Skip to contentSkip to conversation

xen/arm64: io: Decode ldr/str post-indexing instruction

PATCH Superseded Ayan Kumar Halder 3 participants 28 messages Mar 1, 2022 – Mar 8, 2022

v9 of 11 Superseded Latest is v11 →

lore
Share & tools
lore
#0 lore raw

v9 cover xen/arm64: io: Decode ldr/str post-indexing instruction

Hi All,

The patch series introduces support to decode instructions by Xen when ISS is invalid. Currently, when the guest executes post indexing ldr/str instructions on emulated MMIO, these instructions are trapped into Xen as a data abort. Xen reads hsr_dabt.isv == 0, so ISS is invalid. Therefore, it reads the faulting instruction's opcode from guest's PC. It decodes and executes the instruction on the emulated region.

While doing the patch, we found two bugs in the codebase. I have addressed them in patches 3 and 4. These bugs were discussed with Julien on IRC chat. The purpose of addressing these bugs (in this series) is that 1. We should not forget about them. 2. To get clarity if our understanding is correct.

Ayan Kumar Halder (4):
xen/arm64: Decode ldr/str post increment operations xen/arm64: io: Support instructions (for which ISS is not valid) on
emulated MMIO region using MMIO/ioreq handler
xen/arm64: io: Handle the abort due to access to stage1 translation
table
xen/arm64: io: Handle data abort due to cache maintenance instructions

 xen/arch/arm/arm32/traps.c        |  11 +++
 xen/arch/arm/arm64/traps.c        |  47 +++++++++++++
 xen/arch/arm/decode.c             |  80 ++++++++++++++++++++-
 xen/arch/arm/decode.h             |  48 +++++++++++--
 xen/arch/arm/include/asm/domain.h |   4 ++
 xen/arch/arm/include/asm/mmio.h   |  22 +++++-
 xen/arch/arm/include/asm/traps.h  |   2 +
 xen/arch/arm/io.c                 | 112 +++++++++++++++++++++---------
 xen/arch/arm/ioreq.c              |   7 +-
 xen/arch/arm/traps.c              |  93 ++++++++++++++++++++-----
 xen/arch/x86/include/asm/ioreq.h  |   3 +
 xen/include/xen/sched.h           |   2 +
 12 files changed, 369 insertions(+), 62 deletions(-)

Changelog :-
v2 - 1. Updated the rn register after reading from it. (Pointed by Julien,
Stefano)
2. Used a union to represent the instruction opcode (Suggestd by Bertrand)
3. Fixed coding style issues (Pointed by Julien)
4. In the previous patch, I was updating dabt->sign based on the signedness of imm9. This was incorrect. As mentioned in ARMv8 ARM DDI 0487G.b, Page 3221, SSE indicates the signedness of the data item loaded. In our case, the data item loaded is always unsigned.

v3- 1. Handled all the variants of ldr/str (ie 64, 32, 16, 8 bit variants).
Thus, I have removed the check for "instr->code.opc == 0" (Suggested by Andre)
2. Handled the scenario when rn = SP, rt = XZR (Suggested by Jan, Andre)
3. Added restriction for "rt != rn" (Suggested by Andre)
4. Moved union ldr_str_instr_class {} to decode.h. This is the header included by io.c and decode.c (where the union is referred). (Suggested by Jan)
5. Indentation and typo fixes (Suggested by Jan)

v4- 1. Fixed the patch as per Stefano's comments on v3. They are as follows :-
1.1 Use macros to determine the fixed values in the instruction opcode 1.2 Checked if instr != NULL
1.3 Changed some data types and added #define ARM_64 for AArch64 specific
code
1.4 Moved post_increment_register() to decode.c so that the decoding
logic is confined to a single file.
1.5 Moved some checks from post_increment_register() to
decode_loadstore_postindexing()
1.6 Removed a duplicate check
2. Updated the commit message as per Andre's comments.
3. Changed the names of a label and some comments. *32bit* was erroneously mentioned in a label and comments in decode_loadstore_postindexing() although the function handled all variants of ldr/str post indexing.

v5- 1. Renamed decode_loadstore_postindexing() to decode_arm64(). The reason
being this will be extended in future to support more instructions for which hsr_badt.isv = 0
2. Introduce a function try_decode_instruction_invalid_iss() to determine if the instruction needs to be decoded before invoking decode_instruction().

It checks :-
2.1 dabt->s1ptw - Returns IO_UNHANDLED
2.2 dabt->cache - Returns IO_IGNORED. (new enum instroduced to let the
caller know that the instruction needs to be ignored by Xen. Thus the caller needs to increment the PC and return to the guest.

3. Invoked try_decode_instruction_invalid_iss() from the following 2 places :- 3.a - try_handle_mmio() - When we have determined that there is a valid mmio handler.
3.b - try_fwd_ioserv()
When ioserver completes the io request, the acknowledgement is sent via handle_ioserv(). Here, we need to increment the register. As there is no common data shared between try_fwd_ioserv() and handle_ioserv(), we need to decode the instruction again in handle_ioserv() to determine rn, imm9.

(NOTE to Reviewers) - This does not feel correct. However, I could not think of a better approach. Please provide your inputs.

4. Augumented struct hsr_dabt{} with struct hsr_dabt_instr_details{} to hold rn and imm9. This is passed to post_increment_register() to update rn.
5. Other style changes as suggested in v4.

v6 - 1. Split the patch into three parts.

v7 - 1. Merged patch2 and patch3 into a single patch.

v8 - 1. Changes mentioned in the individual patches.

v9 - 1. Added patches 3 and 4 to address the bugs found in the existing codebase.
2. Changes mentioned in the individual patches.

Signature · 1 line
2.17.1
#1 lore raw

PATCH v9 1/4 xen/arm64: Decode ldr/str post increment operations

At the moment, Xen does not decode any of the arm64 instructions. This means that when hsr_dabt.isv == 0, Xen cannot handle those instructions. This will lead to Xen to abort the guests (from which those instructions originate).

With this patch, Xen is able to decode ldr/str post indexing instructions. These are a subset of instructions for which hsr_dabt.isv == 0.

The following instructions are now supported by Xen :-

1.      ldr     x2,    [x1],    #8
2.      ldr     w2,    [x1],    #-4
3.      ldr     x2,    [x1],    #-8
4.      ldr     w2,    [x1],    #4
5.      ldrh    w2,    [x1],    #2
6.      ldrb    w2,    [x1],    #1
7.      str     x2,    [x1],    #8
8.      str     w2,    [x1],    #-4
9.      strh    w2,    [x1],    #2
10.     strb    w2,    [x1],    #1

In the subsequent patch, decode_arm64() will get invoked when hsr_dabt.isv == 0.

---

Changelog :-

v2..v5 - Mentioned in the cover letter.

v6 - 1. Fixed the code style issues as mentioned in v5.

v7 - No change.

v8 - 1. Removed some un-necessary header files inclusion.
2. Some style changes pointed out in v7.

v9 - 1. Rebased on top of the master.
2. Renamed psr_mode_is_32bit to regs_mode_is_32bit.

 xen/arch/arm/decode.c           | 79 ++++++++++++++++++++++++++++++++-
 xen/arch/arm/decode.h           | 48 +++++++++++++++++---
 xen/arch/arm/include/asm/mmio.h |  4 ++
 xen/arch/arm/io.c               |  2 +-
 4 files changed, 124 insertions(+), 9 deletions(-)
xen/arch/arm/decode.c +77 -2
index 792c2e92a7..3add87e83a 100644
@@ -84,6 +84,78 @@ bad_thumb2:
     return 1;
 }
 
+static int decode_arm64(register_t pc, mmio_info_t *info)
+{
+    union instr opcode = {0};
+    struct hsr_dabt *dabt = &info->dabt;
+    struct instr_details *dabt_instr = &info->dabt_instr;
+
+    if ( raw_copy_from_guest(&opcode.value, (void * __user)pc, sizeof (opcode)) )
+    {
+        gprintk(XENLOG_ERR, "Could not copy the instruction from PC\n");
+        return 1;
+    }
+
+    /*
+     * Refer Arm v8 ARM DDI 0487G.b, Page - C6-1107
+     * "Shared decode for all encodings" (under ldr immediate)
+     * If n == t && n != 31, then the return value is implementation defined
+     * (can be WBSUPPRESS, UNKNOWN, UNDEFINED or NOP). Thus, we do not support
+     * this. This holds true for ldrb/ldrh immediate as well.
+     *
+     * Also refer, Page - C6-1384, the above described behaviour is same for
+     * str immediate. This holds true for strb/strh immediate as well
+     */
+    if ( (opcode.ldr_str.rn == opcode.ldr_str.rt) && (opcode.ldr_str.rn != 31) )
+    {
+        gprintk(XENLOG_ERR, "Rn should not be equal to Rt except for r31\n");
+        goto bad_loadstore;
+    }
+
+    /* First, let's check for the fixed values */
+    if ( (opcode.value & POST_INDEX_FIXED_MASK) != POST_INDEX_FIXED_VALUE )
+    {
+        gprintk(XENLOG_ERR,
+                "Decoding instruction 0x%x is not supported\n", opcode.value);
+        goto bad_loadstore;
+    }
+
+    if ( opcode.ldr_str.v != 0 )
+    {
+        gprintk(XENLOG_ERR,
+                "ldr/str post indexing for vector types are not supported\n");
+        goto bad_loadstore;
+    }
+
+    /* Check for STR (immediate) */
+    if ( opcode.ldr_str.opc == 0 )
+        dabt->write = 1;
+    /* Check for LDR (immediate) */
+    else if ( opcode.ldr_str.opc == 1 )
+        dabt->write = 0;
+    else
+    {
+        gprintk(XENLOG_ERR,
+                "Decoding ldr/str post indexing is not supported for this variant\n");
+        goto bad_loadstore;
+    }
+
+    gprintk(XENLOG_INFO,
+            "opcode->ldr_str.rt = 0x%x, opcode->ldr_str.size = 0x%x, opcode->ldr_str.imm9 = %d\n",
+            opcode.ldr_str.rt, opcode.ldr_str.size, opcode.ldr_str.imm9);
+
+    update_dabt(dabt, opcode.ldr_str.rt, opcode.ldr_str.size, false);
+
+    dabt_instr->rn = opcode.ldr_str.rn;
+    dabt_instr->imm9 = opcode.ldr_str.imm9;
+
+    return 0;
+
+ bad_loadstore:
+    gprintk(XENLOG_ERR, "unhandled Arm instruction 0x%x\n", opcode.value);
+    return 1;
+}
+
 static int decode_thumb(register_t pc, struct hsr_dabt *dabt)
 {
     uint16_t instr;
@@ -150,10 +222,13 @@ bad_thumb:
     return 1;
 }
 
-int decode_instruction(const struct cpu_user_regs *regs, struct hsr_dabt *dabt)
+int decode_instruction(const struct cpu_user_regs *regs, mmio_info_t *info)
 {
     if ( is_32bit_domain(current->domain) && regs->cpsr & PSR_THUMB )
-        return decode_thumb(regs->pc, dabt);
+        return decode_thumb(regs->pc, &info->dabt);
+
+    if ( !regs_mode_is_32bit(regs) )
+        return decode_arm64(regs->pc, info);
 
     /* TODO: Handle ARM instruction */
     gprintk(XENLOG_ERR, "unhandled ARM instruction\n");
xen/arch/arm/decode.h +42 -6

61 lines not on this page · raw message

xen/arch/arm/include/asm/mmio.h +4 -0
index 7ab873cb8f..3354d9c635 100644
@@ -29,6 +29,10 @@
 typedef struct
 {
     struct hsr_dabt dabt;
+    struct instr_details {
+        unsigned long rn:5;
+        signed int imm9:9;
+    } dabt_instr;
     paddr_t gpa;
 } mmio_info_t;
 
xen/arch/arm/io.c +1 -1
index 1a066f9ae5..fad103bdbd 100644
@@ -141,7 +141,7 @@ enum io_state try_handle_mmio(struct cpu_user_regs *regs,
     {
         int rc;
 
-        rc = decode_instruction(regs, &info.dabt);
+        rc = decode_instruction(regs, &info);
         if ( rc )
         {
             gprintk(XENLOG_DEBUG, "Unable to decode instruction\n");
Signature · 1 line
2.17.1
Reviewed-by on v9 1/4
#2 lore raw

Re: [XEN v9 1/4] xen/arm64: Decode ldr/str post increment operations

Quoted 20 lines · Ayan Kumar Halder

At the moment, Xen does not decode any of the arm64 instructions. This means that when hsr_dabt.isv == 0, Xen cannot handle those instructions. This will lead to Xen to abort the guests (from which those instructions originate).

With this patch, Xen is able to decode ldr/str post indexing instructions. These are a subset of instructions for which hsr_dabt.isv == 0.

The following instructions are now supported by Xen :-

1.      ldr     x2,    [x1],    #8
2.      ldr     w2,    [x1],    #-4
3.      ldr     x2,    [x1],    #-8
4.      ldr     w2,    [x1],    #4
5.      ldrh    w2,    [x1],    #2
6.      ldrb    w2,    [x1],    #1
7.      str     x2,    [x1],    #8
8.      str     w2,    [x1],    #-4
9.      strh    w2,    [x1],    #2
10.     strb    w2,    [x1],    #1
Quoted 4 lines

In the subsequent patch, decode_arm64() will get invoked when hsr_dabt.isv == 0.

Quoted 218 lines

---

Changelog :-

v2..v5 - Mentioned in the cover letter.

v6 - 1. Fixed the code style issues as mentioned in v5.

v7 - No change.

v8 - 1. Removed some un-necessary header files inclusion.
2. Some style changes pointed out in v7.

v9 - 1. Rebased on top of the master.
2. Renamed psr_mode_is_32bit to regs_mode_is_32bit.

 xen/arch/arm/decode.c           | 79 ++++++++++++++++++++++++++++++++-
 xen/arch/arm/decode.h           | 48 +++++++++++++++++---
 xen/arch/arm/include/asm/mmio.h |  4 ++
 xen/arch/arm/io.c               |  2 +-
 4 files changed, 124 insertions(+), 9 deletions(-)
xen/arch/arm/decode.c +77 -2
index 792c2e92a7..3add87e83a 100644
@@ -84,6 +84,78 @@ bad_thumb2:
     return 1;
 }
 
+static int decode_arm64(register_t pc, mmio_info_t *info)
+{
+    union instr opcode = {0};
+    struct hsr_dabt *dabt = &info->dabt;
+    struct instr_details *dabt_instr = &info->dabt_instr;
+
+    if ( raw_copy_from_guest(&opcode.value, (void * __user)pc, sizeof (opcode)) )
+    {
+        gprintk(XENLOG_ERR, "Could not copy the instruction from PC\n");
+        return 1;
+    }
+
+    /*
+     * Refer Arm v8 ARM DDI 0487G.b, Page - C6-1107
+     * "Shared decode for all encodings" (under ldr immediate)
+     * If n == t && n != 31, then the return value is implementation defined
+     * (can be WBSUPPRESS, UNKNOWN, UNDEFINED or NOP). Thus, we do not support
+     * this. This holds true for ldrb/ldrh immediate as well.
+     *
+     * Also refer, Page - C6-1384, the above described behaviour is same for
+     * str immediate. This holds true for strb/strh immediate as well
+     */
+    if ( (opcode.ldr_str.rn == opcode.ldr_str.rt) && (opcode.ldr_str.rn != 31) )
+    {
+        gprintk(XENLOG_ERR, "Rn should not be equal to Rt except for r31\n");
+        goto bad_loadstore;
+    }
+
+    /* First, let's check for the fixed values */
+    if ( (opcode.value & POST_INDEX_FIXED_MASK) != POST_INDEX_FIXED_VALUE )
+    {
+        gprintk(XENLOG_ERR,
+                "Decoding instruction 0x%x is not supported\n", opcode.value);
+        goto bad_loadstore;
+    }
+
+    if ( opcode.ldr_str.v != 0 )
+    {
+        gprintk(XENLOG_ERR,
+                "ldr/str post indexing for vector types are not supported\n");
+        goto bad_loadstore;
+    }
+
+    /* Check for STR (immediate) */
+    if ( opcode.ldr_str.opc == 0 )
+        dabt->write = 1;
+    /* Check for LDR (immediate) */
+    else if ( opcode.ldr_str.opc == 1 )
+        dabt->write = 0;
+    else
+    {
+        gprintk(XENLOG_ERR,
+                "Decoding ldr/str post indexing is not supported for this variant\n");
+        goto bad_loadstore;
+    }
+
+    gprintk(XENLOG_INFO,
+            "opcode->ldr_str.rt = 0x%x, opcode->ldr_str.size = 0x%x, opcode->ldr_str.imm9 = %d\n",
+            opcode.ldr_str.rt, opcode.ldr_str.size, opcode.ldr_str.imm9);
+
+    update_dabt(dabt, opcode.ldr_str.rt, opcode.ldr_str.size, false);
+
+    dabt_instr->rn = opcode.ldr_str.rn;
+    dabt_instr->imm9 = opcode.ldr_str.imm9;
+
+    return 0;
+
+ bad_loadstore:
+    gprintk(XENLOG_ERR, "unhandled Arm instruction 0x%x\n", opcode.value);
+    return 1;
+}
+
 static int decode_thumb(register_t pc, struct hsr_dabt *dabt)
 {
     uint16_t instr;
@@ -150,10 +222,13 @@ bad_thumb:
     return 1;
 }
 
-int decode_instruction(const struct cpu_user_regs *regs, struct hsr_dabt *dabt)
+int decode_instruction(const struct cpu_user_regs *regs, mmio_info_t *info)
 {
     if ( is_32bit_domain(current->domain) && regs->cpsr & PSR_THUMB )
-        return decode_thumb(regs->pc, dabt);
+        return decode_thumb(regs->pc, &info->dabt);
+
+    if ( !regs_mode_is_32bit(regs) )
+        return decode_arm64(regs->pc, info);
 
     /* TODO: Handle ARM instruction */
     gprintk(XENLOG_ERR, "unhandled ARM instruction\n");
xen/arch/arm/decode.h +42 -6
index 4613763bdb..13db8ac968 100644
@@ -23,19 +23,55 @@
 #include <asm/regs.h>
 #include <asm/processor.h>
 
-/**
+/*
+ * Refer to the ARMv8 ARM (DDI 0487G.b), Section C4.1.4 Loads and Stores
+ * Page 318 specifies the following bit pattern for
+ * "load/store register (immediate post-indexed)".
+ *
+ * 31 30 29  27 26 25  23   21 20              11   9         4       0
+ * ___________________________________________________________________
+ * |size|1 1 1 |V |0 0 |opc |0 |      imm9     |0 1 |  Rn     |  Rt   |
+ * |____|______|__|____|____|__|_______________|____|_________|_______|
+ */
+union instr {
+    uint32_t value;
+    struct {
+        unsigned int rt:5;     /* Rt register */
+        unsigned int rn:5;     /* Rn register */
+        unsigned int fixed1:2; /* value == 01b */
+        signed int imm9:9;     /* imm9 */
+        unsigned int fixed2:1; /* value == 0b */
+        unsigned int opc:2;    /* opc */
+        unsigned int fixed3:2; /* value == 00b */
+        unsigned int v:1;      /* vector */
+        unsigned int fixed4:3; /* value == 111b */
+        unsigned int size:2;   /* size */
+    } ldr_str;
+};
+
+#define POST_INDEX_FIXED_MASK   0x3B200C00
+#define POST_INDEX_FIXED_VALUE  0x38000400
+
+/*
  * Decode an instruction from pc
- * /!\ This function is not intended to fully decode an instruction. It
- * considers that the instruction is valid.
+ * /!\ This function is intended to decode an instruction. It considers that the
+ * instruction is valid.
  *
- * This function will get:
- *  - The transfer register
+ * In case of thumb mode, this function will get:
+ *  - The transfer register (ie Rt)
  *  - Sign bit
  *  - Size
+ *
+ * In case of arm64 mode, this function will get:
+ * - The transfer register (ie Rt)
+ * - The source register (ie Rn)
+ * - Size
+ * - Immediate offset
+ * - Read or write
  */
 
 int decode_instruction(const struct cpu_user_regs *regs,
-                       struct hsr_dabt *dabt);
+                       mmio_info_t *info);
 
 #endif /* __ARCH_ARM_DECODE_H_ */
 
xen/arch/arm/include/asm/mmio.h +4 -0
index 7ab873cb8f..3354d9c635 100644
@@ -29,6 +29,10 @@
 typedef struct
 {
     struct hsr_dabt dabt;
+    struct instr_details {
+        unsigned long rn:5;
+        signed int imm9:9;
+    } dabt_instr;
     paddr_t gpa;
 } mmio_info_t;
 
xen/arch/arm/io.c +1 -1
index 1a066f9ae5..fad103bdbd 100644
@@ -141,7 +141,7 @@ enum io_state try_handle_mmio(struct cpu_user_regs *regs,
     {
         int rc;
 
-        rc = decode_instruction(regs, &info.dabt);
+        rc = decode_instruction(regs, &info);
         if ( rc )
         {
             gprintk(XENLOG_DEBUG, "Unable to decode instruction\n");
Signature · 1 line
2.17.1
#3 lore raw

PATCH v9 2/4 xen/arm64: io: Support instructions (for which ISS is not valid) on emulated MMIO region using MMIO/ioreq handler

In reply to Ayan Kumar Halder

When an instruction is trapped in Xen due to translation fault, Xen checks if the ISS is invalid (for data abort) or it is an instruction abort. If so, Xen tries to resolve the translation fault using p2m page tables. In case of data abort, Xen will try to map the mmio region to the guest (ie tries to emulate the mmio region).

If the ISS is not valid and it is a data abort, then Xen tries to decode the instruction. In case of ioreq, Xen saves the decoding state, rn and imm9 to vcpu_io. Whenever the vcpu handles the ioreq successfully, it will read the decoding state to determine if the instruction decoded was a ldr/str post indexing (ie INSTR_LDR_STR_POSTINDEXING). If so, it uses these details to post increment rn.

In case of mmio handler, if the mmio operation was successful, then Xen retrives the decoding state, rn and imm9. For state == INSTR_LDR_STR_POSTINDEXING, Xen will update rn.

If there is an error encountered while decoding/executing the instruction, Xen will forward the abort to the guest.

---

Changelog :-

v2..v5 - Mentioned in the cover letter.

v6 - 1. Mantained the decoding state of the instruction. This is used by the caller to either abort the guest or retry or ignore or perform read/write on the mmio region.

2. try_decode() invokes decoding for both aarch64 and thumb state. (Previously it used to invoke decoding only for aarch64 state). Thus, it handles all the checking of the registers before invoking any decoding of instruction. try_decode_instruction_invalid_iss() has thus been removed.

3. Introduced a new field('enum instr_decode_state state') inside 'struct instr_details'. This holds the decoding state of the instruction. This is later read by the post_increment_register() to determine if rn needs to be incremented. Also, this is read by the callers of try_decode_instruction() to determine if the instruction was valid or ignored or to be retried or error or decoded successfully.

4. Also stored 'instr_details' inside 'struct ioreq'. This enables arch_ioreq_complete_mmio() to invoke post_increment_register() without decoding the instruction again.

5. Check hsr.dabt.valid in do_trap_stage2_abort_guest(). If it is not valid, then decode the instruction. This ensures that try_handle_mmio() is invoked only when the instruction is either valid or decoded successfully.

6. Inside do_trap_stage2_abort_guest(), if hsr.dabt.valid is not set, then resolve the translation fault before trying to decode the instruction. If translation fault is resolved, then return to the guest to execute the instruction again.

v7 - 1. Moved the decoding instruction details ie instr_details from 'struct ioreq' to 'struct vcpu_io'.

2. The instruction is decoded only when we get a data abort.

3. Replaced ASSERT_UNREACHABLE() with domain_crash(). The reason being asserts can be disabled in some builds. In this scenario when the guest's cpsr is in an erroneous state, Xen should crash the guest.

4. Introduced check_p2m() which invokes p2m_resolve_translation_fault() and try_map_mmio() to resolve translation fault by configuring the page tables. This gets invoked first if ISS is invalid and it is an instruction abort. If it is a data abort and hsr.dabt.s1ptw is set or try_handle_mmio() returns IO_UNHANDLED, then check_p2m() gets invoked again.

v8 - 1. Removed the handling of data abort when info->dabt.cache is set. This will be implemented in a subsequent patch. (Not as part of this series)

2. When the data abort is due to access to stage 1 translation tables, Xen will try to fix the mapping of the page table for the corresponding address. If this returns an error, Xen will abort the guest. Else, it will ask the guest to retry the instruction.

3. Changed v->io.info.dabt_instr from pointer to variable. The reason being that arch_ioreq_complete_mmio() is called from leave_hypervisor_to_guest(). That is after do_trap_stage2_abort_guest() has been invoked. So the original variable will be no longer valid.

4. Some other style issues pointed out in v7.

v9 - 1. Ensure that "Erratum 766422" is handled only when ISS is valid.

2. Whenever Xen receives and instruction abort or data abort (with invalid ISS), Xen should first try to resolve the p2m translation fault or see if it it needs to map a MMIO region. If it succeeds, it should return to the guest to retry the instruction.

3. Removed handling of "dabt.s1ptw == 1" aborts. This is addressed in patch3 as it is an existing bug in codebase.

4. Various style issues pointed by Julien in v8.

 xen/arch/arm/arm32/traps.c        | 11 ++++
 xen/arch/arm/arm64/traps.c        | 47 ++++++++++++++++
 xen/arch/arm/decode.c             |  1 +
 xen/arch/arm/include/asm/domain.h |  4 ++
 xen/arch/arm/include/asm/mmio.h   | 17 +++++-
 xen/arch/arm/include/asm/traps.h  |  2 +
 xen/arch/arm/io.c                 | 90 +++++++++++++++++++------------
 xen/arch/arm/ioreq.c              |  7 ++-
 xen/arch/arm/traps.c              | 77 ++++++++++++++++++++------
 xen/arch/x86/include/asm/ioreq.h  |  3 ++
 xen/include/xen/sched.h           |  2 +
 11 files changed, 207 insertions(+), 54 deletions(-)
xen/arch/arm/arm32/traps.c +11 -0
index 9c9790a6d1..159e3cef8b 100644
@@ -18,9 +18,11 @@
 
 #include <xen/lib.h>
 #include <xen/kernel.h>
+#include <xen/sched.h>
 
 #include <public/xen.h>
 
+#include <asm/mmio.h>
 #include <asm/processor.h>
 #include <asm/traps.h>
 
@@ -82,6 +84,15 @@ void do_trap_data_abort(struct cpu_user_regs *regs)
         do_unexpected_trap("Data Abort", regs);
 }
 
+void post_increment_register(const struct instr_details *instr)
+{
+    /*
+     * We have not implemented decoding of post indexing instructions for 32 bit.
+     * Thus, this should be unreachable.
+     */
+    domain_crash(current->domain);
+}
+
 /*
  * Local variables:
  * mode: C
xen/arch/arm/arm64/traps.c +47 -0

59 lines not on this page · raw message

xen/arch/arm/decode.c +1 -0
index 3add87e83a..16ad0747bb 100644
@@ -146,6 +146,7 @@ static int decode_arm64(register_t pc, mmio_info_t *info)
 
     update_dabt(dabt, opcode.ldr_str.rt, opcode.ldr_str.size, false);
 
+    dabt_instr->state = INSTR_LDR_STR_POSTINDEXING;
     dabt_instr->rn = opcode.ldr_str.rn;
     dabt_instr->imm9 = opcode.ldr_str.imm9;
 
xen/arch/arm/include/asm/domain.h +4 -0
index c56f6e4398..ed63c2b6f9 100644
@@ -281,6 +281,10 @@ static inline void arch_vcpu_block(struct vcpu *v) {}
 /* vPCI is not available on Arm */
 #define has_vpci(d)    ({ (void)(d); false; })
 
+struct arch_vcpu_io {
+    struct instr_details dabt_instr; /* when the instruction is decoded */
+};
+
 #endif /* __ASM_DOMAIN_H__ */
 
 /*
xen/arch/arm/include/asm/mmio.h +15 -2
index 3354d9c635..ef2c57a2d5 100644
@@ -26,12 +26,24 @@
 
 #define MAX_IO_HANDLER  16
 
+enum instr_decode_state
+{
+    INSTR_ERROR,                    /* Error encountered while decoding instr */
+    INSTR_VALID,                    /* ISS is valid, so no need to decode */
+    /*
+     * Instruction is decoded successfully. It is a ldr/str post indexing
+     * instruction.
+     */
+    INSTR_LDR_STR_POSTINDEXING
+};
+
 typedef struct
 {
     struct hsr_dabt dabt;
     struct instr_details {
         unsigned long rn:5;
         signed int imm9:9;
+        enum instr_decode_state state;
     } dabt_instr;
     paddr_t gpa;
 } mmio_info_t;
@@ -69,14 +81,15 @@ struct vmmio {
 };
 
 enum io_state try_handle_mmio(struct cpu_user_regs *regs,
-                              const union hsr hsr,
-                              paddr_t gpa);
+                              mmio_info_t *info);
 void register_mmio_handler(struct domain *d,
                            const struct mmio_handler_ops *ops,
                            paddr_t addr, paddr_t size, void *priv);
 int domain_io_init(struct domain *d, int max_count);
 void domain_io_free(struct domain *d);
 
+void try_decode_instruction(const struct cpu_user_regs *regs,
+                            mmio_info_t *info);
 
 #endif  /* __ASM_ARM_MMIO_H__ */
 
xen/arch/arm/include/asm/traps.h +2 -0
index 2ed2b85c6f..95c46ad391 100644
@@ -109,6 +109,8 @@ static inline register_t sign_extend(const struct hsr_dabt dabt, register_t r)
     return r;
 }
 
+void post_increment_register(const struct instr_details *instr);
+
 #endif /* __ASM_ARM_TRAPS__ */
 /*
  * Local variables:
xen/arch/arm/io.c +56 -34

113 lines not on this page · raw message

xen/arch/arm/ioreq.c +5 -2
index 308650b400..58cd320b5a 100644
@@ -47,6 +47,7 @@ enum io_state try_fwd_ioserv(struct cpu_user_regs *regs,
                              struct vcpu *v, mmio_info_t *info)
 {
     struct vcpu_io *vio = &v->io;
+    struct dabt_instr instr = info->dabt_instr;
     ioreq_t p = {
         .type = IOREQ_TYPE_COPY,
         .addr = info->gpa,
@@ -76,10 +77,10 @@ enum io_state try_fwd_ioserv(struct cpu_user_regs *regs,
     if ( !s )
         return IO_UNHANDLED;
 
-    if ( !info->dabt.valid )
-        return IO_ABORT;
+    ASSERT(dabt.valid);
 
     vio->req = p;
+    vio->info.dabt_instr = instr;
 
     rc = ioreq_send(s, &p, 0);
     if ( rc != IO_RETRY || v->domain->is_shutting_down )
@@ -95,6 +96,7 @@ enum io_state try_fwd_ioserv(struct cpu_user_regs *regs,
 bool arch_ioreq_complete_mmio(void)
 {
     struct vcpu *v = current;
+    struct instr_details dabt_instr = v->io.info.dabt_instr;
     struct cpu_user_regs *regs = guest_cpu_user_regs();
     const union hsr hsr = { .bits = regs->hsr };
 
@@ -106,6 +108,7 @@ bool arch_ioreq_complete_mmio(void)
 
     if ( handle_ioserv(regs, v) == IO_HANDLED )
     {
+        post_increment_register(&dabt_instr);
         advance_pc(regs, hsr);
         return true;
     }
xen/arch/arm/traps.c +61 -16

115 lines not on this page · raw message

xen/arch/x86/include/asm/ioreq.h +3 -0
index d06ce9a6ea..ecfe7f9fdb 100644
@@ -26,6 +26,9 @@
 #include <asm/hvm/ioreq.h>
 #endif
 
+struct arch_vcpu_io {
+};
+
 #endif /* __ASM_X86_IOREQ_H__ */
 
 /*
xen/include/xen/sched.h +2 -0
index 10ea969c7a..406d9bc610 100644
@@ -160,6 +160,8 @@ struct vcpu_io {
     /* I/O request in flight to device model. */
     enum vio_completion  completion;
     ioreq_t              req;
+    /* Arch specific info pertaining to the io request */
+    struct arch_vcpu_io  info;
 };
 
 struct vcpu
Signature · 1 line
2.17.1
#4 lore raw

Re: [XEN v9 2/4] xen/arm64: io: Support instructions (for which ISS is not valid) on emulated MMIO region using MMIO/ioreq handler

Quoted 429 lines · Ayan Kumar Halder

When an instruction is trapped in Xen due to translation fault, Xen checks if the ISS is invalid (for data abort) or it is an instruction abort. If so, Xen tries to resolve the translation fault using p2m page tables. In case of data abort, Xen will try to map the mmio region to the guest (ie tries to emulate the mmio region).

If the ISS is not valid and it is a data abort, then Xen tries to decode the instruction. In case of ioreq, Xen saves the decoding state, rn and imm9 to vcpu_io. Whenever the vcpu handles the ioreq successfully, it will read the decoding state to determine if the instruction decoded was a ldr/str post indexing (ie INSTR_LDR_STR_POSTINDEXING). If so, it uses these details to post increment rn.

In case of mmio handler, if the mmio operation was successful, then Xen retrives the decoding state, rn and imm9. For state == INSTR_LDR_STR_POSTINDEXING, Xen will update rn.

If there is an error encountered while decoding/executing the instruction, Xen will forward the abort to the guest.

---

Changelog :-

v2..v5 - Mentioned in the cover letter.

v6 - 1. Mantained the decoding state of the instruction. This is used by the caller to either abort the guest or retry or ignore or perform read/write on the mmio region.

2. try_decode() invokes decoding for both aarch64 and thumb state. (Previously it used to invoke decoding only for aarch64 state). Thus, it handles all the checking of the registers before invoking any decoding of instruction. try_decode_instruction_invalid_iss() has thus been removed.

3. Introduced a new field('enum instr_decode_state state') inside 'struct instr_details'. This holds the decoding state of the instruction. This is later read by the post_increment_register() to determine if rn needs to be incremented. Also, this is read by the callers of try_decode_instruction() to determine if the instruction was valid or ignored or to be retried or error or decoded successfully.

4. Also stored 'instr_details' inside 'struct ioreq'. This enables arch_ioreq_complete_mmio() to invoke post_increment_register() without decoding the instruction again.

5. Check hsr.dabt.valid in do_trap_stage2_abort_guest(). If it is not valid, then decode the instruction. This ensures that try_handle_mmio() is invoked only when the instruction is either valid or decoded successfully.

6. Inside do_trap_stage2_abort_guest(), if hsr.dabt.valid is not set, then resolve the translation fault before trying to decode the instruction. If translation fault is resolved, then return to the guest to execute the instruction again.

v7 - 1. Moved the decoding instruction details ie instr_details from 'struct ioreq' to 'struct vcpu_io'.

2. The instruction is decoded only when we get a data abort.

3. Replaced ASSERT_UNREACHABLE() with domain_crash(). The reason being asserts can be disabled in some builds. In this scenario when the guest's cpsr is in an erroneous state, Xen should crash the guest.

4. Introduced check_p2m() which invokes p2m_resolve_translation_fault() and try_map_mmio() to resolve translation fault by configuring the page tables. This gets invoked first if ISS is invalid and it is an instruction abort. If it is a data abort and hsr.dabt.s1ptw is set or try_handle_mmio() returns IO_UNHANDLED, then check_p2m() gets invoked again.

v8 - 1. Removed the handling of data abort when info->dabt.cache is set. This will be implemented in a subsequent patch. (Not as part of this series)

2. When the data abort is due to access to stage 1 translation tables, Xen will try to fix the mapping of the page table for the corresponding address. If this returns an error, Xen will abort the guest. Else, it will ask the guest to retry the instruction.

3. Changed v->io.info.dabt_instr from pointer to variable. The reason being that arch_ioreq_complete_mmio() is called from leave_hypervisor_to_guest(). That is after do_trap_stage2_abort_guest() has been invoked. So the original variable will be no longer valid.

4. Some other style issues pointed out in v7.

v9 - 1. Ensure that "Erratum 766422" is handled only when ISS is valid.

2. Whenever Xen receives and instruction abort or data abort (with invalid ISS), Xen should first try to resolve the p2m translation fault or see if it it needs to map a MMIO region. If it succeeds, it should return to the guest to retry the instruction.

3. Removed handling of "dabt.s1ptw == 1" aborts. This is addressed in patch3 as it is an existing bug in codebase.

4. Various style issues pointed by Julien in v8.

 xen/arch/arm/arm32/traps.c        | 11 ++++
 xen/arch/arm/arm64/traps.c        | 47 ++++++++++++++++
 xen/arch/arm/decode.c             |  1 +
 xen/arch/arm/include/asm/domain.h |  4 ++
 xen/arch/arm/include/asm/mmio.h   | 17 +++++-
 xen/arch/arm/include/asm/traps.h  |  2 +
 xen/arch/arm/io.c                 | 90 +++++++++++++++++++------------
 xen/arch/arm/ioreq.c              |  7 ++-
 xen/arch/arm/traps.c              | 77 ++++++++++++++++++++------
 xen/arch/x86/include/asm/ioreq.h  |  3 ++
 xen/include/xen/sched.h           |  2 +
 11 files changed, 207 insertions(+), 54 deletions(-)
xen/arch/arm/arm32/traps.c +11 -0
index 9c9790a6d1..159e3cef8b 100644
@@ -18,9 +18,11 @@
 
 #include <xen/lib.h>
 #include <xen/kernel.h>
+#include <xen/sched.h>
 
 #include <public/xen.h>
 
+#include <asm/mmio.h>
 #include <asm/processor.h>
 #include <asm/traps.h>
 
@@ -82,6 +84,15 @@ void do_trap_data_abort(struct cpu_user_regs *regs)
         do_unexpected_trap("Data Abort", regs);
 }
 
+void post_increment_register(const struct instr_details *instr)
+{
+    /*
+     * We have not implemented decoding of post indexing instructions for 32 bit.
+     * Thus, this should be unreachable.
+     */
+    domain_crash(current->domain);
+}
+
 /*
  * Local variables:
  * mode: C
xen/arch/arm/arm64/traps.c +47 -0
index 9113a15c7a..e18b6b2626 100644
@@ -17,6 +17,7 @@
  */
 
 #include <xen/lib.h>
+#include <xen/sched.h>
 
 #include <asm/hsr.h>
 #include <asm/system.h>
@@ -44,6 +45,52 @@ void do_bad_mode(struct cpu_user_regs *regs, int reason)
     panic("bad mode\n");
 }
 
+void post_increment_register(const struct instr_details *instr)
+{
+    struct cpu_user_regs *regs = guest_cpu_user_regs();
+    register_t val = 0;
+
+    /* Currently, we handle only ldr/str post indexing instructions */
+    if ( instr->state != INSTR_LDR_STR_POSTINDEXING )
+        return;
+
+    /*
+     * Handle when rn = SP
+     * Refer ArmV8 ARM DDI 0487G.b, Page - D1-2463 "Stack pointer register
+     * selection"
+     * t = SP_EL0
+     * h = SP_ELx
+     * and M[3:0] (Page - C5-474 "When exception taken from AArch64 state:")
+     */
+    if (instr->rn == 31 )
+    {
+        if ( (regs->cpsr & PSR_MODE_MASK) == PSR_MODE_EL1h )
+            val = regs->sp_el1;
+        else if ( ((regs->cpsr & PSR_MODE_MASK) == PSR_MODE_EL1t) ||
+                    ((regs->cpsr & PSR_MODE_MASK) == PSR_MODE_EL0t) )
+            val = regs->sp_el0;
+        else
+        {
+            domain_crash(current->domain);
+            return;
+        }
+    }
+    else
+        val = get_user_reg(regs, instr->rn);
+
+    val += instr->imm9;
+
+    if ( instr->rn == 31 )
+    {
+        if ( (regs->cpsr & PSR_MODE_MASK) == PSR_MODE_EL1h )
+            regs->sp_el1 = val;
+        else
+            regs->sp_el0 = val;
+    }
+    else
+        set_user_reg(regs, instr->rn, val);
+}
+
 /*
  * Local variables:
  * mode: C
xen/arch/arm/decode.c +1 -0
index 3add87e83a..16ad0747bb 100644
@@ -146,6 +146,7 @@ static int decode_arm64(register_t pc, mmio_info_t *info)
 
     update_dabt(dabt, opcode.ldr_str.rt, opcode.ldr_str.size, false);
 
+    dabt_instr->state = INSTR_LDR_STR_POSTINDEXING;
     dabt_instr->rn = opcode.ldr_str.rn;
     dabt_instr->imm9 = opcode.ldr_str.imm9;
 
xen/arch/arm/include/asm/domain.h +4 -0
index c56f6e4398..ed63c2b6f9 100644
@@ -281,6 +281,10 @@ static inline void arch_vcpu_block(struct vcpu *v) {}
 /* vPCI is not available on Arm */
 #define has_vpci(d)    ({ (void)(d); false; })
 
+struct arch_vcpu_io {
+    struct instr_details dabt_instr; /* when the instruction is decoded */
+};
+
 #endif /* __ASM_DOMAIN_H__ */
 
 /*
xen/arch/arm/include/asm/mmio.h +15 -2
index 3354d9c635..ef2c57a2d5 100644
@@ -26,12 +26,24 @@
 
 #define MAX_IO_HANDLER  16
 
+enum instr_decode_state
+{
+    INSTR_ERROR,                    /* Error encountered while decoding instr */
+    INSTR_VALID,                    /* ISS is valid, so no need to decode */
+    /*
+     * Instruction is decoded successfully. It is a ldr/str post indexing
+     * instruction.
+     */
+    INSTR_LDR_STR_POSTINDEXING
+};
+
 typedef struct
 {
     struct hsr_dabt dabt;
     struct instr_details {
         unsigned long rn:5;
         signed int imm9:9;
+        enum instr_decode_state state;
     } dabt_instr;
     paddr_t gpa;
 } mmio_info_t;
@@ -69,14 +81,15 @@ struct vmmio {
 };
 
 enum io_state try_handle_mmio(struct cpu_user_regs *regs,
-                              const union hsr hsr,
-                              paddr_t gpa);
+                              mmio_info_t *info);
 void register_mmio_handler(struct domain *d,
                            const struct mmio_handler_ops *ops,
                            paddr_t addr, paddr_t size, void *priv);
 int domain_io_init(struct domain *d, int max_count);
 void domain_io_free(struct domain *d);
 
+void try_decode_instruction(const struct cpu_user_regs *regs,
+                            mmio_info_t *info);
 
 #endif  /* __ASM_ARM_MMIO_H__ */
 
xen/arch/arm/include/asm/traps.h +2 -0
index 2ed2b85c6f..95c46ad391 100644
@@ -109,6 +109,8 @@ static inline register_t sign_extend(const struct hsr_dabt dabt, register_t r)
     return r;
 }
 
+void post_increment_register(const struct instr_details *instr);
+
 #endif /* __ASM_ARM_TRAPS__ */
 /*
  * Local variables:
xen/arch/arm/io.c +56 -34
index fad103bdbd..bea69ffb08 100644
@@ -102,57 +102,79 @@ static const struct mmio_handler *find_mmio_handler(struct domain *d,
     return handler;
 }
 
+void try_decode_instruction(const struct cpu_user_regs *regs,
+                            mmio_info_t *info)
+{
+    int rc;
+
+    if ( info->dabt.valid )
+    {
+        info->dabt_instr.state = INSTR_VALID;
+
+        /*
+         * Erratum 766422: Thumb store translation fault to Hypervisor may
+         * not have correct HSR Rt value.
+         */
+        if ( check_workaround_766422() && (regs->cpsr & PSR_THUMB) &&
+             info->dabt.write )
+        {
+            rc = decode_instruction(regs, info);
+            if ( rc )
+            {
+                gprintk(XENLOG_DEBUG, "Unable to decode instruction\n");
+                info->dabt_instr.state = INSTR_ERROR;
+            }
+        }
+        return;
+    }
+
+    /*
+     * Armv8 processor does not provide a valid syndrome for decoding some
+     * instructions. So in order to process these instructions, Xen must
+     * decode them.
+     */
+    rc = decode_instruction(regs, info);
+    if ( rc )
+    {
+        gprintk(XENLOG_ERR, "Unable to decode instruction\n");
+        info->dabt_instr.state = INSTR_ERROR;
+    }
+}
+
 enum io_state try_handle_mmio(struct cpu_user_regs *regs,
-                              const union hsr hsr,
-                              paddr_t gpa)
+                              mmio_info_t *info)
 {
     struct vcpu *v = current;
     const struct mmio_handler *handler = NULL;
-    const struct hsr_dabt dabt = hsr.dabt;
-    mmio_info_t info = {
-        .gpa = gpa,
-        .dabt = dabt
-    };
+    int rc;
 
-    ASSERT(hsr.ec == HSR_EC_DATA_ABORT_LOWER_EL);
+    ASSERT(info->dabt.ec == HSR_EC_DATA_ABORT_LOWER_EL);
 
-    handler = find_mmio_handler(v->domain, info.gpa);
-    if ( !handler )
+    if ( !((info->dabt_instr.state == INSTR_VALID) || (info->dabt_instr.state == INSTR_LDR_STR_POSTINDEXING)) )
     {
-        int rc;
+        ASSERT_UNREACHABLE();
+        return IO_ABORT;
+    }
 
-        rc = try_fwd_ioserv(regs, v, &info);
+    handler = find_mmio_handler(v->domain, info->gpa);
+    if ( !handler )
+    {
+        rc = try_fwd_ioserv(regs, v, info);
         if ( rc == IO_HANDLED )
             return handle_ioserv(regs, v);
 
         return rc;
     }
 
-    /* All the instructions used on emulated MMIO region should be valid */
-    if ( !dabt.valid )
-        return IO_ABORT;
-
     /*
-     * Erratum 766422: Thumb store translation fault to Hypervisor may
-     * not have correct HSR Rt value.
+     * At this point, we know that the instruction is either valid or has been
+     * decoded successfully. Thus, Xen should be allowed to execute the
+     * instruction on the emulated MMIO region.
      */
-    if ( check_workaround_766422() && (regs->cpsr & PSR_THUMB) &&
-         dabt.write )
-    {
-        int rc;
-
-        rc = decode_instruction(regs, &info);
-        if ( rc )
-        {
-            gprintk(XENLOG_DEBUG, "Unable to decode instruction\n");
-            return IO_ABORT;
-        }
-    }
-
-    if ( info.dabt.write )
-        return handle_write(handler, v, &info);
+    if ( info->dabt.write )
+        return handle_write(handler, v, info);
     else
-        return handle_read(handler, v, &info);
+        return handle_read(handler, v, info);
 }
 
 void register_mmio_handler(struct domain *d,
xen/arch/arm/ioreq.c +1 -0
index 308650b400..58cd320b5a 100644
@@ -47,6 +47,7 @@ enum io_state try_fwd_ioserv(struct cpu_user_regs *regs,
                              struct vcpu *v, mmio_info_t *info)
 {
     struct vcpu_io *vio = &v->io;
+    struct dabt_instr instr = info->dabt_instr;
     ioreq_t p = {
         .type = IOREQ_TYPE_COPY,
         .addr = info->gpa,

xen/arch/arm/ioreq.c enum io_state try_fwd_ioserv(struct cpu_user_regs *regs, · line 80

Quoted 6 lines
xen/arch/arm/ioreq.c +1 -2
@@ -76,10 +77,10 @@ enum io_state try_fwd_ioserv(struct cpu_user_regs *regs,
     if ( !s )
         return IO_UNHANDLED;
 
-    if ( !info->dabt.valid )
-        return IO_ABORT;
+    ASSERT(dabt.valid);

I cannot see where we set dabt.valid on successfully decoding the instruction. It looks like we don't? If we don't, then here the ASSERT would fail in case of postindexing instructions, right?

If we don't, then we should probably just get rid of this ASSERT: it is not worth setting dabt.valid just so that this ASSERT would succeed.

#5 lore raw

Re: [XEN v9 2/4] xen/arm64: io: Support instructions (for which ISS is not valid) on emulated MMIO region using MMIO/ioreq handler

Hi Stefano,

Quoted 17 lines · Stefano Stabellini
Quoted 17 lines

void register_mmio_handler(struct domain *d,

xen/arch/arm/ioreq.c +1 -0
index 308650b400..58cd320b5a 100644
@@ -47,6 +47,7 @@ enum io_state try_fwd_ioserv(struct cpu_user_regs *regs,
                               struct vcpu *v, mmio_info_t *info)
  {
      struct vcpu_io *vio = &v->io;
+    struct dabt_instr instr = info->dabt_instr;
      ioreq_t p = {
          .type = IOREQ_TYPE_COPY,
          .addr = info->gpa,
@@ -76,10 +77,10 @@ enum io_state try_fwd_ioserv(struct cpu_user_regs *regs,
      if ( !s )
          return IO_UNHANDLED;
 
Quoted 7 lines

xen/arch/arm/ioreq.c enum io_state try_fwd_ioserv(struct cpu_user_regs *regs, · line 80

Quoted 3 lines
xen/arch/arm/ioreq.c +1 -2
-    if ( !info->dabt.valid )
-        return IO_ABORT;
+    ASSERT(dabt.valid);

I cannot see where we set dabt.valid on successfully decoding the instruction. It looks like we don't? If we don't, then here the ASSERT would fail in case of postindexing instructions, right?

We don't currently set dabt.valid. There are other reasons to set it (see my reply to Ayan). So...

Quoted 3 lines

If we don't, then we should probably just get rid of this ASSERT: it is not worth setting dabt.valid just so that this ASSERT would succeed.

... I would keep the ASSERT.

Cheers,

Signature · 1 line
Julien Grall
#6 lore raw

Re: [XEN v9 2/4] xen/arm64: io: Support instructions (for which ISS is not valid) on emulated MMIO region using MMIO/ioreq handler

Quoted 28 lines · Julien Grall

Hi Stefano,

Quoted 16 lines · Stefano Stabellini
Quoted 16 lines

void register_mmio_handler(struct domain *d,

xen/arch/arm/ioreq.c +1 -0
index 308650b400..58cd320b5a 100644
@@ -47,6 +47,7 @@ enum io_state try_fwd_ioserv(struct cpu_user_regs *regs,
                               struct vcpu *v, mmio_info_t *info)
  {
      struct vcpu_io *vio = &v->io;
+    struct dabt_instr instr = info->dabt_instr;
      ioreq_t p = {
          .type = IOREQ_TYPE_COPY,
          .addr = info->gpa,
@@ -76,10 +77,10 @@ enum io_state try_fwd_ioserv(struct cpu_user_regs
 *regs,
      if ( !s )
Quoted 8 lines

xen/arch/arm/ioreq.c enum io_state try_fwd_ioserv(struct cpu_user_regs *regs, · line 80

Quoted 4 lines
xen/arch/arm/ioreq.c +1 -1
          return IO_UNHANDLED;
  -    if ( !info->dabt.valid )
-        return IO_ABORT;
+    ASSERT(dabt.valid);

I cannot see where we set dabt.valid on successfully decoding the instruction. It looks like we don't? If we don't, then here the ASSERT would fail in case of postindexing instructions, right?

Quoted 8 lines

We don't currently set dabt.valid. There are other reasons to set it (see my reply to Ayan). So...

Quoted 3 lines

If we don't, then we should probably just get rid of this ASSERT: it is not worth setting dabt.valid just so that this ASSERT would succeed.

... I would keep the ASSERT.

OK

#7 lore raw

Re: [XEN v9 2/4] xen/arm64: io: Support instructions (for which ISS is not valid) on emulated MMIO region using MMIO/ioreq handler

In reply to Ayan Kumar Halder

Hi Ayan,

Quoted 13 lines · Ayan Kumar Halder
xen/arch/arm/arm64/traps.c +13 -0
+void post_increment_register(const struct instr_details *instr)
+{
+    struct cpu_user_regs *regs = guest_cpu_user_regs();
+    register_t val = 0;
+
+    /* Currently, we handle only ldr/str post indexing instructions */
+    if ( instr->state != INSTR_LDR_STR_POSTINDEXING )
+        return;
+
+    /*
+     * Handle when rn = SP
+     * Refer ArmV8 ARM DDI 0487G.b, Page - D1-2463 "Stack pointer register
+     * selection"

xen/arch/arm/arm64/traps.c void do_bad_mode(struct cpu_user_regs *regs, int reason) · line 70

Quoted 10 lines
xen/arch/arm/arm64/traps.c +10 -0
+     * t = SP_EL0
+     * h = SP_ELx
+     * and M[3:0] (Page - C5-474 "When exception taken from AArch64 state:")
+     */
+    if (instr->rn == 31 )
+    {
+        if ( (regs->cpsr & PSR_MODE_MASK) == PSR_MODE_EL1h )
+            val = regs->sp_el1;
+        else if ( ((regs->cpsr & PSR_MODE_MASK) == PSR_MODE_EL1t) ||
+                    ((regs->cpsr & PSR_MODE_MASK) == PSR_MODE_EL0t) )

You are using 3 times regs->cpsr & PSR_MODE_MASK. Can you introduce a temporary variable?

Alternatively, a switch could be used here.

[...]

Quoted 7 lines
xen/arch/arm/include/asm/mmio.h +0 -0
index 3354d9c635..ef2c57a2d5 100644
@@ -26,12 +26,24 @@
 
  #define MAX_IO_HANDLER  16

xen/arch/arm/include/asm/mmio.h line 37

Quoted 10 lines
xen/arch/arm/include/asm/mmio.h +9 -0
 
+enum instr_decode_state
+{
+    INSTR_ERROR,                    /* Error encountered while decoding instr */
+    INSTR_VALID,                    /* ISS is valid, so no need to decode */
+    /*
+     * Instruction is decoded successfully. It is a ldr/str post indexing
+     * instruction.
+     */
+    INSTR_LDR_STR_POSTINDEXING

NIT: Please add ',' even for the last item. This would reduce the diff if we add new one.

Quoted 56 lines
xen/arch/arm/include/asm/mmio.h +6 -2
+};
+
  typedef struct
  {
      struct hsr_dabt dabt;
      struct instr_details {
          unsigned long rn:5;
          signed int imm9:9;
+        enum instr_decode_state state;
      } dabt_instr;
      paddr_t gpa;
  } mmio_info_t;
@@ -69,14 +81,15 @@ struct vmmio {
  };
 
  enum io_state try_handle_mmio(struct cpu_user_regs *regs,
-                              const union hsr hsr,
-                              paddr_t gpa);
+                              mmio_info_t *info);
  void register_mmio_handler(struct domain *d,
                             const struct mmio_handler_ops *ops,
                             paddr_t addr, paddr_t size, void *priv);
  int domain_io_init(struct domain *d, int max_count);
  void domain_io_free(struct domain *d);
 
+void try_decode_instruction(const struct cpu_user_regs *regs,
+                            mmio_info_t *info);
 
  #endif  /* __ASM_ARM_MMIO_H__ */
 
xen/arch/arm/include/asm/traps.h +2 -0
index 2ed2b85c6f..95c46ad391 100644
@@ -109,6 +109,8 @@ static inline register_t sign_extend(const struct hsr_dabt dabt, register_t r)
      return r;
  }
 
+void post_increment_register(const struct instr_details *instr);
+
  #endif /* __ASM_ARM_TRAPS__ */
  /*
   * Local variables:
xen/arch/arm/io.c +5 -0
index fad103bdbd..bea69ffb08 100644
@@ -102,57 +102,79 @@ static const struct mmio_handler *find_mmio_handler(struct domain *d,
      return handler;
  }
 
+void try_decode_instruction(const struct cpu_user_regs *regs,
+                            mmio_info_t *info)
+{
+    int rc;
+

xen/arch/arm/io.c static const struct mmio_handler *find_mmio_handler(struct domain *d, · line 119

Quoted 10 lines
xen/arch/arm/io.c +10 -0
+    if ( info->dabt.valid )
+    {
+        info->dabt_instr.state = INSTR_VALID;
+
+        /*
+         * Erratum 766422: Thumb store translation fault to Hypervisor may
+         * not have correct HSR Rt value.
+         */
+        if ( check_workaround_766422() && (regs->cpsr & PSR_THUMB) &&
+             info->dabt.write )

This change is not explained in the commit message. TBH, I think it should be separate but I am not going to request that at v9.

Quoted 33 lines
xen/arch/arm/io.c +24 -5
+        {
+            rc = decode_instruction(regs, info);
+            if ( rc )
+            {
+                gprintk(XENLOG_DEBUG, "Unable to decode instruction\n");
+                info->dabt_instr.state = INSTR_ERROR;
+        }
+        return;
+    }
+
+    /*
+     * Armv8 processor does not provide a valid syndrome for decoding some
+     * instructions. So in order to process these instructions, Xen must
+     * decode them.
+     */
+    rc = decode_instruction(regs, info);
+    if ( rc )
+    {
+        gprintk(XENLOG_ERR, "Unable to decode instruction\n");
+        info->dabt_instr.state = INSTR_ERROR;
+    }
+}
+
  enum io_state try_handle_mmio(struct cpu_user_regs *regs,
-                              const union hsr hsr,
-                              paddr_t gpa)
+                              mmio_info_t *info)
  {
      struct vcpu *v = current;
      const struct mmio_handler *handler = NULL;
-    const struct hsr_dabt dabt = hsr.dabt;
-    mmio_info_t info = {
-        .gpa = gpa,

xen/arch/arm/io.c static const struct mmio_handler *find_mmio_handler(struct domain *d, · line 153

Quoted 10 lines
xen/arch/arm/io.c +3 -5
-        .dabt = dabt
-    };
+    int rc;
 
-    ASSERT(hsr.ec == HSR_EC_DATA_ABORT_LOWER_EL);
+    ASSERT(info->dabt.ec == HSR_EC_DATA_ABORT_LOWER_EL);
 
-    handler = find_mmio_handler(v->domain, info.gpa);
-    if ( !handler )
+    if ( !((info->dabt_instr.state == INSTR_VALID) || (info->dabt_instr.state == INSTR_LDR_STR_POSTINDEXING)) )

This check will become quite large if we decode more class. I would instead set the dabt.valid bit whenever we successfully decoded the instruction and check that if dabt.valid here.

Quoted 5 lines
xen/arch/arm/io.c +3 -1
      {
-        int rc;
+        ASSERT_UNREACHABLE();
+        return IO_ABORT;
+    }

[...]

Quoted 10 lines
xen/arch/arm/io.c +0 -3
@@ -1982,21 +2030,18 @@ static void do_trap_stage2_abort_guest(struct cpu_user_regs *regs,
              case IO_UNHANDLED:
                  /* IO unhandled, try another way to handle it. */
                  break;
-            }
          }
 
          /*
-         * First check if the translation fault can be resolved by the
-         * P2M subsystem. If that's the case nothing else to do.

xen/arch/arm/io.c static void do_trap_stage2_abort_guest(struct cpu_user_regs *regs, · line 2040

Quoted 10 lines
xen/arch/arm/io.c +4 -5
+         * If the instruction syndrome was invalid, then we already checked if
+         * this was due to a P2M fault. So no point to check again as the result
+         * will be the same.
           */
-        if ( p2m_resolve_translation_fault(current->domain,
-                                           gaddr_to_gfn(gpa)) )
-            return;
-
-        if ( is_data && try_map_mmio(gaddr_to_gfn(gpa)) )
+        if ( info.dabt.valid && check_p2m(is_data, gpa) )

This check would need to be adjusted to check the instruction state is INSTR_VALID.

Cheers,

Signature · 1 line
Julien Grall
#8 lore raw

Re: [XEN v9 2/4] xen/arm64: io: Support instructions (for which ISS is not valid) on emulated MMIO region using MMIO/ioreq handler

Hi Julien,

Many thanks for the feedback.

I have some clarifications.

Quoted 31 lines · Julien Grall

Hi Ayan,

Quoted 17 lines · Ayan Kumar Halder
xen/arch/arm/arm64/traps.c +12 -0
+void post_increment_register(const struct instr_details *instr)
+{
+    struct cpu_user_regs *regs = guest_cpu_user_regs();
+    register_t val = 0;
+
+    /* Currently, we handle only ldr/str post indexing instructions */
+    if ( instr->state != INSTR_LDR_STR_POSTINDEXING )
+        return;
+
+    /*
+     * Handle when rn = SP
+     * Refer ArmV8 ARM DDI 0487G.b, Page - D1-2463 "Stack pointer 

register
+     * selection"
+     * t = SP_EL0
+     * h = SP_ELx
+     * and M[3:0] (Page - C5-474 "When exception taken from AArch64

Quoted 8 lines

state:")
+     */
+    if (instr->rn == 31 )
+    {
+        if ( (regs->cpsr & PSR_MODE_MASK) == PSR_MODE_EL1h ) +            val = regs->sp_el1;
+        else if ( ((regs->cpsr & PSR_MODE_MASK) == PSR_MODE_EL1t) || +                    ((regs->cpsr & PSR_MODE_MASK) == PSR_MODE_EL0t) )

You are using 3 times regs->cpsr & PSR_MODE_MASK. Can you introduce a temporary variable?

Quoted 1 line

Alternatively, a switch could be used here.

Yes, a switch is better. I will address that in v10.

Quoted 15 lines

[...]

Quoted 12 lines
xen/arch/arm/include/asm/mmio.h +0 -0
index 3354d9c635..ef2c57a2d5 100644
@@ -26,12 +26,24 @@
     #define MAX_IO_HANDLER  16

  +enum instr_decode_state
+{
+    INSTR_ERROR,                    /* Error encountered while decoding instr */
+    INSTR_VALID,                    /* ISS is valid, so no need to decode */

Quoted 8 lines
Quoted 6 lines

+    /*
+     * Instruction is decoded successfully. It is a ldr/str post indexing
+     * instruction.
+     */
+    INSTR_LDR_STR_POSTINDEXING

NIT: Please add ',' even for the last item. This would reduce the diff if we add new one.

Ack (To be addressed in v10)

Quoted 58 lines
Quoted 58 lines
xen/arch/arm/include/asm/mmio.h +2 -0
+};
+

  typedef struct
  {
      struct hsr_dabt dabt;
      struct instr_details {
          unsigned long rn:5;
          signed int imm9:9;
+        enum instr_decode_state state;
      } dabt_instr;
      paddr_t gpa;
  } mmio_info_t;

xen/arch/arm/include/asm/mmio.h +2 -2
@@ -69,14 +81,15 @@ struct vmmio {
   };
     enum io_state try_handle_mmio(struct cpu_user_regs *regs,
-                              const union hsr hsr,
-                              paddr_t gpa);
+                              mmio_info_t *info);
   void register_mmio_handler(struct domain *d,
                              const struct mmio_handler_ops *ops,
                              paddr_t addr, paddr_t size, void *priv);
   int domain_io_init(struct domain *d, int max_count);
   void domain_io_free(struct domain *d);
   +void try_decode_instruction(const struct cpu_user_regs *regs,
+                            mmio_info_t *info);
     #endif  /* __ASM_ARM_MMIO_H__ */

  diff --git a/xen/arch/arm/include/asm/traps.h b/xen/arch/arm/include/asm/traps.h index 2ed2b85c6f..95c46ad391 100644

xen/arch/arm/include/asm/traps.h +1 -0
@@ -109,6 +109,8 @@ static inline register_t sign_extend(const struct 
 hsr_dabt dabt, register_t r)
       return r;
   }
   +void post_increment_register(const struct instr_details *instr);
+
   #endif /* __ASM_ARM_TRAPS__ */
   /*

   * Local variables:

xen/arch/arm/io.c +9 -0
index fad103bdbd..bea69ffb08 100644
@@ -102,57 +102,79 @@ static const struct mmio_handler 
 *find_mmio_handler(struct domain *d,
       return handler;
   }
   +void try_decode_instruction(const struct cpu_user_regs *regs,
+                            mmio_info_t *info)
+{
+    int rc;
+
+    if ( info->dabt.valid )
+    {
+        info->dabt_instr.state = INSTR_VALID;
+
+        /*
Quoted 8 lines
Quoted 6 lines
xen/arch/arm/io.c +1 -0
+         * Erratum 766422: Thumb store translation fault to 

Hypervisor may
+         * not have correct HSR Rt value.
+         */
+        if ( check_workaround_766422() && (regs->cpsr & PSR_THUMB) && +             info->dabt.write )

This change is not explained in the commit message. TBH, I think it should be separate but I am not going to request that at v9.

I will leave this as it is then. I will explain in the commit message that the logic to infer the type of instruction has been moved from try_handle_mmio() to try_decode_instruction() which is called before. try_handle_mmio() is solely responsible for handling the mmio operation.

Quoted 40 lines
Quoted 40 lines

+        {
+            rc = decode_instruction(regs, info);
+            if ( rc )
+            {
+                gprintk(XENLOG_DEBUG, "Unable to decode instruction\n");
+                info->dabt_instr.state = INSTR_ERROR;
+        }
+        return;
+    }
+
+    /*
+     * Armv8 processor does not provide a valid syndrome for decoding some
+     * instructions. So in order to process these instructions, Xen must
+     * decode them.
+     */
+    rc = decode_instruction(regs, info);
+    if ( rc )
+    {
+        gprintk(XENLOG_ERR, "Unable to decode instruction\n"); +        info->dabt_instr.state = INSTR_ERROR;
+    }

xen/arch/arm/io.c +2 -0
+}
+

  enum io_state try_handle_mmio(struct cpu_user_regs *regs, -                              const union hsr hsr, -                              paddr_t gpa)
+                              mmio_info_t *info) {
      struct vcpu *v = current;
      const struct mmio_handler *handler = NULL;
-    const struct hsr_dabt dabt = hsr.dabt;
-    mmio_info_t info = {
-        .gpa = gpa,
-        .dabt = dabt
-    };
+    int rc;

Quoted 8 lines
Quoted 6 lines

  -    ASSERT(hsr.ec == HSR_EC_DATA_ABORT_LOWER_EL); +    ASSERT(info->dabt.ec == HSR_EC_DATA_ABORT_LOWER_EL); -    handler = find_mmio_handler(v->domain, info.gpa); -    if ( !handler )
+    if ( !((info->dabt_instr.state == INSTR_VALID) || (info->dabt_instr.state == INSTR_LDR_STR_POSTINDEXING)) )

This check will become quite large if we decode more class. I would instead set the dabt.valid bit whenever we successfully decoded the instruction and check that if dabt.valid here.

Actually the main reason to introduce INSTR_LDR_STR_POSTINDEXING is to distinguish the scenario where the ISS was valid vs when instruction was decoded manually.

In the later scenario, one would need to do the post increment of the rn.

It makes sense to me to have a unque 'info->dabt_instr.state' for each type of instruction decoded as the post processing will vary. In this case, the post processing logic checks that the instruction is ldr_str_postindexing.

However your concern that the check will become large is valid. I would introduce a function as follows :-

bool check_instr_is_valid(enum instr_decode_state state)

{

    if (state == INSTR_VALID) || (state == INSTR_LDR_STR_POSTINDEXING) || ...)

        return true;

    else

        return false;

}

And then in

enum io_state try_handle_mmio(struct cpu_user_regs *regs, ...)

{

...

    if ( !check_instr_is_valid(info->dabt_instr.state) )

    {

        ASSERT_UNREACHABLE();
        return IO_ABORT;

    }

...

}

Please let me know your thoughts,

Quoted 25 lines
Quoted 6 lines

      {
-        int rc;
+        ASSERT_UNREACHABLE();
+        return IO_ABORT;
+    }

[...]

Quoted 16 lines
xen/arch/arm/io.c +0 -0
@@ -1982,21 +2030,18 @@ static void do_trap_stage2_abort_guest(struct 
 cpu_user_regs *regs,

              case IO_UNHANDLED:
                  /* IO unhandled, try another way to handle it. */ break;
-            }
          }
            /*
-         * First check if the translation fault can be resolved by the -         * P2M subsystem. If that's the case nothing else to do. +         * If the instruction syndrome was invalid, then we already checked if
+         * this was due to a P2M fault. So no point to check again as the result
+         * will be the same.
           */

Quoted 8 lines
Quoted 6 lines

-        if ( p2m_resolve_translation_fault(current->domain, -                                           gaddr_to_gfn(gpa)) ) -            return;
-
-        if ( is_data && try_map_mmio(gaddr_to_gfn(gpa)) ) +        if ( info.dabt.valid && check_p2m(is_data, gpa) )

This check would need to be adjusted to check the instruction state is INSTR_VALID.

Ack (To be addressed in v10).

- Ayan

Quoted 3 lines

Cheers,

#9 lore raw

Re: [XEN v9 2/4] xen/arm64: io: Support instructions (for which ISS is not valid) on emulated MMIO region using MMIO/ioreq handler

Hi,

Quoted 49 lines · Ayan Kumar Halder
Quoted 40 lines
Quoted 40 lines

+        {
+            rc = decode_instruction(regs, info);
+            if ( rc )
+            {
+                gprintk(XENLOG_DEBUG, "Unable to decode instruction\n");
+                info->dabt_instr.state = INSTR_ERROR;
+        }
+        return;
+    }
+
+    /*
+     * Armv8 processor does not provide a valid syndrome for decoding some
+     * instructions. So in order to process these instructions, Xen must
+     * decode them.
+     */
+    rc = decode_instruction(regs, info);
+    if ( rc )
+    {
+        gprintk(XENLOG_ERR, "Unable to decode instruction\n"); +        info->dabt_instr.state = INSTR_ERROR;
+    }

(quoted diff) +2 -0
+}
+

  enum io_state try_handle_mmio(struct cpu_user_regs *regs, -                              const union hsr hsr, -                              paddr_t gpa)
+                              mmio_info_t *info) {
      struct vcpu *v = current;
      const struct mmio_handler *handler = NULL;
-    const struct hsr_dabt dabt = hsr.dabt;
-    mmio_info_t info = {
-        .gpa = gpa,
-        .dabt = dabt
-    };
+    int rc;

Quoted 8 lines
Quoted 6 lines

  -    ASSERT(hsr.ec == HSR_EC_DATA_ABORT_LOWER_EL); +    ASSERT(info->dabt.ec == HSR_EC_DATA_ABORT_LOWER_EL); -    handler = find_mmio_handler(v->domain, info.gpa); -    if ( !handler )
+    if ( !((info->dabt_instr.state == INSTR_VALID) || (info->dabt_instr.state == INSTR_LDR_STR_POSTINDEXING)) )

This check will become quite large if we decode more class. I would instead set the dabt.valid bit whenever we successfully decoded the instruction and check that if dabt.valid here.

Quoted 5 lines

Actually the main reason to introduce INSTR_LDR_STR_POSTINDEXING is to distinguish the scenario where the ISS was valid vs when instruction was decoded manually.

In the later scenario, one would need to do the post increment of the rn.

It makes sense to me to have a unque 'info->dabt_instr.state' for each type of instruction decoded as the post processing will vary. In this case, the post processing logic checks that the instruction is ldr_str_postindexing.

So I agree we want to have a unique state for type of instruction. I wasn't suggesting to remove it. Instead, I was suggesting to use dabt.valid as "This is a valid instruction for accessing an emulated MMIO".

Quoted 33 lines

However your concern that the check will become large is valid. I would introduce a function as follows :-

bool check_instr_is_valid(enum instr_decode_state state)

{

    if (state == INSTR_VALID) || (state == INSTR_LDR_STR_POSTINDEXING)
|| ...)

        return true;

    else

        return false;

}

And then in

enum io_state try_handle_mmio(struct cpu_user_regs *regs, ...)

{

...

    if ( !check_instr_is_valid(info->dabt_instr.state) )

    {

        ASSERT_UNREACHABLE();
        return IO_ABORT;

Quoted 7 lines

    }

...

}

Please let me know your thoughts,

This is only moving the check to a separate function. It doesn't help with the fact that the check in check_instr_is_valid() is going to grow.

I can see two options:
* Using dabt.valid as "The instruction was fully decoded".
* Check that the state is not INSTR_ERROR

Above, I was suggesting the former. But I am open to use latter.

Cheers,

Signature · 1 line
Julien Grall
#10 lore raw

PATCH v9 3/4 xen/arm64: io: Handle the abort due to access to stage1 translation table

In reply to Ayan Kumar Halder

If the abort was caused due to access to stage1 translation table, Xen will assume that the stage1 translation table is in the non MMIO region. It will try to resolve the translation fault. If it succeeds, it will return to the guest to retry the instruction. If not, then it means that the table is in MMIO region which is not expected by Xen. Thus, Xen will forward the abort to the guest.

---

Changelog :-

v1..v8 - NA

v9 - 1. Extracted this change from "[XEN v8 2/2] xen/arm64: io: Support instructions (for which ISS is not..." into a separate patch of its own. The reason being this is an existing bug in the codebase.

 xen/arch/arm/io.c    | 11 +++++++++++
 xen/arch/arm/traps.c | 12 +++++++++++-
 2 files changed, 22 insertions(+), 1 deletion(-)
xen/arch/arm/io.c +11 -0
index bea69ffb08..ebcb8ed548 100644
@@ -128,6 +128,17 @@ void try_decode_instruction(const struct cpu_user_regs *regs,
         return;
     }
 
+    /*
+     * At this point, we know that the stage1 translation table is in the MMIO
+     * region. This is not expected by Xen and thus it forwards the abort to the
+     * guest.
+     */
+    if ( info->dabt.s1ptw )
+    {
+        info->dabt_instr.state = INSTR_ERROR;
+        return;
+    }
+
     /*
      * Armv8 processor does not provide a valid syndrome for decoding some
      * instructions. So in order to process these instructions, Xen must
xen/arch/arm/traps.c +11 -1
index 120c971b0f..e491ca15d7 100644
@@ -1923,6 +1923,7 @@ static void do_trap_stage2_abort_guest(struct cpu_user_regs *regs,
     bool is_data = (hsr.ec == HSR_EC_DATA_ABORT_LOWER_EL);
     mmio_info_t info;
     enum io_state state;
+    bool check_mmio_region = true;
 
     /*
      * If this bit has been set, it means that this stage-2 abort is caused
@@ -1987,7 +1988,16 @@ static void do_trap_stage2_abort_guest(struct cpu_user_regs *regs,
          */
         if ( !is_data || !info.dabt.valid )
         {
-            if ( check_p2m(is_data, gpa) )
+            /*
+             * If the translation fault was caused due to access to stage 1
+             * translation table, then we try to set the translation table entry
+             * for page1 translation table (assuming that it is in the non mmio
+             * region).
+             */
+            if ( xabt.s1ptw )
+                check_mmio_region = false;
+
+            if ( check_p2m((is_data && check_mmio_region), gpa) )
                 return;
 
             /*
Signature · 1 line
2.17.1
#11 lore raw

Re: [XEN v9 3/4] xen/arm64: io: Handle the abort due to access to stage1 translation table

Quoted 57 lines · Ayan Kumar Halder

If the abort was caused due to access to stage1 translation table, Xen will assume that the stage1 translation table is in the non MMIO region. It will try to resolve the translation fault. If it succeeds, it will return to the guest to retry the instruction. If not, then it means that the table is in MMIO region which is not expected by Xen. Thus, Xen will forward the abort to the guest.

---

Changelog :-

v1..v8 - NA

v9 - 1. Extracted this change from "[XEN v8 2/2] xen/arm64: io: Support instructions (for which ISS is not..." into a separate patch of its own. The reason being this is an existing bug in the codebase.

 xen/arch/arm/io.c    | 11 +++++++++++
 xen/arch/arm/traps.c | 12 +++++++++++-
 2 files changed, 22 insertions(+), 1 deletion(-)
xen/arch/arm/io.c +11 -0
index bea69ffb08..ebcb8ed548 100644
@@ -128,6 +128,17 @@ void try_decode_instruction(const struct cpu_user_regs *regs,
         return;
     }
 
+    /*
+     * At this point, we know that the stage1 translation table is in the MMIO
+     * region. This is not expected by Xen and thus it forwards the abort to the
+     * guest.
+     */
+    if ( info->dabt.s1ptw )
+    {
+        info->dabt_instr.state = INSTR_ERROR;
+        return;
+    }
+
     /*
      * Armv8 processor does not provide a valid syndrome for decoding some
      * instructions. So in order to process these instructions, Xen must
xen/arch/arm/traps.c +1 -0
index 120c971b0f..e491ca15d7 100644
@@ -1923,6 +1923,7 @@ static void do_trap_stage2_abort_guest(struct cpu_user_regs *regs,
     bool is_data = (hsr.ec == HSR_EC_DATA_ABORT_LOWER_EL);
     mmio_info_t info;
     enum io_state state;
+    bool check_mmio_region = true;
 
     /*
      * If this bit has been set, it means that this stage-2 abort is caused

xen/arch/arm/traps.c static void do_trap_stage2_abort_guest(struct cpu_user_regs *regs, · line 1994

Quoted 8 lines
xen/arch/arm/traps.c +4 -1
@@ -1987,7 +1988,16 @@ static void do_trap_stage2_abort_guest(struct cpu_user_regs *regs,
          */
         if ( !is_data || !info.dabt.valid )
         {
-            if ( check_p2m(is_data, gpa) )
+            /*
+             * If the translation fault was caused due to access to stage 1
+             * translation table, then we try to set the translation table entry
+             * for page1 translation table (assuming that it is in the non mmio

^ stage1

Do you mean to say maybe:

If the translation fault was caused by an access to stage 1 translation table, then no need to change the stage 2 p2m.

?

Quoted 12 lines
xen/arch/arm/traps.c +6 -0
+             * region).
+             */
+            if ( xabt.s1ptw )
+                check_mmio_region = false;
+
+            if ( check_p2m((is_data && check_mmio_region), gpa) )
                 return;
 
             /*
Signature · 1 line
2.17.1
#12 lore raw

Re: [XEN v9 3/4] xen/arm64: io: Handle the abort due to access to stage1 translation table

Hi Stefano,

Quoted 68 lines · Stefano Stabellini
Quoted 57 lines · Ayan Kumar Halder

If the abort was caused due to access to stage1 translation table, Xen will assume that the stage1 translation table is in the non MMIO region. It will try to resolve the translation fault. If it succeeds, it will return to the guest to retry the instruction. If not, then it means that the table is in MMIO region which is not expected by Xen. Thus, Xen will forward the abort to the guest.

---

Changelog :-

v1..v8 - NA

v9 - 1. Extracted this change from "[XEN v8 2/2] xen/arm64: io: Support instructions (for which ISS is not..." into a separate patch of its own. The reason being this is an existing bug in the codebase.

  xen/arch/arm/io.c    | 11 +++++++++++
  xen/arch/arm/traps.c | 12 +++++++++++-
  2 files changed, 22 insertions(+), 1 deletion(-)
xen/arch/arm/io.c +11 -0
index bea69ffb08..ebcb8ed548 100644
@@ -128,6 +128,17 @@ void try_decode_instruction(const struct cpu_user_regs *regs,
          return;
      }
 
+    /*
+     * At this point, we know that the stage1 translation table is in the MMIO
+     * region. This is not expected by Xen and thus it forwards the abort to the
+     * guest.
+     */
+    if ( info->dabt.s1ptw )
+    {
+        info->dabt_instr.state = INSTR_ERROR;
+        return;
+    }
+
      /*
       * Armv8 processor does not provide a valid syndrome for decoding some
       * instructions. So in order to process these instructions, Xen must
xen/arch/arm/traps.c +1 -0
index 120c971b0f..e491ca15d7 100644
@@ -1923,6 +1923,7 @@ static void do_trap_stage2_abort_guest(struct cpu_user_regs *regs,
      bool is_data = (hsr.ec == HSR_EC_DATA_ABORT_LOWER_EL);
      mmio_info_t info;
      enum io_state state;
+    bool check_mmio_region = true;
 
      /*
       * If this bit has been set, it means that this stage-2 abort is caused

xen/arch/arm/traps.c static void do_trap_stage2_abort_guest(struct cpu_user_regs *regs, · line 1994

Quoted 8 lines
xen/arch/arm/traps.c +4 -1
@@ -1987,7 +1988,16 @@ static void do_trap_stage2_abort_guest(struct cpu_user_regs *regs,
           */
          if ( !is_data || !info.dabt.valid )
          {
-            if ( check_p2m(is_data, gpa) )
+            /*
+             * If the translation fault was caused due to access to stage 1
+             * translation table, then we try to set the translation table entry
+             * for page1 translation table (assuming that it is in the non mmio

^ stage1

Quoted 1 line

Do you mean to say maybe:

Yes, it should be stage1. Sorry for typo.

Quoted 5 lines

If the translation fault was caused by an access to stage 1 translation table, then no need to change the stage 2 p2m.

?

The translation fault was caused due to access to stage1 translation table. As per my understanding, the address of stage1 tables is in stage2 translation table entries. Thus, Xen needs to modify the corresponding stage2 p2m entries.

- Ayan

Quoted 15 lines
Quoted 15 lines
xen/arch/arm/traps.c +6 -0
+             * region).
+             */
+            if ( xabt.s1ptw )
+                check_mmio_region = false;
+
+            if ( check_p2m((is_data && check_mmio_region), gpa) )
                  return;
 
              /*
Signature · 1 line
2.17.1
#13 lore raw

Re: [XEN v9 3/4] xen/arm64: io: Handle the abort due to access to stage1 translation table

Quoted 80 lines · Ayan Kumar Halder
Quoted 76 lines · Stefano Stabellini
Quoted 68 lines · Ayan Kumar Halder

If the abort was caused due to access to stage1 translation table, Xen will assume that the stage1 translation table is in the non MMIO region. It will try to resolve the translation fault. If it succeeds, it will return to the guest to retry the instruction. If not, then it means that the table is in MMIO region which is not expected by Xen. Thus, Xen will forward the abort to the guest.

---

Changelog :-

v1..v8 - NA

v9 - 1. Extracted this change from "[XEN v8 2/2] xen/arm64: io: Support instructions (for which ISS is not..." into a separate patch of its own. The reason being this is an existing bug in the codebase.

  xen/arch/arm/io.c    | 11 +++++++++++
  xen/arch/arm/traps.c | 12 +++++++++++-
  2 files changed, 22 insertions(+), 1 deletion(-)
xen/arch/arm/io.c +1 -0
index bea69ffb08..ebcb8ed548 100644
@@ -128,6 +128,17 @@ void try_decode_instruction(const struct
 cpu_user_regs *regs,
          return;
      }
  +    /*
+     * At this point, we know that the stage1 translation table is in the

MMIO
+ * region. This is not expected by Xen and thus it forwards the abort
to the

xen/arch/arm/io.c +8 -0
+     * guest.
+     */
+    if ( info->dabt.s1ptw )
+    {
+        info->dabt_instr.state = INSTR_ERROR;
+        return;
+    }
+
      /*
       * Armv8 processor does not provide a valid syndrome for decoding

some
* instructions. So in order to process these instructions, Xen must

xen/arch/arm/traps.c +1 -0
index 120c971b0f..e491ca15d7 100644
@@ -1923,6 +1923,7 @@ static void do_trap_stage2_abort_guest(struct
 cpu_user_regs *regs,
      bool is_data = (hsr.ec == HSR_EC_DATA_ABORT_LOWER_EL);
      mmio_info_t info;
      enum io_state state;
+    bool check_mmio_region = true;
        /*
       * If this bit has been set, it means that this stage-2 abort is

caused

xen/arch/arm/traps.c +2 -1
@@ -1987,7 +1988,16 @@ static void do_trap_stage2_abort_guest(struct
 cpu_user_regs *regs,
           */
          if ( !is_data || !info.dabt.valid )
          {
-            if ( check_p2m(is_data, gpa) )
+            /*
+             * If the translation fault was caused due to access to stage
Quoted 5 lines

1
+ * translation table, then we try to set the translation
table entry
+ * for page1 translation table (assuming that it is in the
non mmio

^ stage1

Quoted 1 line

Do you mean to say maybe:

Yes, it should be stage1. Sorry for typo.

Quoted 1 line
Quoted 8 lines
Quoted 4 lines

If the translation fault was caused by an access to stage 1 translation table, then no need to change the stage 2 p2m.

?

The translation fault was caused due to access to stage1 translation table. As per my understanding, the address of stage1 tables is in stage2 translation table entries. Thus, Xen needs to modify the corresponding stage2 p2m entries.

OK, I follow what you are saying and what this patch is doing now. I suggest:

If the translation fault was caused due to access to the stage 1 translation table, then we try to set the p2m entry for the stage 1 translation table, but we don't handle stage 1 translation tables in MMIO regions.

#14 lore raw

Re: [XEN v9 3/4] xen/arm64: io: Handle the abort due to access to stage1 translation table

In reply to Ayan Kumar Halder

Hi Ayan,

Quoted 27 lines · Ayan Kumar Halder

If the abort was caused due to access to stage1 translation table, Xen will assume that the stage1 translation table is in the non MMIO region. It will try to resolve the translation fault. If it succeeds, it will return to the guest to retry the instruction. If not, then it means that the table is in MMIO region which is not expected by Xen. Thus, Xen will forward the abort to the guest.

---

Changelog :-

v1..v8 - NA

v9 - 1. Extracted this change from "[XEN v8 2/2] xen/arm64: io: Support instructions (for which ISS is not..." into a separate patch of its own. The reason being this is an existing bug in the codebase.

  xen/arch/arm/io.c    | 11 +++++++++++
  xen/arch/arm/traps.c | 12 +++++++++++-
  2 files changed, 22 insertions(+), 1 deletion(-)
xen/arch/arm/io.c +0 -0
index bea69ffb08..ebcb8ed548 100644

xen/arch/arm/io.c void try_decode_instruction(const struct cpu_user_regs *regs, · line 133

Quoted 6 lines
xen/arch/arm/io.c +3 -0
@@ -128,6 +128,17 @@ void try_decode_instruction(const struct cpu_user_regs *regs,
          return;
      }
 
+    /*
+     * At this point, we know that the stage1 translation table is in the MMIO
+     * region. This is not expected by Xen and thus it forwards the abort to the

We don't know that. We only know that there are no corresponding valid mapping in the P2M. So the address may be part of an emulated MMIO region or invalid.

For both cases, we will want to send an abort.

Furthermore, I would say "emulated MMIO region" rather than MMIO region because the P2M can also contain MMIO mapping (we usually call then "direct MMIO").

Cheers,

Signature · 1 line
Julien Grall
#15 lore raw

Re: [XEN v9 3/4] xen/arm64: io: Handle the abort due to access to stage1 translation table

Hi Julien,

One clarification.

Quoted 41 lines · Julien Grall

Hi Ayan,

Quoted 30 lines · Ayan Kumar Halder

If the abort was caused due to access to stage1 translation table, Xen will assume that the stage1 translation table is in the non MMIO region. It will try to resolve the translation fault. If it succeeds, it will return to the guest to retry the instruction. If not, then it means that the table is in MMIO region which is not expected by Xen. Thus, Xen will forward the abort to the guest.

---

Changelog :-

v1..v8 - NA

v9 - 1. Extracted this change from "[XEN v8 2/2] xen/arm64: io: Support instructions (for which ISS is not..." into a separate patch of its own. The reason being this is an existing bug in the codebase.

  xen/arch/arm/io.c    | 11 +++++++++++
  xen/arch/arm/traps.c | 12 +++++++++++-

  2 files changed, 22 insertions(+), 1 deletion(-)
xen/arch/arm/io.c +0 -0
index bea69ffb08..ebcb8ed548 100644
@@ -128,6 +128,17 @@ void try_decode_instruction(const struct 
 cpu_user_regs *regs,
           return;
       }
Quoted 5 lines

  +    /*
+     * At this point, we know that the stage1 translation table is in the MMIO
+     * region. This is not expected by Xen and thus it forwards the abort to the

We don't know that. We only know that there are no corresponding valid mapping in the P2M. So the address may be part of an emulated MMIO region or invalid.

Quoted 3 lines

For both cases, we will want to send an abort.

Furthermore, I would say "emulated MMIO region" rather than MMIO region because the P2M can also contain MMIO mapping (we usually call then "direct MMIO").

Can I say MMIO region (to indicate both emulated and direct) ? The reason being the assumption that stage1 page tables cannot be in the MMIO region. And thus, when check_p2m() is invoked, we do not invoke try_map_mmio(gaddr_to_gfn(gpa).

See this snippet :-

            if ( xabt.s1ptw )
                check_mmio_region = false;

            if ( check_p2m((is_data && check_mmio_region), gpa) ) return;

- Ayan

Quoted 3 lines

Cheers,

#16 lore raw

Re: [XEN v9 3/4] xen/arm64: io: Handle the abort due to access to stage1 translation table

Quoted 1 line · Ayan Kumar Halder

Hi Julien,

Hi Ayan,

Quoted 9 lines

One clarification.

Quoted 5 lines · Julien Grall

Hi Ayan,

Quoted 2 lines · Ayan Kumar Halder

If the abort was caused due to access to stage1 translation table, Xen will assume that the stage1 translation table is in the non MMIO region.

Reading this commit message again. I think you want to explain why we want to do that because, from my understanding, this is technically not forbidden by the Arm Arm.

From the previous discussion, we want to do this because we can't easily handle such fault on emulated region (we have no away to the walker the value read).

Quoted 38 lines
Quoted 38 lines
Quoted 28 lines

It will try to resolve the translation fault. If it succeeds, it will return to the guest to retry the instruction. If not, then it means that the table is in MMIO region which is not expected by Xen. Thus, Xen will forward the abort to the guest.

---

Changelog :-

v1..v8 - NA

v9 - 1. Extracted this change from "[XEN v8 2/2] xen/arm64: io: Support instructions (for which ISS is not..." into a separate patch of its own. The reason being this is an existing bug in the codebase.

  xen/arch/arm/io.c    | 11 +++++++++++
  xen/arch/arm/traps.c | 12 +++++++++++-

  2 files changed, 22 insertions(+), 1 deletion(-)
xen/arch/arm/io.c +0 -0
index bea69ffb08..ebcb8ed548 100644
@@ -128,6 +128,17 @@ void try_decode_instruction(const struct 
 cpu_user_regs *regs,
           return;
       }
Quoted 5 lines

  +    /*
+     * At this point, we know that the stage1 translation table is in the MMIO
+     * region. This is not expected by Xen and thus it forwards the abort to the

We don't know that. We only know that there are no corresponding valid mapping in the P2M. So the address may be part of an emulated MMIO region or invalid.

For both cases, we will want to send an abort.

Quoted 8 lines
Quoted 1 line

Furthermore, I would say "emulated MMIO region" rather than MMIO region because the P2M can also contain MMIO mapping (we usually call then "direct MMIO").

Can I say MMIO region (to indicate both emulated and direct) ? The reason being the assumption that stage1 page tables cannot be in the MMIO region. And thus, when check_p2m() is invoked, we do not invoke try_map_mmio(gaddr_to_gfn(gpa).

See this snippet :-

            if ( xabt.s1ptw )
                check_mmio_region = false;

Thinking a bit more of this. I would actually drop this check. We don't need to decode the instruction, so I don't think there are much benefits here to restrict access for direct MMIO. Did I miss anything?

Cheers,

Signature · 1 line
Julien Grall
#17 lore raw

Re: [XEN v9 3/4] xen/arm64: io: Handle the abort due to access to stage1 translation table

Quoted 6 lines · Julien Grall
Quoted 1 line · Ayan Kumar Halder

Hi Julien,

Hi Ayan,

Hi Julien,

I need a bit of clarification to understand this.

Quoted 7 lines
Quoted 7 lines

One clarification.

Quoted 2 lines · Julien Grall

Hi Ayan,

Quoted 7 lines
Quoted 3 lines
Quoted 3 lines · Ayan Kumar Halder
Quoted 2 lines

If the abort was caused due to access to stage1 translation table, Xen will assume that the stage1 translation table is in the non MMIO region.

Reading this commit message again. I think you want to explain why we want to do that because, from my understanding, this is technically not forbidden by the Arm Arm.

From the previous discussion, we want to do this because we can't easily handle such fault on emulated region (we have no away to the walker the value read).

Sorry, Can you explain this a bit more ? Do you mean that if the page table is located in the emulated region, map_domain_page() (called from p2m_next_level()) will fail.

But for emulated region, shouldn't pages be already mapped for Xen to access them ?

Quoted 41 lines
Quoted 41 lines
Quoted 40 lines
Quoted 29 lines

It will try to resolve the translation fault. If it succeeds, it will return to the guest to retry the instruction. If not, then it means that the table is in MMIO region which is not expected by Xen. Thus, Xen will forward the abort to the guest.

---

Changelog :-

v1..v8 - NA

v9 - 1. Extracted this change from "[XEN v8 2/2] xen/arm64: io: Support instructions (for which ISS is not..." into a separate patch of its own. The reason being this is an existing bug in the codebase.

  xen/arch/arm/io.c    | 11 +++++++++++
  xen/arch/arm/traps.c | 12 +++++++++++-

  2 files changed, 22 insertions(+), 1 deletion(-)
xen/arch/arm/io.c +0 -0
index bea69ffb08..ebcb8ed548 100644
@@ -128,6 +128,17 @@ void try_decode_instruction(const struct 
 cpu_user_regs *regs,
           return;
       }
Quoted 5 lines

  +    /*
+     * At this point, we know that the stage1 translation table is in the MMIO
+     * region. This is not expected by Xen and thus it forwards the abort to the

We don't know that. We only know that there are no corresponding valid mapping in the P2M. So the address may be part of an emulated MMIO region or invalid.

For both cases, we will want to send an abort.

Furthermore, I would say "emulated MMIO region" rather than MMIO region because the P2M can also contain MMIO mapping (we usually call then "direct MMIO").

Quoted 9 lines
Quoted 6 lines

Can I say MMIO region (to indicate both emulated and direct) ? The reason being the assumption that stage1 page tables cannot be in the MMIO region. And thus, when check_p2m() is invoked, we do not invoke try_map_mmio(gaddr_to_gfn(gpa).

See this snippet :-

             if ( xabt.s1ptw )
                 check_mmio_region = false;

Thinking a bit more of this. I would actually drop this check. We don't need to decode the instruction, so I don't think there are much benefits here to restrict access for direct MMIO. Did I miss anything?

Can Linux or any OS keep its page tables in the direct MMIO space ? If yes, then try_map_mmio() needs to be invoked to map the region, so that OS can access it. If not, then Xen needs to return abort because the OS may be behaving maliciously.

My understanding from previous discussion was that it does not make sense for linux or any OS to keep its page tables in any MMIO region (emulated or direct). Please correct me if mistaken.

- Ayan

Quoted 2 lines

Cheers,

#18 lore raw

Re: [XEN v9 3/4] xen/arm64: io: Handle the abort due to access to stage1 translation table

Hi,

Quoted 21 lines · Ayan Kumar Halder
Quoted 6 lines · Julien Grall
Quoted 1 line · Ayan Kumar Halder

Hi Julien,

Hi Ayan,

Hi Julien,

I need a bit of clarification to understand this.

Quoted 8 lines
Quoted 8 lines

One clarification.

Quoted 3 lines · Julien Grall

Hi Ayan,

On 01/03/2022 12:40, Ayan Kumar Halder wrote:

Quoted 8 lines
Quoted 6 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

If the abort was caused due to access to stage1 translation table, Xen will assume that the stage1 translation table is in the non MMIO region.

Reading this commit message again. I think you want to explain why we want to do that because, from my understanding, this is technically not forbidden by the Arm Arm.

From the previous discussion, we want to do this because we can't easily handle such fault on emulated region (we have no away to the walker the value read).

Sorry, Can you explain this a bit more ? Do you mean that if the page table is located in the emulated region, map_domain_page() (called from p2m_next_level()) will fail.

For data abort with valid syndrome, you will have a register to write back the value read. When the data abort has s1ptw == 1, AFAICT, we have no information how to return the value.

Quoted 2 lines

But for emulated region, shouldn't pages be already mapped for Xen to access them ?

I am not sure which "pages" you are referring to here. The implementation of emulated regions is left to the discretion of Xen. This may be reading/writing to a buffer allocated by Xen or return a fixed value.

Quoted 44 lines
Quoted 44 lines
Quoted 44 lines
Quoted 38 lines
Quoted 30 lines

It will try to resolve the translation fault. If it succeeds, it will return to the guest to retry the instruction. If not, then it means that the table is in MMIO region which is not expected by Xen. Thus, Xen will forward the abort to the guest.

---

Changelog :-

v1..v8 - NA

v9 - 1. Extracted this change from "[XEN v8 2/2] xen/arm64: io: Support instructions (for which ISS is not..." into a separate patch of its own. The reason being this is an existing bug in the codebase.

  xen/arch/arm/io.c    | 11 +++++++++++
  xen/arch/arm/traps.c | 12 +++++++++++-

  2 files changed, 22 insertions(+), 1 deletion(-)
xen/arch/arm/io.c +0 -0
index bea69ffb08..ebcb8ed548 100644
@@ -128,6 +128,17 @@ void try_decode_instruction(const struct 
 cpu_user_regs *regs,
           return;
       }
Quoted 5 lines

  +    /*
+     * At this point, we know that the stage1 translation table is in the MMIO
+     * region. This is not expected by Xen and thus it forwards the abort to the

We don't know that. We only know that there are no corresponding valid mapping in the P2M. So the address may be part of an emulated MMIO region or invalid.

Quoted 3 lines

For both cases, we will want to send an abort.

Furthermore, I would say "emulated MMIO region" rather than MMIO region because the P2M can also contain MMIO mapping (we usually call then "direct MMIO").

Can I say MMIO region (to indicate both emulated and direct) ? The reason being the assumption that stage1 page tables cannot be in the MMIO region. And thus, when check_p2m() is invoked, we do not invoke try_map_mmio(gaddr_to_gfn(gpa).

Quoted 8 lines
Quoted 7 lines
Quoted 4 lines

See this snippet :-

             if ( xabt.s1ptw )
                 check_mmio_region = false;

Thinking a bit more of this. I would actually drop this check. We don't need to decode the instruction, so I don't think there are much benefits here to restrict access for direct MMIO. Did I miss anything?

Can Linux or any OS keep its page tables in the direct MMIO space ? If yes, then try_map_mmio() needs to be invoked to map the region, so that OS can access it. If not, then Xen needs to return abort because the OS may be behaving maliciously.

I think what matters is whether the Arm Arm would or would not allow it. From what I can tell there are no such restriction. So we would need to be cautious to restrict it further than necessary.

Quoted 2 lines

My understanding from previous discussion was that it does not make sense for linux or any OS to keep its page tables in any MMIO region (emulated or direct). Please correct me if mistaken.

At the moment, none of the regions emulated by Xen could be used for page-tables. I am also not sure how we should handle such access if they arise. So it is more convenient to simply forbid them.

Regarding direct MMIO, see above. Correct me if I am wrong, but it should not be a problem for Xen to deal with them. So while I agree this doesn't seem to make sense, the restriction seems unnecessary.

Cheers,

Signature · 1 line
Julien Grall
#19 lore raw

Re: [XEN v9 3/4] xen/arm64: io: Handle the abort due to access to stage1 translation table

Hi Julien,

Quoted 27 lines · Julien Grall

Hi,

Quoted 24 lines · Ayan Kumar Halder
Quoted 6 lines · Julien Grall
Quoted 1 line · Ayan Kumar Halder

Hi Julien,

Hi Ayan,

Hi Julien,

I need a bit of clarification to understand this.

Quoted 11 lines
Quoted 10 lines

One clarification.

Quoted 5 lines · Julien Grall

Hi Ayan,

Quoted 2 lines · Ayan Kumar Halder

If the abort was caused due to access to stage1 translation table, Xen will assume that the stage1 translation table is in the non MMIO region.

Quoted 7 lines
Quoted 5 lines
Quoted 3 lines

Reading this commit message again. I think you want to explain why we want to do that because, from my understanding, this is technically not forbidden by the Arm Arm.

From the previous discussion, we want to do this because we can't easily handle such fault on emulated region (we have no away to the walker the value read).

Sorry, Can you explain this a bit more ? Do you mean that if the page table is located in the emulated region, map_domain_page() (called from p2m_next_level()) will fail.

For data abort with valid syndrome, you will have a register to write back the value read. When the data abort has s1ptw == 1, AFAICT, we have no information how to return the value.

Do you mean that for s1ptw, we get an intermediate physical address ?

    if ( hpfar_is_valid(xabt.s1ptw, fsc) )
        gpa = get_faulting_ipa(gva);

If the IPA corresponds to an emulated region, then Xen can read the emulated address, but can't return the value to the guest OS.

(I actually want to understand this very well).

Quoted 59 lines
Quoted 3 lines

But for emulated region, shouldn't pages be already mapped for Xen to access them ?

I am not sure which "pages" you are referring to here. The implementation of emulated regions is left to the discretion of Xen. This may be reading/writing to a buffer allocated by Xen or return a fixed value.

Quoted 44 lines
Quoted 44 lines
Quoted 44 lines
Quoted 38 lines
Quoted 30 lines

It will try to resolve the translation fault. If it succeeds, it will return to the guest to retry the instruction. If not, then it means that the table is in MMIO region which is not expected by Xen. Thus, Xen will forward the abort to the guest.

---

Changelog :-

v1..v8 - NA

v9 - 1. Extracted this change from "[XEN v8 2/2] xen/arm64: io: Support instructions (for which ISS is not..." into a separate patch of its own. The reason being this is an existing bug in the codebase.

  xen/arch/arm/io.c    | 11 +++++++++++
  xen/arch/arm/traps.c | 12 +++++++++++-

  2 files changed, 22 insertions(+), 1 deletion(-)
xen/arch/arm/io.c +0 -0
index bea69ffb08..ebcb8ed548 100644
@@ -128,6 +128,17 @@ void try_decode_instruction(const struct 
 cpu_user_regs *regs,
           return;
       }
Quoted 5 lines

  +    /*
+     * At this point, we know that the stage1 translation table is in the MMIO
+     * region. This is not expected by Xen and thus it forwards the abort to the

We don't know that. We only know that there are no corresponding valid mapping in the P2M. So the address may be part of an emulated MMIO region or invalid.

Quoted 3 lines

For both cases, we will want to send an abort.

Furthermore, I would say "emulated MMIO region" rather than MMIO region because the P2M can also contain MMIO mapping (we usually call then "direct MMIO").

Can I say MMIO region (to indicate both emulated and direct) ? The reason being the assumption that stage1 page tables cannot be in the MMIO region. And thus, when check_p2m() is invoked, we do not invoke try_map_mmio(gaddr_to_gfn(gpa).

Quoted 8 lines
Quoted 7 lines
Quoted 4 lines

See this snippet :-

             if ( xabt.s1ptw )
                 check_mmio_region = false;

Thinking a bit more of this. I would actually drop this check. We don't need to decode the instruction, so I don't think there are much benefits here to restrict access for direct MMIO. Did I miss anything?

Can Linux or any OS keep its page tables in the direct MMIO space ? If yes, then try_map_mmio() needs to be invoked to map the region, so that OS can access it. If not, then Xen needs to return abort because the OS may be behaving maliciously.

Quoted 8 lines

I think what matters is whether the Arm Arm would or would not allow it. From what I can tell there are no such restriction. So we would need to be cautious to restrict it further than necessary.

Quoted 2 lines

My understanding from previous discussion was that it does not make sense for linux or any OS to keep its page tables in any MMIO region (emulated or direct). Please correct me if mistaken.

At the moment, none of the regions emulated by Xen could be used for page-tables. I am also not sure how we should handle such access if they arise. So it is more convenient to simply forbid them.

Regarding direct MMIO, see above. Correct me if I am wrong, but it should not be a problem for Xen to deal with them. So while I agree this doesn't seem to make sense, the restriction seems unnecessary.

So the behavior will be :-

1. If the stage1 translation table is in the non MMIO region or 'direct mapped' MMIO region, then invoke p2m_resolve_translation_fault() and try_map_mmio() to resolve the fault. If it succeeds, then return to the guest to retry.

2. If the previous step fails and for any other scenario (ie stage1 translation table is in emulated MMIO region or the address is invalid), return the abort to the guest.

- Ayan

Quoted 3 lines

Cheers,

#20 lore raw

Re: [XEN v9 3/4] xen/arm64: io: Handle the abort due to access to stage1 translation table

Hi,

Quoted 34 lines · Ayan Kumar Halder

Hi Julien,

Quoted 31 lines · Julien Grall

Hi,

Quoted 28 lines · Ayan Kumar Halder
Quoted 6 lines · Julien Grall
Quoted 1 line · Ayan Kumar Halder

Hi Julien,

Hi Ayan,

Hi Julien,

I need a bit of clarification to understand this.

Quoted 14 lines
Quoted 10 lines

One clarification.

Quoted 5 lines · Julien Grall

Hi Ayan,

Quoted 2 lines · Ayan Kumar Halder

If the abort was caused due to access to stage1 translation table, Xen will assume that the stage1 translation table is in the non MMIO region.

Reading this commit message again. I think you want to explain why we want to do that because, from my understanding, this is technically not forbidden by the Arm Arm.

From the previous discussion, we want to do this because we can't easily handle such fault on emulated region (we have no away to the walker the value read).

Quoted 8 lines
Quoted 3 lines
Quoted 1 line

Sorry, Can you explain this a bit more ? Do you mean that if the page table is located in the emulated region, map_domain_page() (called from p2m_next_level()) will fail.

For data abort with valid syndrome, you will have a register to write back the value read. When the data abort has s1ptw == 1, AFAICT, we have no information how to return the value.

Do you mean that for s1ptw, we get an intermediate physical address ?

    if ( hpfar_is_valid(xabt.s1ptw, fsc) )
        gpa = get_faulting_ipa(gva);

No. That's not relevant here. We always need a IPA in order to deal with the fault.

Quoted 2 lines

If the IPA corresponds to an emulated region, then Xen can read the emulated address, but can't return the value to the guest OS.

That wouldn't be the guest OS. But the page-table walker in the CPU.

Quoted 7 lines
Quoted 7 lines
Quoted 1 line

Can Linux or any OS keep its page tables in the direct MMIO space ? If yes, then try_map_mmio() needs to be invoked to map the region, so that OS can access it. If not, then Xen needs to return abort because the OS may be behaving maliciously.

I think what matters is whether the Arm Arm would or would not allow it. From what I can tell there are no such restriction. So we would need to be cautious to restrict it further than necessary.

Quoted 2 lines

My understanding from previous discussion was that it does not make sense for linux or any OS to keep its page tables in any MMIO region (emulated or direct). Please correct me if mistaken.

Quoted 7 lines
Quoted 3 lines

At the moment, none of the regions emulated by Xen could be used for page-tables. I am also not sure how we should handle such access if they arise. So it is more convenient to simply forbid them.

Regarding direct MMIO, see above. Correct me if I am wrong, but it should not be a problem for Xen to deal with them. So while I agree this doesn't seem to make sense, the restriction seems unnecessary.

So the behavior will be :-

1. If the stage1 translation table is in the non MMIO region or 'direct mapped' MMIO region, then invoke p2m_resolve_translation_fault() and try_map_mmio() to resolve the fault. If it succeeds, then return to the guest to retry.

When 1. happens we don't know yet if the stage1 will be a non MMIO region or 'direct mapped'. Instead, we only know that the ISS syndrome is invalid.

If it is a prefetch abort or the syndrome is invalid, then we should call check_p2m().

Quoted 2 lines

2. If the previous step fails and for any other scenario (ie stage1 translation table is in emulated MMIO region or the address is invalid), return the abort to the guest.

If check_p2m() didn't success and it is a fault on stage-1 walk, then we should send an abort.

Cheers,

Signature · 1 line
Julien Grall
#21 lore raw

PATCH v9 4/4 xen/arm64: io: Handle data abort due to cache maintenance instructions

In reply to Ayan Kumar Halder

When the data abort is caused due to cache maintenance for an address, there are two scenarios:-

1. Address belonging to a non emulated region - For this, Xen should set the corresponding bit in the translation table entry to valid and return to the guest to retry the instruction. This can happen sometimes as Xen need to set the translation table entry to invalid. (for eg 'Break-Before-Make' sequence).

2. Address belongs to an emulated region - Xen should ignore the instruction (ie increment the PC) and return to the guest.

We try to deal with scenario#1, by invoking check_p2m(). If this is unsuccessful, then we assume scenario#2.

---

Changelog:-

v1...v8 - NA

v9 - Extracted this change from "[XEN v7 2/2] xen/arm64: io: Support instructions (for which ISS is not ..." into a separate patch of its own. The reason being this addresses an existing bug in the codebase.

 xen/arch/arm/include/asm/mmio.h |  3 ++-
 xen/arch/arm/io.c               | 11 +++++++++++
 xen/arch/arm/traps.c            |  6 ++++++
 3 files changed, 19 insertions(+), 1 deletion(-)
xen/arch/arm/include/asm/mmio.h +2 -1
index ef2c57a2d5..75d362d5f5 100644
@@ -34,7 +34,8 @@ enum instr_decode_state
      * Instruction is decoded successfully. It is a ldr/str post indexing
      * instruction.
      */
-    INSTR_LDR_STR_POSTINDEXING
+    INSTR_LDR_STR_POSTINDEXING,
+    INSTR_IGNORE                    /* Instruction is ignored */
 };
 
 typedef struct
xen/arch/arm/io.c +11 -0
index ebcb8ed548..7e9dd4bb08 100644
@@ -139,6 +139,17 @@ void try_decode_instruction(const struct cpu_user_regs *regs,
         return;
     }
 
+    /*
+     * When the data abort is caused due to cache maintenance, Xen should ignore
+     * this instruction as the cache maintenance was caused on an address belonging
+     * to the emulated region.
+     */
+    if ( info->dabt.cache )
+    {
+        info->dabt_instr.state = INSTR_IGNORE;
+        return;
+    }
+
     /*
      * Armv8 processor does not provide a valid syndrome for decoding some
      * instructions. So in order to process these instructions, Xen must
xen/arch/arm/traps.c +6 -0
index e491ca15d7..5879640b73 100644
@@ -2011,6 +2011,12 @@ static void do_trap_stage2_abort_guest(struct cpu_user_regs *regs,
 
         try_decode_instruction(regs, &info);
 
+        if ( info.dabt_instr.state == INSTR_IGNORE )
+        {
+            advance_pc(regs, hsr);
+            return;
+        }
+
         /*
          * If Xen could not decode the instruction or encountered an error
          * while decoding, then it should forward the abort to the guest.
Signature · 1 line
2.17.1
Acked-by on v9 4/4
#22 lore raw

Re: [XEN v9 4/4] xen/arm64: io: Handle data abort due to cache maintenance instructions

Quoted 9 lines · Ayan Kumar Halder

When the data abort is caused due to cache maintenance for an address, there are two scenarios:-

1. Address belonging to a non emulated region - For this, Xen should set the corresponding bit in the translation table entry to valid and return to the guest to retry the instruction. This can happen sometimes as Xen need to set the translation table entry to invalid. (for eg 'Break-Before-Make' sequence).

Quoted 7 lines

2. Address belongs to an emulated region - Xen should ignore the instruction (ie increment the PC) and return to the guest.

We try to deal with scenario#1, by invoking check_p2m(). If this is unsuccessful, then we assume scenario#2.

Quoted 71 lines

---

Changelog:-

v1...v8 - NA

v9 - Extracted this change from "[XEN v7 2/2] xen/arm64: io: Support instructions (for which ISS is not ..." into a separate patch of its own. The reason being this addresses an existing bug in the codebase.

 xen/arch/arm/include/asm/mmio.h |  3 ++-
 xen/arch/arm/io.c               | 11 +++++++++++
 xen/arch/arm/traps.c            |  6 ++++++
 3 files changed, 19 insertions(+), 1 deletion(-)
xen/arch/arm/include/asm/mmio.h +2 -1
index ef2c57a2d5..75d362d5f5 100644
@@ -34,7 +34,8 @@ enum instr_decode_state
      * Instruction is decoded successfully. It is a ldr/str post indexing
      * instruction.
      */
-    INSTR_LDR_STR_POSTINDEXING
+    INSTR_LDR_STR_POSTINDEXING,
+    INSTR_IGNORE                    /* Instruction is ignored */
 };
 
 typedef struct
xen/arch/arm/io.c +11 -0
index ebcb8ed548..7e9dd4bb08 100644
@@ -139,6 +139,17 @@ void try_decode_instruction(const struct cpu_user_regs *regs,
         return;
     }
 
+    /*
+     * When the data abort is caused due to cache maintenance, Xen should ignore
+     * this instruction as the cache maintenance was caused on an address belonging
+     * to the emulated region.
+     */
+    if ( info->dabt.cache )
+    {
+        info->dabt_instr.state = INSTR_IGNORE;
+        return;
+    }
+
     /*
      * Armv8 processor does not provide a valid syndrome for decoding some
      * instructions. So in order to process these instructions, Xen must
xen/arch/arm/traps.c +6 -0
index e491ca15d7..5879640b73 100644
@@ -2011,6 +2011,12 @@ static void do_trap_stage2_abort_guest(struct cpu_user_regs *regs,
 
         try_decode_instruction(regs, &info);
 
+        if ( info.dabt_instr.state == INSTR_IGNORE )
+        {
+            advance_pc(regs, hsr);
+            return;
+        }
+
         /*
          * If Xen could not decode the instruction or encountered an error
          * while decoding, then it should forward the abort to the guest.
Signature · 1 line
2.17.1
#23 lore raw

Re: [XEN v9 4/4] xen/arm64: io: Handle data abort due to cache maintenance instructions

In reply to Ayan Kumar Halder

Hi Ayan,

Quoted 11 lines · Ayan Kumar Halder

When the data abort is caused due to cache maintenance for an address, there are two scenarios:-

1. Address belonging to a non emulated region - For this, Xen should set the corresponding bit in the translation table entry to valid and return to the guest to retry the instruction. This can happen sometimes as Xen need to set the translation table entry to invalid. (for eg 'Break-Before-Make' sequence).

2. Address belongs to an emulated region - Xen should ignore the instruction (ie increment the PC) and return to the guest.

I would be explicit and say something along the lines:

"Xen doesn't cache data for emulated regions. So we can safely ignore them".

There is a third scenarios:

The address belongs to neither an emulated region nor has a valid mapping in the P2M.

Quoted 3 lines

We try to deal with scenario#1, by invoking check_p2m(). If this is unsuccessful, then we assume scenario#2.

This means that you will ignore cache maintenance on invalid region. I think we should send an abort to the guest rather than let them believe it worked.

Cheers,

Signature · 1 line
Julien Grall
#24 lore raw

Re: [XEN v9 4/4] xen/arm64: io: Handle data abort due to cache maintenance instructions

Hi Julien,

Quoted 15 lines · Julien Grall

Hi Ayan,

Quoted 11 lines · Ayan Kumar Halder

When the data abort is caused due to cache maintenance for an address, there are two scenarios:-

1. Address belonging to a non emulated region - For this, Xen should set the corresponding bit in the translation table entry to valid and return to the guest to retry the instruction. This can happen sometimes as Xen need to set the translation table entry to invalid. (for eg 'Break-Before-Make' sequence).

2. Address belongs to an emulated region - Xen should ignore the instruction (ie increment the PC) and return to the guest.

Quoted 7 lines

I would be explicit and say something along the lines:

"Xen doesn't cache data for emulated regions. So we can safely ignore them".

There is a third scenarios:

The address belongs to neither an emulated region nor has a valid mapping in the P2M.

To check this, we should test "try_handle_mmio() == IO_UNHANDLED". If so then send an abort to the guest.

Is this correct ?

- Ayan

Quoted 8 lines
Quoted 3 lines

We try to deal with scenario#1, by invoking check_p2m(). If this is unsuccessful, then we assume scenario#2.

This means that you will ignore cache maintenance on invalid region. I think we should send an abort to the guest rather than let them believe it worked.

Cheers,

#25 lore raw

Re: [XEN v9 4/4] xen/arm64: io: Handle data abort due to cache maintenance instructions

Quoted 1 line · Ayan Kumar Halder

Hi Julien,

Hi,

Quoted 21 lines · Julien Grall
Quoted 19 lines

Hi Ayan,

Quoted 11 lines · Ayan Kumar Halder

When the data abort is caused due to cache maintenance for an address, there are two scenarios:-

1. Address belonging to a non emulated region - For this, Xen should set the corresponding bit in the translation table entry to valid and return to the guest to retry the instruction. This can happen sometimes as Xen need to set the translation table entry to invalid. (for eg 'Break-Before-Make' sequence).

2. Address belongs to an emulated region - Xen should ignore the instruction (ie increment the PC) and return to the guest.

I would be explicit and say something along the lines:

"Xen doesn't cache data for emulated regions. So we can safely ignore them".

Quoted 7 lines
Quoted 3 lines

There is a third scenarios:

The address belongs to neither an emulated region nor has a valid mapping in the P2M.

To check this, we should test "try_handle_mmio() == IO_UNHANDLED". If so then send an abort to the guest.

Is this correct ?

I think it would be too late because if the region is emulated, then we would have already tried to handle it.

Instead, I think we need to check after we confirmed that the region is emulated or we need to forward to an IOREQ server.

So the check would have to be duplicated here.

Cheers,

Signature · 1 line
Julien Grall
#26 lore raw

Re: [XEN v9 4/4] xen/arm64: io: Handle data abort due to cache maintenance instructions

Hi Julien,

I have a question.

Quoted 32 lines · Julien Grall
Quoted 1 line · Ayan Kumar Halder

Hi Julien,

Hi,

Quoted 25 lines · Julien Grall
Quoted 22 lines

Hi Ayan,

Quoted 11 lines · Ayan Kumar Halder

When the data abort is caused due to cache maintenance for an address, there are two scenarios:-

1. Address belonging to a non emulated region - For this, Xen should set the corresponding bit in the translation table entry to valid and return to the guest to retry the instruction. This can happen sometimes as Xen need to set the translation table entry to invalid. (for eg 'Break-Before-Make' sequence).

2. Address belongs to an emulated region - Xen should ignore the instruction (ie increment the PC) and return to the guest.

I would be explicit and say something along the lines:

"Xen doesn't cache data for emulated regions. So we can safely ignore them".

There is a third scenarios:

The address belongs to neither an emulated region nor has a valid mapping in the P2M.

Quoted 8 lines
Quoted 3 lines

To check this, we should test "try_handle_mmio() == IO_UNHANDLED". If so then send an abort to the guest.

Is this correct ?

I think it would be too late because if the region is emulated, then we would have already tried to handle it.

Instead, I think we need to check after we confirmed that the region is emulated or we need to forward to an IOREQ server.

So the check would have to be duplicated here.

When do we know that a particular address does not belong to an emulated MMIO region ?

Is this after both "find_mmio_handler()" and "ioreq_server_select()" have returned NULL ?

- Ayan

Quoted 3 lines

Cheers,

#27 lore raw

Re: [XEN v9 4/4] xen/arm64: io: Handle data abort due to cache maintenance instructions

Quoted 1 line · Ayan Kumar Halder

Hi Julien,

Hi,

Quoted 40 lines

I have a question.

Quoted 37 lines · Julien Grall
Quoted 1 line · Ayan Kumar Halder

Hi Julien,

Hi,

Quoted 21 lines · Julien Grall
Quoted 19 lines

Hi Ayan,

Quoted 11 lines · Ayan Kumar Halder

When the data abort is caused due to cache maintenance for an address, there are two scenarios:-

1. Address belonging to a non emulated region - For this, Xen should set the corresponding bit in the translation table entry to valid and return to the guest to retry the instruction. This can happen sometimes as Xen need to set the translation table entry to invalid. (for eg 'Break-Before-Make' sequence).

2. Address belongs to an emulated region - Xen should ignore the instruction (ie increment the PC) and return to the guest.

I would be explicit and say something along the lines:

"Xen doesn't cache data for emulated regions. So we can safely ignore them".

Quoted 7 lines
Quoted 3 lines

There is a third scenarios:

The address belongs to neither an emulated region nor has a valid mapping in the P2M.

To check this, we should test "try_handle_mmio() == IO_UNHANDLED". If so then send an abort to the guest.

Is this correct ?

I think it would be too late because if the region is emulated, then we would have already tried to handle it.

Quoted 7 lines
Quoted 3 lines

Instead, I think we need to check after we confirmed that the region is emulated or we need to forward to an IOREQ server.

So the check would have to be duplicated here.

When do we know that a particular address does not belong to an emulated MMIO region ?

Is this after both "find_mmio_handler()" and "ioreq_server_select()" have returned NULL ?

Correct.

Cheers,

Signature · 1 line
Julien Grall