_SP_usr - UREGS_sp) /* SP, LR, SPSR, PC */
         push    {r0-r12}                       /* Save R0-R12 */
@@ -127,12 +146,39 @@ skip_check:
 
         mrs     r11, SPSR_hyp
         str     r11, [sp, #UREGS_cpsr]
-        and     r11, #PSR_MODE_MASK
-        cmp     r11, #PSR_MODE_HYP
-        blne    save_guest_regs
 
+        /*
+         * We need to distinguish whether we came from guest or
+         * hypervisor context.
+         */
+        and     r0, r11, #PSR_MODE_MASK
+        cmp     r0, #PSR_MODE_HYP
+
+        bne     1f
+        /*
+         * Trap from the hypervisor
+         *
+         * Inherit the state of the interrupts from the hypervisor
+         * context. For that we need to use SPSR (stored in r11) and
+         * modify CPSR accordingly.
+         *
+         * CPSR = (CPSR & ~hyp_iflags) | (SPSR & hyp_iflags)
+         */
+        mrs     r10, cpsr
+        bic     r10, r10, #\hyp_iflags
+        and     r11, r11, #\hyp_iflags
+        orr     r10, r10, r11
+        msr     cpsr_cx, r10
+        b       2f
+
+1:
+        /* Trap from the guest */
+        bl      save_guest_regs
+        .if     \guest_iflags != n
+        cpsie   \guest_iflags
+        .endif
+2:
         /* We are ready to handle the trap, setup the registers and jump. */
-        cpsie   \iflags
         adr     lr, return_from_trap
         mov     r0, sp
         /*
@@ -144,20 +190,6 @@ skip_check:
         b       do_trap_\trap
         .endm
 
-#define __DEFINE_TRAP_ENTRY(trap, iflags)                               \
-        ALIGN;                                                          \
-trap_##trap:                                                            \
-        vector trap, iflags
-
-/* Trap handler which unmask IRQ/Abort, keep FIQ masked */
-#define DEFINE_TRAP_ENTRY(trap) __DEFINE_TRAP_ENTRY(trap, ai)
-
-/* Trap handler which unmask Abort, keep IRQ/FIQ masked */
-#define DEFINE_TRAP_ENTRY_NOIRQ(trap) __DEFINE_TRAP_ENTRY(trap, a)
-
-/* Trap handler which unmask IRQ, keep Abort/FIQ masked */
-#define DEFINE_TRAP_ENTRY_NOABORT(trap) __DEFINE_TRAP_ENTRY(trap, i)
-
         .align 5
 GLOBAL(hyp_traps_vector)
         b trap_reset                    /* 0x00 - Reset */
@@ -228,14 +260,62 @@ decode_vectors:
 
 #endif /* CONFIG_HARDEN_BRANCH_PREDICTOR */
 
-DEFINE_TRAP_ENTRY(reset)
-DEFINE_TRAP_ENTRY(undefined_instruction)
-DEFINE_TRAP_ENTRY(hypervisor_call)
-DEFINE_TRAP_ENTRY(prefetch_abort)
-DEFINE_TRAP_ENTRY(guest_sync)
-DEFINE_TRAP_ENTRY_NOIRQ(irq)
-DEFINE_TRAP_ENTRY_NOIRQ(fiq)
-DEFINE_TRAP_ENTRY_NOABORT(data_abort)
+/* Vector not used by the Hypervisor. */
+trap_reset:
+        vector reset
+
+/*
+ * Vector only used by the Hypervisor.
+ *
+ * While the exception can be executed with all the interrupts (e.g.
+ * IRQ) unmasked, the interrupted context may have purposefully masked
+ * some of them. So we want to inherit the state from the interrupted
+ * context.
+ */
+trap_undefined_instruction:
+        vector undefined_instruction, hyp_iflags=IFLAGS_AIF
+
+/* We should never reach this trap */
+trap_hypervisor_call:
+        vector hypervisor_call
+
+/*
+ * Vector only used by the hypervisor.
+ *
+ * While the exception can be executed with all the interrupts (e.g.
+ * IRQ) unmasked, the interrupted context may have purposefully masked
+ * some of them. So we want to inherit the state from the interrupted
+ * context.
+ */
+trap_prefetch_abort:
+       vector prefetch_abort, hyp_iflags=IFLAGS_AIF
+
+/*
+ * Vector only used by the hypervisor.
+ *
+ * Data Abort should be rare and most likely fatal. It is best to not
+ * unmask any interrupts to limit the amount of code that can run before
+ * the Data Abort is treated.
+ */
+trap_data_abort:
+        vector data_abort
+
+/* Vector only used by the guest. We can unmask Abort/IRQ. */
+trap_guest_sync:
+        vector guest_sync, guest_iflags=ai
+
+
+/* Vector used by the hypervisor and the guest. */
+trap_irq:
+        vector irq, guest_iflags=a, hyp_iflags=IFLAGS_A_F
+
+/*
+ * Vector used by the hypervisor and the guest.
+ *
+ * FIQ are not meant to happen, so we don't unmask any interrupts.
+ */
+trap_fiq:
+        vector fiq
 
 return_from_trap:
         /*
--
generated by git-patchbot for /home/xen/git/xen.git#staging

_______________________________________________
Xen-changelog mailing list
Xen-changelog@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/xen-changelog

 


Rackspace

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