Skip to contentSkip to conversation

Nested virtualization fixes for QEMU

PATCH Applied Anup Patel 2 participants 19 messages Nov 8, 2022 – Jan 19, 2023 RISC-V

v2 of 3 Applied Latest is v3 →

lore
Share & tools
lore

Anup Patel

author
#0 lore raw

v2 cover Nested virtualization fixes for QEMU

This series mainly includes fixes discovered while developing nested virtualization running on QEMU.

These patches can also be found in the riscv_nested_fixes_v2 branch at:
https://github.com/avpatel/qemu.git

Changes since v1:
- Added Alistair's Reviewed-by tags to appropriate patches
- Added detailed comment block in PATCH4

Anup Patel (5):
target/riscv: Typo fix in sstc() predicate target/riscv: Update VS timer whenever htimedelta changes target/riscv: Don't clear mask in riscv_cpu_update_mip() for VSTIP target/riscv: No need to re-start QEMU timer when timecmp ==

    UINT64_MAX

target/riscv: Ensure opcode is saved for all relevant instructions

 target/riscv/cpu_helper.c                   |  2 --
 target/riscv/csr.c                          | 18 ++++++++++-
 target/riscv/insn_trans/trans_rva.c.inc     | 10 ++++--
 target/riscv/insn_trans/trans_rvd.c.inc     |  2 ++
 target/riscv/insn_trans/trans_rvf.c.inc     |  2 ++
 target/riscv/insn_trans/trans_rvh.c.inc     |  3 ++
 target/riscv/insn_trans/trans_rvi.c.inc     |  2 ++
 target/riscv/insn_trans/trans_rvzfh.c.inc   |  2 ++
 target/riscv/insn_trans/trans_svinval.c.inc |  3 ++
 target/riscv/time_helper.c                  | 36 ++++++++++++++++++---
 10 files changed, 70 insertions(+), 10 deletions(-)
Signature · 1 line
2.34.1
author
#1 lore raw

PATCH v2 1/5 target/riscv: Typo fix in sstc() predicate

We should use "&&" instead of "&" when checking hcounteren.TM and henvcfg.STCE bits.

Fixes 3ec0fe18a31f ("target/riscv: Add vstimecmp suppor")
---
 target/riscv/csr.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
target/riscv/csr.c +1 -1
index 5c9a7ee287..716f9d960e 100644
@@ -838,7 +838,7 @@ static RISCVException sstc(CPURISCVState *env, int csrno)
     }
 
     if (riscv_cpu_virt_enabled(env)) {
-        if (!(get_field(env->hcounteren, COUNTEREN_TM) &
+        if (!(get_field(env->hcounteren, COUNTEREN_TM) &&
               get_field(env->henvcfg, HENVCFG_STCE))) {
             return RISCV_EXCP_VIRT_INSTRUCTION_FAULT;
         }
Signature · 1 line
2.34.1
author
#2 lore raw

PATCH v2 2/5 target/riscv: Update VS timer whenever htimedelta changes

In reply to Anup Patel

The htimedelta[h] CSR has impact on the VS timer comparison so we should call riscv_timer_write_timecmp() whenever htimedelta changes.

Fixes 3ec0fe18a31f ("target/riscv: Add vstimecmp suppor")
---
 target/riscv/csr.c | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
target/riscv/csr.c +16 -0
index 716f9d960e..4b1a608260 100644
@@ -2722,6 +2722,8 @@ static RISCVException read_htimedelta(CPURISCVState *env, int csrno,
 static RISCVException write_htimedelta(CPURISCVState *env, int csrno,
                                        target_ulong val)
 {
+    RISCVCPU *cpu = env_archcpu(env);
+
     if (!env->rdtime_fn) {
         return RISCV_EXCP_ILLEGAL_INST;
     }
@@ -2731,6 +2733,12 @@ static RISCVException write_htimedelta(CPURISCVState *env, int csrno,
     } else {
         env->htimedelta = val;
     }
+
+    if (cpu->cfg.ext_sstc && env->rdtime_fn) {
+        riscv_timer_write_timecmp(cpu, env->vstimer, env->vstimecmp,
+                                  env->htimedelta, MIP_VSTIP);
+    }
+
     return RISCV_EXCP_NONE;
 }
 
@@ -2748,11 +2756,19 @@ static RISCVException read_htimedeltah(CPURISCVState *env, int csrno,
 static RISCVException write_htimedeltah(CPURISCVState *env, int csrno,
                                         target_ulong val)
 {
+    RISCVCPU *cpu = env_archcpu(env);
+
     if (!env->rdtime_fn) {
         return RISCV_EXCP_ILLEGAL_INST;
     }
 
     env->htimedelta = deposit64(env->htimedelta, 32, 32, (uint64_t)val);
+
+    if (cpu->cfg.ext_sstc && env->rdtime_fn) {
+        riscv_timer_write_timecmp(cpu, env->vstimer, env->vstimecmp,
+                                  env->htimedelta, MIP_VSTIP);
+    }
+
     return RISCV_EXCP_NONE;
 }
 
Signature · 1 line
2.34.1
author
#13 lore raw

PATCH v2 3/5 target/riscv: Don't clear mask in riscv_cpu_update_mip() for VSTIP

In reply to Anup Patel

Instead of clearing mask in riscv_cpu_update_mip() for VSTIP, we should call riscv_cpu_update_mip() with mask == 0 from timer_helper.c for VSTIP.

Fixes 3ec0fe18a31f ("target/riscv: Add vstimecmp suppor")
---
 target/riscv/cpu_helper.c  |  2 --
 target/riscv/time_helper.c | 12 ++++++++----
 2 files changed, 8 insertions(+), 6 deletions(-)
target/riscv/cpu_helper.c +0 -2
index 5d66246c2c..a403825e49 100644
@@ -617,8 +617,6 @@ uint64_t riscv_cpu_update_mip(RISCVCPU *cpu, uint64_t mask, uint64_t value)
         vsgein = (env->hgeip & (1ULL << gein)) ? MIP_VSEIP : 0;
     }
 
-    /* No need to update mip for VSTIP */
-    mask = ((mask == MIP_VSTIP) && env->vstime_irq) ? 0 : mask;
     vstip = env->vstime_irq ? MIP_VSTIP : 0;
 
     if (!qemu_mutex_iothread_locked()) {
target/riscv/time_helper.c +8 -4
index 8cce667dfd..4fb2a471a9 100644
@@ -27,7 +27,7 @@ static void riscv_vstimer_cb(void *opaque)
     RISCVCPU *cpu = opaque;
     CPURISCVState *env = &cpu->env;
     env->vstime_irq = 1;
-    riscv_cpu_update_mip(cpu, MIP_VSTIP, BOOL_TO_MASK(1));
+    riscv_cpu_update_mip(cpu, 0, BOOL_TO_MASK(1));
 }
 
 static void riscv_stimer_cb(void *opaque)
@@ -57,16 +57,20 @@ void riscv_timer_write_timecmp(RISCVCPU *cpu, QEMUTimer *timer,
          */
         if (timer_irq == MIP_VSTIP) {
             env->vstime_irq = 1;
+            riscv_cpu_update_mip(cpu, 0, BOOL_TO_MASK(1));
+        } else {
+            riscv_cpu_update_mip(cpu, MIP_STIP, BOOL_TO_MASK(1));
         }
-        riscv_cpu_update_mip(cpu, timer_irq, BOOL_TO_MASK(1));
         return;
     }
 
+    /* Clear the [VS|S]TIP bit in mip */
     if (timer_irq == MIP_VSTIP) {
         env->vstime_irq = 0;
+        riscv_cpu_update_mip(cpu, 0, BOOL_TO_MASK(0));
+    } else {
+        riscv_cpu_update_mip(cpu, timer_irq, BOOL_TO_MASK(0));
     }
-    /* Clear the [V]STIP bit in mip */
-    riscv_cpu_update_mip(cpu, timer_irq, BOOL_TO_MASK(0));
 
     /* otherwise, set up the future timer interrupt */
     diff = timecmp - rtc_r;
Signature · 1 line
2.34.1
author
#15 lore raw

PATCH v2 5/5 target/riscv: Ensure opcode is saved for all relevant instructions

In reply to Anup Patel

We should call decode_save_opc() for all relevant instructions which can potentially generate a virtual instruction fault or a guest page fault because generating transformed instruction upon guest page fault expects opcode to be available. Without this, hypervisor will see transformed instruction as zero in htinst CSR for guest MMIO emulation which makes MMIO emulation in hypervisor slow and also breaks nested virtualization.

Fixes a9814e3e08d2 ("target/riscv: Minimize the calls to decode_save_opc")
---
 target/riscv/insn_trans/trans_rva.c.inc     | 10 +++++++---
 target/riscv/insn_trans/trans_rvd.c.inc     |  2 ++
 target/riscv/insn_trans/trans_rvf.c.inc     |  2 ++
 target/riscv/insn_trans/trans_rvh.c.inc     |  3 +++
 target/riscv/insn_trans/trans_rvi.c.inc     |  2 ++
 target/riscv/insn_trans/trans_rvzfh.c.inc   |  2 ++
 target/riscv/insn_trans/trans_svinval.c.inc |  3 +++
 7 files changed, 21 insertions(+), 3 deletions(-)
target/riscv/insn_trans/trans_rva.c.inc +7 -3
index 45db82c9be..5f194a447b 100644
@@ -20,8 +20,10 @@
 
 static bool gen_lr(DisasContext *ctx, arg_atomic *a, MemOp mop)
 {
-    TCGv src1 = get_address(ctx, a->rs1, 0);
+    TCGv src1;
 
+    decode_save_opc(ctx);
+    src1 = get_address(ctx, a->rs1, 0);
     if (a->rl) {
         tcg_gen_mb(TCG_MO_ALL | TCG_BAR_STRL);
     }
@@ -43,6 +45,7 @@ static bool gen_sc(DisasContext *ctx, arg_atomic *a, MemOp mop)
     TCGLabel *l1 = gen_new_label();
     TCGLabel *l2 = gen_new_label();
 
+    decode_save_opc(ctx);
     src1 = get_address(ctx, a->rs1, 0);
     tcg_gen_brcond_tl(TCG_COND_NE, load_res, src1, l1);
 
@@ -81,9 +84,10 @@ static bool gen_amo(DisasContext *ctx, arg_atomic *a,
                     MemOp mop)
 {
     TCGv dest = dest_gpr(ctx, a->rd);
-    TCGv src1 = get_address(ctx, a->rs1, 0);
-    TCGv src2 = get_gpr(ctx, a->rs2, EXT_NONE);
+    TCGv src1, src2 = get_gpr(ctx, a->rs2, EXT_NONE);
 
+    decode_save_opc(ctx);
+    src1 = get_address(ctx, a->rs1, 0);
     func(dest, src1, src2, ctx->mem_idx, mop);
 
     gen_set_gpr(ctx, a->rd, dest);
target/riscv/insn_trans/trans_rvd.c.inc +2 -0
index 1397c1ce1c..6e3159b797 100644
@@ -38,6 +38,7 @@ static bool trans_fld(DisasContext *ctx, arg_fld *a)
     REQUIRE_FPU;
     REQUIRE_EXT(ctx, RVD);
 
+    decode_save_opc(ctx);
     addr = get_address(ctx, a->rs1, a->imm);
     tcg_gen_qemu_ld_i64(cpu_fpr[a->rd], addr, ctx->mem_idx, MO_TEUQ);
 
@@ -52,6 +53,7 @@ static bool trans_fsd(DisasContext *ctx, arg_fsd *a)
     REQUIRE_FPU;
     REQUIRE_EXT(ctx, RVD);
 
+    decode_save_opc(ctx);
     addr = get_address(ctx, a->rs1, a->imm);
     tcg_gen_qemu_st_i64(cpu_fpr[a->rs2], addr, ctx->mem_idx, MO_TEUQ);
     return true;
target/riscv/insn_trans/trans_rvf.c.inc +2 -0
index a1d3eb52ad..965e1f8d11 100644
@@ -38,6 +38,7 @@ static bool trans_flw(DisasContext *ctx, arg_flw *a)
     REQUIRE_FPU;
     REQUIRE_EXT(ctx, RVF);
 
+    decode_save_opc(ctx);
     addr = get_address(ctx, a->rs1, a->imm);
     dest = cpu_fpr[a->rd];
     tcg_gen_qemu_ld_i64(dest, addr, ctx->mem_idx, MO_TEUL);
@@ -54,6 +55,7 @@ static bool trans_fsw(DisasContext *ctx, arg_fsw *a)
     REQUIRE_FPU;
     REQUIRE_EXT(ctx, RVF);
 
+    decode_save_opc(ctx);
     addr = get_address(ctx, a->rs1, a->imm);
     tcg_gen_qemu_st_i64(cpu_fpr[a->rs2], addr, ctx->mem_idx, MO_TEUL);
     return true;
target/riscv/insn_trans/trans_rvh.c.inc +3 -0
index 4f8aecddc7..9248b48c36 100644
@@ -36,6 +36,7 @@ static bool do_hlv(DisasContext *ctx, arg_r2 *a, MemOp mop)
 #ifdef CONFIG_USER_ONLY
     return false;
 #else
+    decode_save_opc(ctx);
     if (check_access(ctx)) {
         TCGv dest = dest_gpr(ctx, a->rd);
         TCGv addr = get_gpr(ctx, a->rs1, EXT_NONE);
@@ -82,6 +83,7 @@ static bool do_hsv(DisasContext *ctx, arg_r2_s *a, MemOp mop)
 #ifdef CONFIG_USER_ONLY
     return false;
 #else
+    decode_save_opc(ctx);
     if (check_access(ctx)) {
         TCGv addr = get_gpr(ctx, a->rs1, EXT_NONE);
         TCGv data = get_gpr(ctx, a->rs2, EXT_NONE);
@@ -135,6 +137,7 @@ static bool trans_hsv_d(DisasContext *ctx, arg_hsv_d *a)
 static bool do_hlvx(DisasContext *ctx, arg_r2 *a,
                     void (*func)(TCGv, TCGv_env, TCGv))
 {
+    decode_save_opc(ctx);
     if (check_access(ctx)) {
         TCGv dest = dest_gpr(ctx, a->rd);
         TCGv addr = get_gpr(ctx, a->rs1, EXT_NONE);
target/riscv/insn_trans/trans_rvi.c.inc +2 -0
index c49dbec0eb..1665efb639 100644
@@ -261,6 +261,7 @@ static bool gen_load_i128(DisasContext *ctx, arg_lb *a, MemOp memop)
 
 static bool gen_load(DisasContext *ctx, arg_lb *a, MemOp memop)
 {
+    decode_save_opc(ctx);
     if (get_xl(ctx) == MXL_RV128) {
         return gen_load_i128(ctx, a, memop);
     } else {
@@ -350,6 +351,7 @@ static bool gen_store_i128(DisasContext *ctx, arg_sb *a, MemOp memop)
 
 static bool gen_store(DisasContext *ctx, arg_sb *a, MemOp memop)
 {
+    decode_save_opc(ctx);
     if (get_xl(ctx) == MXL_RV128) {
         return gen_store_i128(ctx, a, memop);
     } else {
target/riscv/insn_trans/trans_rvzfh.c.inc +2 -0
index 5d07150cd0..2ad5716312 100644
@@ -49,6 +49,7 @@ static bool trans_flh(DisasContext *ctx, arg_flh *a)
     REQUIRE_FPU;
     REQUIRE_ZFH_OR_ZFHMIN(ctx);
 
+    decode_save_opc(ctx);
     t0 = get_gpr(ctx, a->rs1, EXT_NONE);
     if (a->imm) {
         TCGv temp = temp_new(ctx);
@@ -71,6 +72,7 @@ static bool trans_fsh(DisasContext *ctx, arg_fsh *a)
     REQUIRE_FPU;
     REQUIRE_ZFH_OR_ZFHMIN(ctx);
 
+    decode_save_opc(ctx);
     t0 = get_gpr(ctx, a->rs1, EXT_NONE);
     if (a->imm) {
         TCGv temp = tcg_temp_new();
target/riscv/insn_trans/trans_svinval.c.inc +3 -0
index 2682bd969f..f3cd7d5c0b 100644
@@ -28,6 +28,7 @@ static bool trans_sinval_vma(DisasContext *ctx, arg_sinval_vma *a)
     /* Do the same as sfence.vma currently */
     REQUIRE_EXT(ctx, RVS);
 #ifndef CONFIG_USER_ONLY
+    decode_save_opc(ctx);
     gen_helper_tlb_flush(cpu_env);
     return true;
 #endif
@@ -56,6 +57,7 @@ static bool trans_hinval_vvma(DisasContext *ctx, arg_hinval_vvma *a)
     /* Do the same as hfence.vvma currently */
     REQUIRE_EXT(ctx, RVH);
 #ifndef CONFIG_USER_ONLY
+    decode_save_opc(ctx);
     gen_helper_hyp_tlb_flush(cpu_env);
     return true;
 #endif
@@ -68,6 +70,7 @@ static bool trans_hinval_gvma(DisasContext *ctx, arg_hinval_gvma *a)
     /* Do the same as hfence.gvma currently */
     REQUIRE_EXT(ctx, RVH);
 #ifndef CONFIG_USER_ONLY
+    decode_save_opc(ctx);
     gen_helper_hyp_gvma_tlb_flush(cpu_env);
     return true;
 #endif
Signature · 1 line
2.34.1
author
#14 lore raw

PATCH v2 4/5 target/riscv: No need to re-start QEMU timer when timecmp == UINT64_MAX

In reply to Anup Patel

The time CSR will wrap-around immediately after reaching UINT64_MAX so we don't need to re-start QEMU timer when timecmp == UINT64_MAX in riscv_timer_write_timecmp().

---
 target/riscv/time_helper.c | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)
target/riscv/time_helper.c +24 -0
index 4fb2a471a9..b654f91af9 100644
@@ -72,6 +72,30 @@ void riscv_timer_write_timecmp(RISCVCPU *cpu, QEMUTimer *timer,
         riscv_cpu_update_mip(cpu, timer_irq, BOOL_TO_MASK(0));
     }
 
+    /*
+     * Sstc specification says the following about timer interrupt:
+     * "A supervisor timer interrupt becomes pending - as reflected in
+     * the STIP bit in the mip and sip registers - whenever time contains
+     * a value greater than or equal to stimecmp, treating the values
+     * as unsigned integers. Writes to stimecmp are guaranteed to be
+     * reflected in STIP eventually, but not necessarily immediately.
+     * The interrupt remains posted until stimecmp becomes greater
+     * than time - typically as a result of writing stimecmp."
+     *
+     * When timecmp = UINT64_MAX, the time CSR will eventually reach
+     * timecmp value but on next timer tick the time CSR will wrap-around
+     * and become zero which is less than UINT64_MAX. Now, the timer
+     * interrupt behaves like a level triggered interrupt so it will
+     * become 1 when time = timecmp = UINT64_MAX and next timer tick
+     * it will become 0 again because time = 0 < timecmp = UINT64_MAX.
+     *
+     * Based on above, we don't re-start the QEMU timer when timecmp
+     * equals UINT64_MAX.
+     */
+    if (timecmp == UINT64_MAX) {
+        return;
+    }
+
     /* otherwise, set up the future timer interrupt */
     diff = timecmp - rtc_r;
     /* back to ns (note args switched in muldiv64) */
Signature · 1 line
2.34.1
author
#17 lore raw
In reply to Anup Patel

Hi Alistair,

Quoted 11 lines · Anup Patel

This series mainly includes fixes discovered while developing nested virtualization running on QEMU.

These patches can also be found in the riscv_nested_fixes_v2 branch at:
https://github.com/avpatel/qemu.git

Changes since v1:
- Added Alistair's Reviewed-by tags to appropriate patches
- Added detailed comment block in PATCH4

Quoted 7 lines

Anup Patel (5):
target/riscv: Typo fix in sstc() predicate target/riscv: Update VS timer whenever htimedelta changes target/riscv: Don't clear mask in riscv_cpu_update_mip() for VSTIP target/riscv: No need to re-start QEMU timer when timecmp ==

    UINT64_MAX

target/riscv: Ensure opcode is saved for all relevant instructions

Friendly ping ?

Regards,
Anup

Quoted 16 lines
 target/riscv/cpu_helper.c                   |  2 --
 target/riscv/csr.c                          | 18 ++++++++++-
 target/riscv/insn_trans/trans_rva.c.inc     | 10 ++++--
 target/riscv/insn_trans/trans_rvd.c.inc     |  2 ++
 target/riscv/insn_trans/trans_rvf.c.inc     |  2 ++
 target/riscv/insn_trans/trans_rvh.c.inc     |  3 ++
 target/riscv/insn_trans/trans_rvi.c.inc     |  2 ++
 target/riscv/insn_trans/trans_rvzfh.c.inc   |  2 ++
 target/riscv/insn_trans/trans_svinval.c.inc |  3 ++
 target/riscv/time_helper.c                  | 36 ++++++++++++++++++---
 10 files changed, 70 insertions(+), 10 deletions(-)

--
2.34.1

Reviewed-by on v2 5/5
#16 lore raw

Re: [PATCH v2 5/5] target/riscv: Ensure opcode is saved for all relevant instructions

In reply to Anup Patel
Quoted 11 lines · Anup Patel

We should call decode_save_opc() for all relevant instructions which can potentially generate a virtual instruction fault or a guest page fault because generating transformed instruction upon guest page fault expects opcode to be available. Without this, hypervisor will see transformed instruction as zero in htinst CSR for guest MMIO emulation which makes MMIO emulation in hypervisor slow and also breaks nested virtualization.

Fixes a9814e3e08d2 ("target/riscv: Minimize the calls to decode_save_opc")

Alistair

Quoted 187 lines
---
 target/riscv/insn_trans/trans_rva.c.inc     | 10 +++++++---
 target/riscv/insn_trans/trans_rvd.c.inc     |  2 ++
 target/riscv/insn_trans/trans_rvf.c.inc     |  2 ++
 target/riscv/insn_trans/trans_rvh.c.inc     |  3 +++
 target/riscv/insn_trans/trans_rvi.c.inc     |  2 ++
 target/riscv/insn_trans/trans_rvzfh.c.inc   |  2 ++
 target/riscv/insn_trans/trans_svinval.c.inc |  3 +++
 7 files changed, 21 insertions(+), 3 deletions(-)
target/riscv/insn_trans/trans_rva.c.inc +7 -3
index 45db82c9be..5f194a447b 100644
@@ -20,8 +20,10 @@
 
 static bool gen_lr(DisasContext *ctx, arg_atomic *a, MemOp mop)
 {
-    TCGv src1 = get_address(ctx, a->rs1, 0);
+    TCGv src1;
 
+    decode_save_opc(ctx);
+    src1 = get_address(ctx, a->rs1, 0);
     if (a->rl) {
         tcg_gen_mb(TCG_MO_ALL | TCG_BAR_STRL);
     }
@@ -43,6 +45,7 @@ static bool gen_sc(DisasContext *ctx, arg_atomic *a, MemOp mop)
     TCGLabel *l1 = gen_new_label();
     TCGLabel *l2 = gen_new_label();
 
+    decode_save_opc(ctx);
     src1 = get_address(ctx, a->rs1, 0);
     tcg_gen_brcond_tl(TCG_COND_NE, load_res, src1, l1);
 
@@ -81,9 +84,10 @@ static bool gen_amo(DisasContext *ctx, arg_atomic *a,
                     MemOp mop)
 {
     TCGv dest = dest_gpr(ctx, a->rd);
-    TCGv src1 = get_address(ctx, a->rs1, 0);
-    TCGv src2 = get_gpr(ctx, a->rs2, EXT_NONE);
+    TCGv src1, src2 = get_gpr(ctx, a->rs2, EXT_NONE);
 
+    decode_save_opc(ctx);
+    src1 = get_address(ctx, a->rs1, 0);
     func(dest, src1, src2, ctx->mem_idx, mop);
 
     gen_set_gpr(ctx, a->rd, dest);
target/riscv/insn_trans/trans_rvd.c.inc +2 -0
index 1397c1ce1c..6e3159b797 100644
@@ -38,6 +38,7 @@ static bool trans_fld(DisasContext *ctx, arg_fld *a)
     REQUIRE_FPU;
     REQUIRE_EXT(ctx, RVD);
 
+    decode_save_opc(ctx);
     addr = get_address(ctx, a->rs1, a->imm);
     tcg_gen_qemu_ld_i64(cpu_fpr[a->rd], addr, ctx->mem_idx, MO_TEUQ);
 
@@ -52,6 +53,7 @@ static bool trans_fsd(DisasContext *ctx, arg_fsd *a)
     REQUIRE_FPU;
     REQUIRE_EXT(ctx, RVD);
 
+    decode_save_opc(ctx);
     addr = get_address(ctx, a->rs1, a->imm);
     tcg_gen_qemu_st_i64(cpu_fpr[a->rs2], addr, ctx->mem_idx, MO_TEUQ);
     return true;
target/riscv/insn_trans/trans_rvf.c.inc +2 -0
index a1d3eb52ad..965e1f8d11 100644
@@ -38,6 +38,7 @@ static bool trans_flw(DisasContext *ctx, arg_flw *a)
     REQUIRE_FPU;
     REQUIRE_EXT(ctx, RVF);
 
+    decode_save_opc(ctx);
     addr = get_address(ctx, a->rs1, a->imm);
     dest = cpu_fpr[a->rd];
     tcg_gen_qemu_ld_i64(dest, addr, ctx->mem_idx, MO_TEUL);
@@ -54,6 +55,7 @@ static bool trans_fsw(DisasContext *ctx, arg_fsw *a)
     REQUIRE_FPU;
     REQUIRE_EXT(ctx, RVF);
 
+    decode_save_opc(ctx);
     addr = get_address(ctx, a->rs1, a->imm);
     tcg_gen_qemu_st_i64(cpu_fpr[a->rs2], addr, ctx->mem_idx, MO_TEUL);
     return true;
target/riscv/insn_trans/trans_rvh.c.inc +3 -0
index 4f8aecddc7..9248b48c36 100644
@@ -36,6 +36,7 @@ static bool do_hlv(DisasContext *ctx, arg_r2 *a, MemOp mop)
 #ifdef CONFIG_USER_ONLY
     return false;
 #else
+    decode_save_opc(ctx);
     if (check_access(ctx)) {
         TCGv dest = dest_gpr(ctx, a->rd);
         TCGv addr = get_gpr(ctx, a->rs1, EXT_NONE);
@@ -82,6 +83,7 @@ static bool do_hsv(DisasContext *ctx, arg_r2_s *a, MemOp mop)
 #ifdef CONFIG_USER_ONLY
     return false;
 #else
+    decode_save_opc(ctx);
     if (check_access(ctx)) {
         TCGv addr = get_gpr(ctx, a->rs1, EXT_NONE);
         TCGv data = get_gpr(ctx, a->rs2, EXT_NONE);
@@ -135,6 +137,7 @@ static bool trans_hsv_d(DisasContext *ctx, arg_hsv_d *a)
 static bool do_hlvx(DisasContext *ctx, arg_r2 *a,
                     void (*func)(TCGv, TCGv_env, TCGv))
 {
+    decode_save_opc(ctx);
     if (check_access(ctx)) {
         TCGv dest = dest_gpr(ctx, a->rd);
         TCGv addr = get_gpr(ctx, a->rs1, EXT_NONE);
target/riscv/insn_trans/trans_rvi.c.inc +2 -0
index c49dbec0eb..1665efb639 100644
@@ -261,6 +261,7 @@ static bool gen_load_i128(DisasContext *ctx, arg_lb *a, MemOp memop)
 
 static bool gen_load(DisasContext *ctx, arg_lb *a, MemOp memop)
 {
+    decode_save_opc(ctx);
     if (get_xl(ctx) == MXL_RV128) {
         return gen_load_i128(ctx, a, memop);
     } else {
@@ -350,6 +351,7 @@ static bool gen_store_i128(DisasContext *ctx, arg_sb *a, MemOp memop)
 
 static bool gen_store(DisasContext *ctx, arg_sb *a, MemOp memop)
 {
+    decode_save_opc(ctx);
     if (get_xl(ctx) == MXL_RV128) {
         return gen_store_i128(ctx, a, memop);
     } else {
target/riscv/insn_trans/trans_rvzfh.c.inc +2 -0
index 5d07150cd0..2ad5716312 100644
@@ -49,6 +49,7 @@ static bool trans_flh(DisasContext *ctx, arg_flh *a)
     REQUIRE_FPU;
     REQUIRE_ZFH_OR_ZFHMIN(ctx);
 
+    decode_save_opc(ctx);
     t0 = get_gpr(ctx, a->rs1, EXT_NONE);
     if (a->imm) {
         TCGv temp = temp_new(ctx);
@@ -71,6 +72,7 @@ static bool trans_fsh(DisasContext *ctx, arg_fsh *a)
     REQUIRE_FPU;
     REQUIRE_ZFH_OR_ZFHMIN(ctx);
 
+    decode_save_opc(ctx);
     t0 = get_gpr(ctx, a->rs1, EXT_NONE);
     if (a->imm) {
         TCGv temp = tcg_temp_new();
target/riscv/insn_trans/trans_svinval.c.inc +3 -1
index 2682bd969f..f3cd7d5c0b 100644
@@ -28,6 +28,7 @@ static bool trans_sinval_vma(DisasContext *ctx, arg_sinval_vma *a)
     /* Do the same as sfence.vma currently */
     REQUIRE_EXT(ctx, RVS);
 #ifndef CONFIG_USER_ONLY
+    decode_save_opc(ctx);
     gen_helper_tlb_flush(cpu_env);
     return true;
 #endif
@@ -56,6 +57,7 @@ static bool trans_hinval_vvma(DisasContext *ctx, arg_hinval_vvma *a)
     /* Do the same as hfence.vvma currently */
     REQUIRE_EXT(ctx, RVH);
 #ifndef CONFIG_USER_ONLY
+    decode_save_opc(ctx);
     gen_helper_hyp_tlb_flush(cpu_env);
     return true;
 #endif
@@ -68,6 +70,7 @@ static bool trans_hinval_gvma(DisasContext *ctx, arg_hinval_gvma *a)
     /* Do the same as hfence.gvma currently */
     REQUIRE_EXT(ctx, RVH);
 #ifndef CONFIG_USER_ONLY
+    decode_save_opc(ctx);
     gen_helper_hyp_gvma_tlb_flush(cpu_env);
     return true;
 #endif
--

2.34.1

#18 lore raw
In reply to Anup Patel
Quoted 11 lines · Anup Patel

This series mainly includes fixes discovered while developing nested virtualization running on QEMU.

These patches can also be found in the riscv_nested_fixes_v2 branch at:
https://github.com/avpatel/qemu.git

Changes since v1:
- Added Alistair's Reviewed-by tags to appropriate patches
- Added detailed comment block in PATCH4

Quoted 7 lines

Anup Patel (5):
target/riscv: Typo fix in sstc() predicate target/riscv: Update VS timer whenever htimedelta changes target/riscv: Don't clear mask in riscv_cpu_update_mip() for VSTIP target/riscv: No need to re-start QEMU timer when timecmp ==

    UINT64_MAX

target/riscv: Ensure opcode is saved for all relevant instructions

Thanks!

Applied to riscv-to-apply.next

Alistair

Quoted 17 lines
 target/riscv/cpu_helper.c                   |  2 --
 target/riscv/csr.c                          | 18 ++++++++++-
 target/riscv/insn_trans/trans_rva.c.inc     | 10 ++++--
 target/riscv/insn_trans/trans_rvd.c.inc     |  2 ++
 target/riscv/insn_trans/trans_rvf.c.inc     |  2 ++
 target/riscv/insn_trans/trans_rvh.c.inc     |  3 ++
 target/riscv/insn_trans/trans_rvi.c.inc     |  2 ++
 target/riscv/insn_trans/trans_rvzfh.c.inc   |  2 ++
 target/riscv/insn_trans/trans_svinval.c.inc |  3 ++
 target/riscv/time_helper.c                  | 36 ++++++++++++++++++---
 10 files changed, 70 insertions(+), 10 deletions(-)

--
2.34.1

#3 lore raw

Re: [PATCH v2 2/5] target/riscv: Update VS timer whenever htimedelta changes

In reply to Anup Patel
Quoted 7 lines · Anup Patel

The htimedelta[h] CSR has impact on the VS timer comparison so we should call riscv_timer_write_timecmp() whenever htimedelta changes.

Fixes 3ec0fe18a31f ("target/riscv: Add vstimecmp suppor")

This patch breaks my Xvisor test. When running OpenSBI and Xvisor like this:

qemu-system-riscv64 -machine virt \
-m 1G -serial mon:stdio -serial null -nographic \
-append 'vmm.console=uart@10000000 vmm.bootcmd="vfs mount initrd
/;vfs run /boot.xscript;vfs cat /system/banner.txt; guest kick guest0;

vserial bind guest0/uart0"' \
    -smp 4 -d guest_errors \
    -bios none \
    -device loader,file=./images/qemuriscv64/vmm.bin,addr=0x80200000 \
    -kernel ./images/qemuriscv64/fw_jump.elf \
    -initrd ./images/qemuriscv64/vmm-disk-linux.img -cpu rv64,h=true

Running:

Xvisor v0.3.0-129-gbc33f339 (Jan 1 1970 00:00:00)

I see this failure:

INIT: bootcmd: guest kick guest0

guest0: Kicked

INIT: bootcmd: vserial bind guest0/uart0

[guest0/uart0] cpu_vcpu_stage2_map: guest_phys=0x000000003B9AC000 size=0x4096 map failed

do_error: CPU3: VCPU=guest0/vcpu0 page fault failed (error -1)

       zero=0x0000000000000000          ra=0x0000000080001B4E

         sp=0x000000008001CF80          gp=0x0000000000000000

         tp=0x0000000000000000          s0=0x000000008001CFB0

         s1=0x0000000000000000          a0=0x0000000010001048

         a1=0x0000000000000000          a2=0x0000000000989680

         a3=0x000000003B9ACA00          a4=0x0000000000000048

         a5=0x0000000000000000          a6=0x0000000000019000

         a7=0x0000000000000000          s2=0x0000000000000000

         s3=0x0000000000000000          s4=0x0000000000000000

         s5=0x0000000000000000          s6=0x0000000000000000

         s7=0x0000000000000000          s8=0x0000000000000000

         s9=0x0000000000000000         s10=0x0000000000000000

        s11=0x0000000000000000          t0=0x0000000000004000

         t1=0x0000000000000100          t2=0x0000000000000000

         t3=0x0000000000000000          t4=0x0000000000000000

         t5=0x0000000000000000          t6=0x0000000000000000

       sepc=0x0000000080001918     sstatus=0x0000000200004120

    hstatus=0x00000002002001C0     sp_exec=0x0000000010A64000

     scause=0x0000000000000017       stval=0x000000003B9ACAF8

      htval=0x000000000EE6B2BE      htinst=0x0000000000D03021

I have tried updating to a newer Xvisor release, but with that I don't get any serial output.

Can you help get the Xvisor tests back up and running?

Alistair

author
#4 lore raw

Re: [PATCH v2 2/5] target/riscv: Update VS timer whenever htimedelta changes

Quoted 80 lines · Alistair Francis
Quoted 7 lines · Anup Patel

The htimedelta[h] CSR has impact on the VS timer comparison so we should call riscv_timer_write_timecmp() whenever htimedelta changes.

Fixes 3ec0fe18a31f ("target/riscv: Add vstimecmp suppor")

This patch breaks my Xvisor test. When running OpenSBI and Xvisor like this:

qemu-system-riscv64 -machine virt \
-m 1G -serial mon:stdio -serial null -nographic \
-append 'vmm.console=uart@10000000 vmm.bootcmd="vfs mount initrd
/;vfs run /boot.xscript;vfs cat /system/banner.txt; guest kick guest0;

vserial bind guest0/uart0"' \
    -smp 4 -d guest_errors \
    -bios none \
    -device loader,file=./images/qemuriscv64/vmm.bin,addr=0x80200000 \
    -kernel ./images/qemuriscv64/fw_jump.elf \
    -initrd ./images/qemuriscv64/vmm-disk-linux.img -cpu rv64,h=true

Running:

Xvisor v0.3.0-129-gbc33f339 (Jan 1 1970 00:00:00)

I see this failure:

INIT: bootcmd: guest kick guest0

guest0: Kicked

INIT: bootcmd: vserial bind guest0/uart0

[guest0/uart0] cpu_vcpu_stage2_map: guest_phys=0x000000003B9AC000 size=0x4096 map failed

do_error: CPU3: VCPU=guest0/vcpu0 page fault failed (error -1)

       zero=0x0000000000000000          ra=0x0000000080001B4E

         sp=0x000000008001CF80          gp=0x0000000000000000

         tp=0x0000000000000000          s0=0x000000008001CFB0

         s1=0x0000000000000000          a0=0x0000000010001048

         a1=0x0000000000000000          a2=0x0000000000989680

         a3=0x000000003B9ACA00          a4=0x0000000000000048

         a5=0x0000000000000000          a6=0x0000000000019000

         a7=0x0000000000000000          s2=0x0000000000000000

         s3=0x0000000000000000          s4=0x0000000000000000

         s5=0x0000000000000000          s6=0x0000000000000000

         s7=0x0000000000000000          s8=0x0000000000000000

         s9=0x0000000000000000         s10=0x0000000000000000

        s11=0x0000000000000000          t0=0x0000000000004000

         t1=0x0000000000000100          t2=0x0000000000000000

         t3=0x0000000000000000          t4=0x0000000000000000

         t5=0x0000000000000000          t6=0x0000000000000000

       sepc=0x0000000080001918     sstatus=0x0000000200004120

    hstatus=0x00000002002001C0     sp_exec=0x0000000010A64000

     scause=0x0000000000000017       stval=0x000000003B9ACAF8

      htval=0x000000000EE6B2BE      htinst=0x0000000000D03021
Quoted 4 lines

I have tried updating to a newer Xvisor release, but with that I don't get any serial output.

Can you help get the Xvisor tests back up and running?

I tried the latest Xvisor-next (https://github.com/avpatel/xvisor-next)
with your QEMU riscv-to-apply.next branch and it works fine (both with and without Sstc).

Here's the QEMU command which I use:

qemu-system-riscv64 -M virt -m 512M -nographic \
-bios opensbi/build/platform/generic/firmware/fw_jump.bin \
-kernel ../xvisor-next/build/vmm.bin \
-initrd rbd_v64.img \
-append "vmm.bootcmd=\"vfs mount initrd /;vfs run /boot.xscript;vfs cat /system/banner.txt\"" \
-smp 4

Also, I will be releasing Xvisor-0.3.2 by the end of Dec 2022 so I suggest using this upcoming release in your test.

Regards,
Anup

#5 lore raw

Re: [PATCH v2 2/5] target/riscv: Update VS timer whenever htimedelta changes

Quoted 82 lines · Anup Patel
Quoted 80 lines · Alistair Francis
Quoted 7 lines · Anup Patel

The htimedelta[h] CSR has impact on the VS timer comparison so we should call riscv_timer_write_timecmp() whenever htimedelta changes.

Fixes 3ec0fe18a31f ("target/riscv: Add vstimecmp suppor")

This patch breaks my Xvisor test. When running OpenSBI and Xvisor like this:

qemu-system-riscv64 -machine virt \
-m 1G -serial mon:stdio -serial null -nographic \
-append 'vmm.console=uart@10000000 vmm.bootcmd="vfs mount initrd
/;vfs run /boot.xscript;vfs cat /system/banner.txt; guest kick guest0;

vserial bind guest0/uart0"' \
    -smp 4 -d guest_errors \
    -bios none \
    -device loader,file=./images/qemuriscv64/vmm.bin,addr=0x80200000 \
    -kernel ./images/qemuriscv64/fw_jump.elf \
    -initrd ./images/qemuriscv64/vmm-disk-linux.img -cpu rv64,h=true

Running:

Xvisor v0.3.0-129-gbc33f339 (Jan 1 1970 00:00:00)

I see this failure:

INIT: bootcmd: guest kick guest0

guest0: Kicked

INIT: bootcmd: vserial bind guest0/uart0

[guest0/uart0] cpu_vcpu_stage2_map: guest_phys=0x000000003B9AC000 size=0x4096 map failed

do_error: CPU3: VCPU=guest0/vcpu0 page fault failed (error -1)

       zero=0x0000000000000000          ra=0x0000000080001B4E

         sp=0x000000008001CF80          gp=0x0000000000000000

         tp=0x0000000000000000          s0=0x000000008001CFB0

         s1=0x0000000000000000          a0=0x0000000010001048

         a1=0x0000000000000000          a2=0x0000000000989680

         a3=0x000000003B9ACA00          a4=0x0000000000000048

         a5=0x0000000000000000          a6=0x0000000000019000

         a7=0x0000000000000000          s2=0x0000000000000000

         s3=0x0000000000000000          s4=0x0000000000000000

         s5=0x0000000000000000          s6=0x0000000000000000

         s7=0x0000000000000000          s8=0x0000000000000000

         s9=0x0000000000000000         s10=0x0000000000000000

        s11=0x0000000000000000          t0=0x0000000000004000

         t1=0x0000000000000100          t2=0x0000000000000000

         t3=0x0000000000000000          t4=0x0000000000000000

         t5=0x0000000000000000          t6=0x0000000000000000

       sepc=0x0000000080001918     sstatus=0x0000000200004120

    hstatus=0x00000002002001C0     sp_exec=0x0000000010A64000

     scause=0x0000000000000017       stval=0x000000003B9ACAF8

      htval=0x000000000EE6B2BE      htinst=0x0000000000D03021
Quoted 8 lines
Quoted 4 lines

I have tried updating to a newer Xvisor release, but with that I don't get any serial output.

Can you help get the Xvisor tests back up and running?

I tried the latest Xvisor-next (https://github.com/avpatel/xvisor-next)
with your QEMU riscv-to-apply.next branch and it works fine (both with and without Sstc).

Does it work with the latest release?

Alistair

Quoted 16 lines

Here's the QEMU command which I use:

qemu-system-riscv64 -M virt -m 512M -nographic \

(quoted diff) +0 -4
-bios opensbi/build/platform/generic/firmware/fw_jump.bin \
-kernel ../xvisor-next/build/vmm.bin \
-initrd rbd_v64.img \
-append "vmm.bootcmd=\"vfs mount initrd /;vfs run /boot.xscript;vfs

cat /system/banner.txt\"" \
-smp 4

Also, I will be releasing Xvisor-0.3.2 by the end of Dec 2022 so I suggest using this upcoming release in your test.

Regards,
Anup

author
#6 lore raw

Re: [PATCH v2 2/5] target/riscv: Update VS timer whenever htimedelta changes

Quoted 87 lines · Alistair Francis
Quoted 85 lines · Anup Patel
Quoted 83 lines · Alistair Francis
Quoted 7 lines · Anup Patel

The htimedelta[h] CSR has impact on the VS timer comparison so we should call riscv_timer_write_timecmp() whenever htimedelta changes.

Fixes 3ec0fe18a31f ("target/riscv: Add vstimecmp suppor")

This patch breaks my Xvisor test. When running OpenSBI and Xvisor like this:

qemu-system-riscv64 -machine virt \
-m 1G -serial mon:stdio -serial null -nographic \
-append 'vmm.console=uart@10000000 vmm.bootcmd="vfs mount initrd
/;vfs run /boot.xscript;vfs cat /system/banner.txt; guest kick guest0;

vserial bind guest0/uart0"' \
    -smp 4 -d guest_errors \
    -bios none \
    -device loader,file=./images/qemuriscv64/vmm.bin,addr=0x80200000 \
    -kernel ./images/qemuriscv64/fw_jump.elf \
    -initrd ./images/qemuriscv64/vmm-disk-linux.img -cpu rv64,h=true

Running:

Xvisor v0.3.0-129-gbc33f339 (Jan 1 1970 00:00:00)

I see this failure:

INIT: bootcmd: guest kick guest0

guest0: Kicked

INIT: bootcmd: vserial bind guest0/uart0

[guest0/uart0] cpu_vcpu_stage2_map: guest_phys=0x000000003B9AC000 size=0x4096 map failed

do_error: CPU3: VCPU=guest0/vcpu0 page fault failed (error -1)

       zero=0x0000000000000000          ra=0x0000000080001B4E

         sp=0x000000008001CF80          gp=0x0000000000000000

         tp=0x0000000000000000          s0=0x000000008001CFB0

         s1=0x0000000000000000          a0=0x0000000010001048

         a1=0x0000000000000000          a2=0x0000000000989680

         a3=0x000000003B9ACA00          a4=0x0000000000000048

         a5=0x0000000000000000          a6=0x0000000000019000

         a7=0x0000000000000000          s2=0x0000000000000000

         s3=0x0000000000000000          s4=0x0000000000000000

         s5=0x0000000000000000          s6=0x0000000000000000

         s7=0x0000000000000000          s8=0x0000000000000000

         s9=0x0000000000000000         s10=0x0000000000000000

        s11=0x0000000000000000          t0=0x0000000000004000

         t1=0x0000000000000100          t2=0x0000000000000000

         t3=0x0000000000000000          t4=0x0000000000000000

         t5=0x0000000000000000          t6=0x0000000000000000

       sepc=0x0000000080001918     sstatus=0x0000000200004120

    hstatus=0x00000002002001C0     sp_exec=0x0000000010A64000

     scause=0x0000000000000017       stval=0x000000003B9ACAF8

      htval=0x000000000EE6B2BE      htinst=0x0000000000D03021

I have tried updating to a newer Xvisor release, but with that I don't get any serial output.

Quoted 7 lines
Quoted 5 lines
Quoted 1 line

Can you help get the Xvisor tests back up and running?

I tried the latest Xvisor-next (https://github.com/avpatel/xvisor-next)
with your QEMU riscv-to-apply.next branch and it works fine (both with and without Sstc).

Does it work with the latest release?

Yes, the latest Xvisor-next repo works for QEMU v7.2.0-rc4 and your riscv-to-apply.next branch (commit 51bb9de2d188)

Regards,
Anup

Quoted 19 lines

Alistair

Quoted 16 lines

Here's the QEMU command which I use:

qemu-system-riscv64 -M virt -m 512M -nographic \

(quoted diff) +0 -4
-bios opensbi/build/platform/generic/firmware/fw_jump.bin \
-kernel ../xvisor-next/build/vmm.bin \
-initrd rbd_v64.img \
-append "vmm.bootcmd=\"vfs mount initrd /;vfs run /boot.xscript;vfs

cat /system/banner.txt\"" \
-smp 4

Also, I will be releasing Xvisor-0.3.2 by the end of Dec 2022 so I suggest using this upcoming release in your test.

Regards,
Anup

#7 lore raw

Re: [PATCH v2 2/5] target/riscv: Update VS timer whenever htimedelta changes

Quoted 91 lines · Anup Patel
Quoted 89 lines · Alistair Francis
Quoted 87 lines · Anup Patel
Quoted 84 lines · Alistair Francis
Quoted 7 lines · Anup Patel

The htimedelta[h] CSR has impact on the VS timer comparison so we should call riscv_timer_write_timecmp() whenever htimedelta changes.

Fixes 3ec0fe18a31f ("target/riscv: Add vstimecmp suppor")

This patch breaks my Xvisor test. When running OpenSBI and Xvisor like this:

qemu-system-riscv64 -machine virt \
-m 1G -serial mon:stdio -serial null -nographic \
-append 'vmm.console=uart@10000000 vmm.bootcmd="vfs mount initrd
/;vfs run /boot.xscript;vfs cat /system/banner.txt; guest kick guest0;

vserial bind guest0/uart0"' \
    -smp 4 -d guest_errors \
    -bios none \
    -device loader,file=./images/qemuriscv64/vmm.bin,addr=0x80200000 \
    -kernel ./images/qemuriscv64/fw_jump.elf \
    -initrd ./images/qemuriscv64/vmm-disk-linux.img -cpu rv64,h=true

Running:

Xvisor v0.3.0-129-gbc33f339 (Jan 1 1970 00:00:00)

I see this failure:

INIT: bootcmd: guest kick guest0

guest0: Kicked

INIT: bootcmd: vserial bind guest0/uart0

[guest0/uart0] cpu_vcpu_stage2_map: guest_phys=0x000000003B9AC000 size=0x4096 map failed

do_error: CPU3: VCPU=guest0/vcpu0 page fault failed (error -1)

       zero=0x0000000000000000          ra=0x0000000080001B4E

         sp=0x000000008001CF80          gp=0x0000000000000000

         tp=0x0000000000000000          s0=0x000000008001CFB0

         s1=0x0000000000000000          a0=0x0000000010001048

         a1=0x0000000000000000          a2=0x0000000000989680

         a3=0x000000003B9ACA00          a4=0x0000000000000048

         a5=0x0000000000000000          a6=0x0000000000019000

         a7=0x0000000000000000          s2=0x0000000000000000

         s3=0x0000000000000000          s4=0x0000000000000000

         s5=0x0000000000000000          s6=0x0000000000000000

         s7=0x0000000000000000          s8=0x0000000000000000

         s9=0x0000000000000000         s10=0x0000000000000000

        s11=0x0000000000000000          t0=0x0000000000004000

         t1=0x0000000000000100          t2=0x0000000000000000

         t3=0x0000000000000000          t4=0x0000000000000000

         t5=0x0000000000000000          t6=0x0000000000000000

       sepc=0x0000000080001918     sstatus=0x0000000200004120

    hstatus=0x00000002002001C0     sp_exec=0x0000000010A64000

     scause=0x0000000000000017       stval=0x000000003B9ACAF8

      htval=0x000000000EE6B2BE      htinst=0x0000000000D03021

I have tried updating to a newer Xvisor release, but with that I don't get any serial output.

Can you help get the Xvisor tests back up and running?

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines

I tried the latest Xvisor-next (https://github.com/avpatel/xvisor-next)
with your QEMU riscv-to-apply.next branch and it works fine (both with and without Sstc).

Does it work with the latest release?

Yes, the latest Xvisor-next repo works for QEMU v7.2.0-rc4 and your riscv-to-apply.next branch (commit 51bb9de2d188)

I can't get anything to work with this patch. I have dropped this and the patches after this.

I'm building the latest Xvisor release with:

export CROSS_COMPILE=riscv64-linux-gnu-
ARCH=riscv make generic-64b-defconfig
make

and running it as above, yet nothing. What am I missing here?

Alistair

Quoted 23 lines

Regards,
Anup

Quoted 19 lines

Alistair

Quoted 16 lines

Here's the QEMU command which I use:

qemu-system-riscv64 -M virt -m 512M -nographic \

(quoted diff) +0 -4
-bios opensbi/build/platform/generic/firmware/fw_jump.bin \
-kernel ../xvisor-next/build/vmm.bin \
-initrd rbd_v64.img \
-append "vmm.bootcmd=\"vfs mount initrd /;vfs run /boot.xscript;vfs

cat /system/banner.txt\"" \
-smp 4

Also, I will be releasing Xvisor-0.3.2 by the end of Dec 2022 so I suggest using this upcoming release in your test.

Regards,
Anup

author
#8 lore raw

Re: [PATCH v2 2/5] target/riscv: Update VS timer whenever htimedelta changes

Quoted 105 lines · Alistair Francis
Quoted 91 lines · Anup Patel
Quoted 89 lines · Alistair Francis
Quoted 87 lines · Anup Patel
Quoted 84 lines · Alistair Francis
Quoted 7 lines · Anup Patel

The htimedelta[h] CSR has impact on the VS timer comparison so we should call riscv_timer_write_timecmp() whenever htimedelta changes.

Fixes 3ec0fe18a31f ("target/riscv: Add vstimecmp suppor")

This patch breaks my Xvisor test. When running OpenSBI and Xvisor like this:

qemu-system-riscv64 -machine virt \
-m 1G -serial mon:stdio -serial null -nographic \
-append 'vmm.console=uart@10000000 vmm.bootcmd="vfs mount initrd
/;vfs run /boot.xscript;vfs cat /system/banner.txt; guest kick guest0;

vserial bind guest0/uart0"' \
    -smp 4 -d guest_errors \
    -bios none \
    -device loader,file=./images/qemuriscv64/vmm.bin,addr=0x80200000 \
    -kernel ./images/qemuriscv64/fw_jump.elf \
    -initrd ./images/qemuriscv64/vmm-disk-linux.img -cpu rv64,h=true

Running:

Xvisor v0.3.0-129-gbc33f339 (Jan 1 1970 00:00:00)

I see this failure:

INIT: bootcmd: guest kick guest0

guest0: Kicked

INIT: bootcmd: vserial bind guest0/uart0

[guest0/uart0] cpu_vcpu_stage2_map: guest_phys=0x000000003B9AC000 size=0x4096 map failed

do_error: CPU3: VCPU=guest0/vcpu0 page fault failed (error -1)

       zero=0x0000000000000000          ra=0x0000000080001B4E

         sp=0x000000008001CF80          gp=0x0000000000000000

         tp=0x0000000000000000          s0=0x000000008001CFB0

         s1=0x0000000000000000          a0=0x0000000010001048

         a1=0x0000000000000000          a2=0x0000000000989680

         a3=0x000000003B9ACA00          a4=0x0000000000000048

         a5=0x0000000000000000          a6=0x0000000000019000

         a7=0x0000000000000000          s2=0x0000000000000000

         s3=0x0000000000000000          s4=0x0000000000000000

         s5=0x0000000000000000          s6=0x0000000000000000

         s7=0x0000000000000000          s8=0x0000000000000000

         s9=0x0000000000000000         s10=0x0000000000000000

        s11=0x0000000000000000          t0=0x0000000000004000

         t1=0x0000000000000100          t2=0x0000000000000000

         t3=0x0000000000000000          t4=0x0000000000000000

         t5=0x0000000000000000          t6=0x0000000000000000

       sepc=0x0000000080001918     sstatus=0x0000000200004120

    hstatus=0x00000002002001C0     sp_exec=0x0000000010A64000

     scause=0x0000000000000017       stval=0x000000003B9ACAF8

      htval=0x000000000EE6B2BE      htinst=0x0000000000D03021

I have tried updating to a newer Xvisor release, but with that I don't get any serial output.

Can you help get the Xvisor tests back up and running?

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines

I tried the latest Xvisor-next (https://github.com/avpatel/xvisor-next)
with your QEMU riscv-to-apply.next branch and it works fine (both with and without Sstc).

Does it work with the latest release?

Yes, the latest Xvisor-next repo works for QEMU v7.2.0-rc4 and your riscv-to-apply.next branch (commit 51bb9de2d188)

I can't get anything to work with this patch. I have dropped this and the patches after this.

Quoted 7 lines

I'm building the latest Xvisor release with:

export CROSS_COMPILE=riscv64-linux-gnu-
ARCH=riscv make generic-64b-defconfig
make

and running it as above, yet nothing. What am I missing here?

I tried multiple times with the latest Xvisor on different machines but still can't reproduce the issue you are seeing.

We generally provide pre-built binaries with every Xvisor release so I will share with you pre-built binaries of the upcoming Xvisor-0.3.2 release. Maybe that would help you ?

Regards,
Anup

Quoted 26 lines

Alistair

Quoted 23 lines

Regards,
Anup

Quoted 19 lines

Alistair

Quoted 16 lines

Here's the QEMU command which I use:

qemu-system-riscv64 -M virt -m 512M -nographic \

(quoted diff) +0 -4
-bios opensbi/build/platform/generic/firmware/fw_jump.bin \
-kernel ../xvisor-next/build/vmm.bin \
-initrd rbd_v64.img \
-append "vmm.bootcmd=\"vfs mount initrd /;vfs run /boot.xscript;vfs

cat /system/banner.txt\"" \
-smp 4

Also, I will be releasing Xvisor-0.3.2 by the end of Dec 2022 so I suggest using this upcoming release in your test.

Regards,
Anup

#9 lore raw

Re: [PATCH v2 2/5] target/riscv: Update VS timer whenever htimedelta changes

Quoted 109 lines · Anup Patel
Quoted 107 lines · Alistair Francis
Quoted 91 lines · Anup Patel
Quoted 89 lines · Alistair Francis
Quoted 87 lines · Anup Patel
Quoted 84 lines · Alistair Francis
Quoted 7 lines · Anup Patel

The htimedelta[h] CSR has impact on the VS timer comparison so we should call riscv_timer_write_timecmp() whenever htimedelta changes.

Fixes 3ec0fe18a31f ("target/riscv: Add vstimecmp suppor")

This patch breaks my Xvisor test. When running OpenSBI and Xvisor like this:

qemu-system-riscv64 -machine virt \
-m 1G -serial mon:stdio -serial null -nographic \
-append 'vmm.console=uart@10000000 vmm.bootcmd="vfs mount initrd
/;vfs run /boot.xscript;vfs cat /system/banner.txt; guest kick guest0;

vserial bind guest0/uart0"' \
    -smp 4 -d guest_errors \
    -bios none \
    -device loader,file=./images/qemuriscv64/vmm.bin,addr=0x80200000 \
    -kernel ./images/qemuriscv64/fw_jump.elf \
    -initrd ./images/qemuriscv64/vmm-disk-linux.img -cpu rv64,h=true

Running:

Xvisor v0.3.0-129-gbc33f339 (Jan 1 1970 00:00:00)

I see this failure:

INIT: bootcmd: guest kick guest0

guest0: Kicked

INIT: bootcmd: vserial bind guest0/uart0

[guest0/uart0] cpu_vcpu_stage2_map: guest_phys=0x000000003B9AC000 size=0x4096 map failed

do_error: CPU3: VCPU=guest0/vcpu0 page fault failed (error -1)

       zero=0x0000000000000000          ra=0x0000000080001B4E

         sp=0x000000008001CF80          gp=0x0000000000000000

         tp=0x0000000000000000          s0=0x000000008001CFB0

         s1=0x0000000000000000          a0=0x0000000010001048

         a1=0x0000000000000000          a2=0x0000000000989680

         a3=0x000000003B9ACA00          a4=0x0000000000000048

         a5=0x0000000000000000          a6=0x0000000000019000

         a7=0x0000000000000000          s2=0x0000000000000000

         s3=0x0000000000000000          s4=0x0000000000000000

         s5=0x0000000000000000          s6=0x0000000000000000

         s7=0x0000000000000000          s8=0x0000000000000000

         s9=0x0000000000000000         s10=0x0000000000000000

        s11=0x0000000000000000          t0=0x0000000000004000

         t1=0x0000000000000100          t2=0x0000000000000000

         t3=0x0000000000000000          t4=0x0000000000000000

         t5=0x0000000000000000          t6=0x0000000000000000

       sepc=0x0000000080001918     sstatus=0x0000000200004120

    hstatus=0x00000002002001C0     sp_exec=0x0000000010A64000

     scause=0x0000000000000017       stval=0x000000003B9ACAF8

      htval=0x000000000EE6B2BE      htinst=0x0000000000D03021

I have tried updating to a newer Xvisor release, but with that I don't get any serial output.

Can you help get the Xvisor tests back up and running?

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines

I tried the latest Xvisor-next (https://github.com/avpatel/xvisor-next)
with your QEMU riscv-to-apply.next branch and it works fine (both with and without Sstc).

Does it work with the latest release?

Yes, the latest Xvisor-next repo works for QEMU v7.2.0-rc4 and your riscv-to-apply.next branch (commit 51bb9de2d188)

I can't get anything to work with this patch. I have dropped this and the patches after this.

I'm building the latest Xvisor release with:

Quoted 8 lines
Quoted 5 lines

export CROSS_COMPILE=riscv64-linux-gnu-
ARCH=riscv make generic-64b-defconfig
make

and running it as above, yet nothing. What am I missing here?

I tried multiple times with the latest Xvisor on different machines but still can't reproduce the issue you are seeing.

Odd

Quoted 4 lines

We generally provide pre-built binaries with every Xvisor release so I will share with you pre-built binaries of the upcoming Xvisor-0.3.2 release. Maybe that would help you ?

That would work. Let me know when the release happens and I can update my images.

Alistair

author
#10 lore raw

Re: [PATCH v2 2/5] target/riscv: Update VS timer whenever htimedelta changes

Hi Alistair,

Quoted 123 lines · Alistair Francis
Quoted 109 lines · Anup Patel
Quoted 107 lines · Alistair Francis
Quoted 91 lines · Anup Patel
Quoted 89 lines · Alistair Francis
Quoted 87 lines · Anup Patel
Quoted 84 lines · Alistair Francis
Quoted 7 lines · Anup Patel

The htimedelta[h] CSR has impact on the VS timer comparison so we should call riscv_timer_write_timecmp() whenever htimedelta changes.

Fixes 3ec0fe18a31f ("target/riscv: Add vstimecmp suppor")

This patch breaks my Xvisor test. When running OpenSBI and Xvisor like this:

qemu-system-riscv64 -machine virt \
-m 1G -serial mon:stdio -serial null -nographic \
-append 'vmm.console=uart@10000000 vmm.bootcmd="vfs mount initrd
/;vfs run /boot.xscript;vfs cat /system/banner.txt; guest kick guest0;

vserial bind guest0/uart0"' \
    -smp 4 -d guest_errors \
    -bios none \
    -device loader,file=./images/qemuriscv64/vmm.bin,addr=0x80200000 \
    -kernel ./images/qemuriscv64/fw_jump.elf \
    -initrd ./images/qemuriscv64/vmm-disk-linux.img -cpu rv64,h=true

Running:

Xvisor v0.3.0-129-gbc33f339 (Jan 1 1970 00:00:00)

I see this failure:

INIT: bootcmd: guest kick guest0

guest0: Kicked

INIT: bootcmd: vserial bind guest0/uart0

[guest0/uart0] cpu_vcpu_stage2_map: guest_phys=0x000000003B9AC000 size=0x4096 map failed

do_error: CPU3: VCPU=guest0/vcpu0 page fault failed (error -1)

       zero=0x0000000000000000          ra=0x0000000080001B4E

         sp=0x000000008001CF80          gp=0x0000000000000000

         tp=0x0000000000000000          s0=0x000000008001CFB0

         s1=0x0000000000000000          a0=0x0000000010001048

         a1=0x0000000000000000          a2=0x0000000000989680

         a3=0x000000003B9ACA00          a4=0x0000000000000048

         a5=0x0000000000000000          a6=0x0000000000019000

         a7=0x0000000000000000          s2=0x0000000000000000

         s3=0x0000000000000000          s4=0x0000000000000000

         s5=0x0000000000000000          s6=0x0000000000000000

         s7=0x0000000000000000          s8=0x0000000000000000

         s9=0x0000000000000000         s10=0x0000000000000000

        s11=0x0000000000000000          t0=0x0000000000004000

         t1=0x0000000000000100          t2=0x0000000000000000

         t3=0x0000000000000000          t4=0x0000000000000000

         t5=0x0000000000000000          t6=0x0000000000000000

       sepc=0x0000000080001918     sstatus=0x0000000200004120

    hstatus=0x00000002002001C0     sp_exec=0x0000000010A64000

     scause=0x0000000000000017       stval=0x000000003B9ACAF8

      htval=0x000000000EE6B2BE      htinst=0x0000000000D03021

I have tried updating to a newer Xvisor release, but with that I don't get any serial output.

Can you help get the Xvisor tests back up and running?

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines

I tried the latest Xvisor-next (https://github.com/avpatel/xvisor-next)
with your QEMU riscv-to-apply.next branch and it works fine (both with and without Sstc).

Does it work with the latest release?

Yes, the latest Xvisor-next repo works for QEMU v7.2.0-rc4 and your riscv-to-apply.next branch (commit 51bb9de2d188)

I can't get anything to work with this patch. I have dropped this and the patches after this.

I'm building the latest Xvisor release with:

Quoted 8 lines
Quoted 5 lines

export CROSS_COMPILE=riscv64-linux-gnu-
ARCH=riscv make generic-64b-defconfig
make

and running it as above, yet nothing. What am I missing here?

I tried multiple times with the latest Xvisor on different machines but still can't reproduce the issue you are seeing.

Odd

Quoted 1 line
Quoted 6 lines
Quoted 3 lines

We generally provide pre-built binaries with every Xvisor release so I will share with you pre-built binaries of the upcoming Xvisor-0.3.2 release. Maybe that would help you ?

That would work. Let me know when the release happens and I can update my images.

Please download the Xvisor v0.3.2 pre-built binary tarball from:
https://xhypervisor.org/tarball/xvisor-0.3.2-bins.tar.xz

After untarring the above tarball, you can try the following command:

$ qemu-system-riscv64 -M virt -m 512M -nographic -bios
opensbi/build/platform/generic/firmware/fw_jump.bin -kernel
xvisor-0.3.2-bins/riscv/rv64/xvisor/vmm.bin -initrd
xvisor-0.3.2-bins/riscv/rv64/guest/virt64/disk-linux-6.1.1-one_guest_virt64.ext2
-append "vmm.bootcmd=\"vfs mount initrd /;vfs run /boot.xscript;vfs
cat /system/banner.txt\""
OR
$ qemu-system-riscv32 -M virt -m 512M -nographic -bios
opensbi/build/platform/generic/firmware/fw_jump.bin -kernel
xvisor-0.3.2-bins/riscv/rv32/xvisor/vmm.bin -initrd
xvisor-0.3.2-bins/riscv/rv32/guest/virt32/disk-linux-6.1.1-one_guest_virt32.ext2
-append "vmm.bootcmd=\"vfs mount initrd /;vfs run /boot.xscript;vfs
cat /system/banner.txt\""

Regards,
Anup

author
#11 lore raw

Re: [PATCH v2 2/5] target/riscv: Update VS timer whenever htimedelta changes

Hi Alistair,

Quoted 138 lines · Anup Patel

Hi Alistair,

Quoted 123 lines · Alistair Francis
Quoted 109 lines · Anup Patel
Quoted 107 lines · Alistair Francis
Quoted 91 lines · Anup Patel
Quoted 89 lines · Alistair Francis
Quoted 87 lines · Anup Patel
Quoted 84 lines · Alistair Francis
Quoted 7 lines · Anup Patel

The htimedelta[h] CSR has impact on the VS timer comparison so we should call riscv_timer_write_timecmp() whenever htimedelta changes.

Fixes 3ec0fe18a31f ("target/riscv: Add vstimecmp suppor")

This patch breaks my Xvisor test. When running OpenSBI and Xvisor like this:

qemu-system-riscv64 -machine virt \
-m 1G -serial mon:stdio -serial null -nographic \
-append 'vmm.console=uart@10000000 vmm.bootcmd="vfs mount initrd
/;vfs run /boot.xscript;vfs cat /system/banner.txt; guest kick guest0;

vserial bind guest0/uart0"' \
    -smp 4 -d guest_errors \
    -bios none \
    -device loader,file=./images/qemuriscv64/vmm.bin,addr=0x80200000 \
    -kernel ./images/qemuriscv64/fw_jump.elf \
    -initrd ./images/qemuriscv64/vmm-disk-linux.img -cpu rv64,h=true

Running:

Xvisor v0.3.0-129-gbc33f339 (Jan 1 1970 00:00:00)

I see this failure:

INIT: bootcmd: guest kick guest0

guest0: Kicked

INIT: bootcmd: vserial bind guest0/uart0

[guest0/uart0] cpu_vcpu_stage2_map: guest_phys=0x000000003B9AC000 size=0x4096 map failed

do_error: CPU3: VCPU=guest0/vcpu0 page fault failed (error -1)

       zero=0x0000000000000000          ra=0x0000000080001B4E

         sp=0x000000008001CF80          gp=0x0000000000000000

         tp=0x0000000000000000          s0=0x000000008001CFB0

         s1=0x0000000000000000          a0=0x0000000010001048

         a1=0x0000000000000000          a2=0x0000000000989680

         a3=0x000000003B9ACA00          a4=0x0000000000000048

         a5=0x0000000000000000          a6=0x0000000000019000

         a7=0x0000000000000000          s2=0x0000000000000000

         s3=0x0000000000000000          s4=0x0000000000000000

         s5=0x0000000000000000          s6=0x0000000000000000

         s7=0x0000000000000000          s8=0x0000000000000000

         s9=0x0000000000000000         s10=0x0000000000000000

        s11=0x0000000000000000          t0=0x0000000000004000

         t1=0x0000000000000100          t2=0x0000000000000000

         t3=0x0000000000000000          t4=0x0000000000000000

         t5=0x0000000000000000          t6=0x0000000000000000

       sepc=0x0000000080001918     sstatus=0x0000000200004120

    hstatus=0x00000002002001C0     sp_exec=0x0000000010A64000

     scause=0x0000000000000017       stval=0x000000003B9ACAF8

      htval=0x000000000EE6B2BE      htinst=0x0000000000D03021

I have tried updating to a newer Xvisor release, but with that I don't get any serial output.

Can you help get the Xvisor tests back up and running?

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines

I tried the latest Xvisor-next (https://github.com/avpatel/xvisor-next)
with your QEMU riscv-to-apply.next branch and it works fine (both with and without Sstc).

Does it work with the latest release?

Yes, the latest Xvisor-next repo works for QEMU v7.2.0-rc4 and your riscv-to-apply.next branch (commit 51bb9de2d188)

I can't get anything to work with this patch. I have dropped this and the patches after this.

I'm building the latest Xvisor release with:

Quoted 8 lines
Quoted 5 lines

export CROSS_COMPILE=riscv64-linux-gnu-
ARCH=riscv make generic-64b-defconfig
make

and running it as above, yet nothing. What am I missing here?

I tried multiple times with the latest Xvisor on different machines but still can't reproduce the issue you are seeing.

Odd

Quoted 1 line
Quoted 6 lines
Quoted 3 lines

We generally provide pre-built binaries with every Xvisor release so I will share with you pre-built binaries of the upcoming Xvisor-0.3.2 release. Maybe that would help you ?

That would work. Let me know when the release happens and I can update my images.

Please download the Xvisor v0.3.2 pre-built binary tarball from:
https://xhypervisor.org/tarball/xvisor-0.3.2-bins.tar.xz

After untarring the above tarball, you can try the following command:

Quoted 13 lines
$ qemu-system-riscv64 -M virt -m 512M -nographic -bios
opensbi/build/platform/generic/firmware/fw_jump.bin -kernel
xvisor-0.3.2-bins/riscv/rv64/xvisor/vmm.bin -initrd
xvisor-0.3.2-bins/riscv/rv64/guest/virt64/disk-linux-6.1.1-one_guest_virt64.ext2
-append "vmm.bootcmd=\"vfs mount initrd /;vfs run /boot.xscript;vfs
cat /system/banner.txt\""
OR
$ qemu-system-riscv32 -M virt -m 512M -nographic -bios
opensbi/build/platform/generic/firmware/fw_jump.bin -kernel
xvisor-0.3.2-bins/riscv/rv32/xvisor/vmm.bin -initrd
xvisor-0.3.2-bins/riscv/rv32/guest/virt32/disk-linux-6.1.1-one_guest_virt32.ext2
-append "vmm.bootcmd=\"vfs mount initrd /;vfs run /boot.xscript;vfs
cat /system/banner.txt\""

Do you want me to rebase and resend the patches which are not merged ?

Regards,
Anup

#12 lore raw

Re: [PATCH v2 2/5] target/riscv: Update VS timer whenever htimedelta changes

Quoted 141 lines · Anup Patel

Hi Alistair,

Quoted 137 lines · Anup Patel

Hi Alistair,

Quoted 123 lines · Alistair Francis
Quoted 109 lines · Anup Patel
Quoted 107 lines · Alistair Francis
Quoted 91 lines · Anup Patel
Quoted 89 lines · Alistair Francis
Quoted 87 lines · Anup Patel
Quoted 84 lines · Alistair Francis
Quoted 7 lines · Anup Patel

The htimedelta[h] CSR has impact on the VS timer comparison so we should call riscv_timer_write_timecmp() whenever htimedelta changes.

Fixes 3ec0fe18a31f ("target/riscv: Add vstimecmp suppor")

This patch breaks my Xvisor test. When running OpenSBI and Xvisor like this:

qemu-system-riscv64 -machine virt \
-m 1G -serial mon:stdio -serial null -nographic \
-append 'vmm.console=uart@10000000 vmm.bootcmd="vfs mount initrd
/;vfs run /boot.xscript;vfs cat /system/banner.txt; guest kick guest0;

vserial bind guest0/uart0"' \
    -smp 4 -d guest_errors \
    -bios none \
    -device loader,file=./images/qemuriscv64/vmm.bin,addr=0x80200000 \
    -kernel ./images/qemuriscv64/fw_jump.elf \
    -initrd ./images/qemuriscv64/vmm-disk-linux.img -cpu rv64,h=true

Running:

Xvisor v0.3.0-129-gbc33f339 (Jan 1 1970 00:00:00)

I see this failure:

INIT: bootcmd: guest kick guest0

guest0: Kicked

INIT: bootcmd: vserial bind guest0/uart0

[guest0/uart0] cpu_vcpu_stage2_map: guest_phys=0x000000003B9AC000 size=0x4096 map failed

do_error: CPU3: VCPU=guest0/vcpu0 page fault failed (error -1)

       zero=0x0000000000000000          ra=0x0000000080001B4E

         sp=0x000000008001CF80          gp=0x0000000000000000

         tp=0x0000000000000000          s0=0x000000008001CFB0

         s1=0x0000000000000000          a0=0x0000000010001048

         a1=0x0000000000000000          a2=0x0000000000989680

         a3=0x000000003B9ACA00          a4=0x0000000000000048

         a5=0x0000000000000000          a6=0x0000000000019000

         a7=0x0000000000000000          s2=0x0000000000000000

         s3=0x0000000000000000          s4=0x0000000000000000

         s5=0x0000000000000000          s6=0x0000000000000000

         s7=0x0000000000000000          s8=0x0000000000000000

         s9=0x0000000000000000         s10=0x0000000000000000

        s11=0x0000000000000000          t0=0x0000000000004000

         t1=0x0000000000000100          t2=0x0000000000000000

         t3=0x0000000000000000          t4=0x0000000000000000

         t5=0x0000000000000000          t6=0x0000000000000000

       sepc=0x0000000080001918     sstatus=0x0000000200004120

    hstatus=0x00000002002001C0     sp_exec=0x0000000010A64000

     scause=0x0000000000000017       stval=0x000000003B9ACAF8

      htval=0x000000000EE6B2BE      htinst=0x0000000000D03021

I have tried updating to a newer Xvisor release, but with that I don't get any serial output.

Can you help get the Xvisor tests back up and running?

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines

I tried the latest Xvisor-next (https://github.com/avpatel/xvisor-next)
with your QEMU riscv-to-apply.next branch and it works fine (both with and without Sstc).

Does it work with the latest release?

Yes, the latest Xvisor-next repo works for QEMU v7.2.0-rc4 and your riscv-to-apply.next branch (commit 51bb9de2d188)

I can't get anything to work with this patch. I have dropped this and the patches after this.

I'm building the latest Xvisor release with:

Quoted 8 lines
Quoted 5 lines

export CROSS_COMPILE=riscv64-linux-gnu-
ARCH=riscv make generic-64b-defconfig
make

and running it as above, yet nothing. What am I missing here?

I tried multiple times with the latest Xvisor on different machines but still can't reproduce the issue you are seeing.

Odd

Quoted 1 line
Quoted 6 lines
Quoted 3 lines

We generally provide pre-built binaries with every Xvisor release so I will share with you pre-built binaries of the upcoming Xvisor-0.3.2 release. Maybe that would help you ?

That would work. Let me know when the release happens and I can update my images.

Please download the Xvisor v0.3.2 pre-built binary tarball from:
https://xhypervisor.org/tarball/xvisor-0.3.2-bins.tar.xz

Quoted 14 lines
Quoted 14 lines

After untarring the above tarball, you can try the following command:

$ qemu-system-riscv64 -M virt -m 512M -nographic -bios
opensbi/build/platform/generic/firmware/fw_jump.bin -kernel
xvisor-0.3.2-bins/riscv/rv64/xvisor/vmm.bin -initrd
xvisor-0.3.2-bins/riscv/rv64/guest/virt64/disk-linux-6.1.1-one_guest_virt64.ext2
-append "vmm.bootcmd=\"vfs mount initrd /;vfs run /boot.xscript;vfs
cat /system/banner.txt\""
OR
$ qemu-system-riscv32 -M virt -m 512M -nographic -bios
opensbi/build/platform/generic/firmware/fw_jump.bin -kernel
xvisor-0.3.2-bins/riscv/rv32/xvisor/vmm.bin -initrd
xvisor-0.3.2-bins/riscv/rv32/guest/virt32/disk-linux-6.1.1-one_guest_virt32.ext2
-append "vmm.bootcmd=\"vfs mount initrd /;vfs run /boot.xscript;vfs
cat /system/banner.txt\""

Great! I have updated my tests and they are passing on the current QEMU master.

Quoted 3 lines

Do you want me to rebase and resend the patches which are not merged ?

Yes please :)

Alistair

Quoted 3 lines

Regards,
Anup