s390x: support zpci in tcg
v2 cover s390x: support zpci in tcg
<blurb from last cover letter>
This patch set (available at git://github.com/cohuck/qemu pci-tcg) wires up zpci under tcg. A guest started with
qemu-system-s390x -M s390-ccw-virtio,accel=tcg -cpu qemu,zpci=on -device virtio-net-pci (...)
and a 4.15 kernel in the guest (needed because the requirement for ais was dropped with
commit 48070c73058be6de9c0d754d441ed7092dfc8f12
Author: Christian Borntraeger <[email protected]> Date: Mon Oct 30 14:38:58 2017 +0100
s390/pci: do not require AIS facility
and we don't have a working ais with tcg yet), I get sensible output out of lspci.
</blurb>
Looks much nicer now after implementing feedback from David (thanks!)
Changes RFC->v2:
- no need to check for the zpci feature in the helpers
- SIC does not set a condition code
- SIC also should not depend on zpci
- use tcg magic to improve the code in translate.c
- clear zpci feature from supported set for !CONFIG_PCI
Cornelia Huck (2):
s390x/tcg: wire up pci instructions s390x/cpumodel: allow zpci features in qemu model
target/s390x/cpu_models.c | 6 +++ target/s390x/gen-features.c | 3 ++ target/s390x/helper.h | 9 ++++ target/s390x/insn-data.def | 15 ++++++ target/s390x/misc_helper.c | 109 ++++++++++++++++++++++++++++++++++++++++++++ target/s390x/translate.c | 101 ++++++++++++++++++++++++++++++++++++++++ 6 files changed, 243 insertions(+)
Signature · 1 line
2.13.6
PATCH v2 1/2 s390x/tcg: wire up pci instructions
On s390x, pci support is implemented via a set of instructions (no mmio). Unfortunately, none of them are documented in the PoP; the code is based upon the existing implementation for KVM and the Linux zpci driver.
--- target/s390x/helper.h | 9 ++++ target/s390x/insn-data.def | 15 +++++++ target/s390x/misc_helper.c | 109 +++++++++++++++++++++++++++++++++++++++++++++ target/s390x/translate.c | 101 +++++++++++++++++++++++++++++++++++++++++ 4 files changed, 234 insertions(+)
target/s390x/helper.h +9 -0
@@ -172,4 +172,13 @@ DEF_HELPER_2(stcrw, void, env, i64) DEF_HELPER_3(stsch, void, env, i64, i64) DEF_HELPER_3(tsch, void, env, i64, i64) DEF_HELPER_2(chsc, void, env, i64) + +DEF_HELPER_2(clp, void, env, i32) +DEF_HELPER_3(pcilg, void, env, i32, i32) +DEF_HELPER_3(pcistg, void, env, i32, i32) +DEF_HELPER_4(stpcifc, void, env, i32, i64, i32) +DEF_HELPER_3(sic, void, env, i64, i64) +DEF_HELPER_3(rpcit, void, env, i32, i32) +DEF_HELPER_5(pcistb, void, env, i32, i32, i64, i32) +DEF_HELPER_4(mpcifc, void, env, i32, i64, i32)#endif
target/s390x/insn-data.def +15 -0
@@ -1067,4 +1067,19 @@ /* ??? Not listed in PoO ninth edition, but there's a linux driver that uses it: "A CHSC subchannel is usually present on LPAR only." */ C(0xb25f, CHSC, RRE, Z, 0, insn, 0, 0, chsc, 0) + +/* zPCI Instructions */ + /* None of these instructions are documented in the PoP, so this is all + based upon target/s390x/kvm.c and Linux code and likely incomplete */ + C(0xebd0, PCISTB, RSY_a, PCI, la2, 0, 0, 0, pcistb, 0) + /* SIC does not really depend on pci, but it is unclear from the code + with which facility it becomes available */ + C(0xebd1, SIC, RSY_a, Z, r1, r3, 0, 0, sic, 0) + C(0xb9a0, CLP, RRF_c, PCI, 0, 0, 0, 0, clp, 0) + C(0xb9d0, PCISTG, RRE, PCI, 0, 0, 0, 0, pcistg, 0) + C(0xb9d2, PCILG, RRE, PCI, 0, 0, 0, 0, pcilg, 0) + C(0xb9d3, RPCIT, RRE, PCI, 0, 0, 0, 0, rpcit, 0) + C(0xe3d0, MPCIFC, RXY_a, PCI, la2, 0, 0, 0, mpcifc, 0) + C(0xe3d4, STPCIFC, RXY_a, PCI, la2, 0, 0, 0, stpcifc, 0) + #endif /* CONFIG_USER_ONLY */
target/s390x/misc_helper.c +109 -0
118 lines not on this page · raw message
target/s390x/translate.c +101 -0
113 lines not on this page · raw message
Signature · 1 line
2.13.6
Re: [Qemu-devel] [PATCH v2 1/2] s390x/tcg: wire up pci instructions
Quoted 291 lines · Cornelia Huck
On s390x, pci support is implemented via a set of instructions (no mmio). Unfortunately, none of them are documented in the PoP; the code is based upon the existing implementation for KVM and the Linux zpci driver.
Signed-off-by Cornelia Huck <[email protected]>--- target/s390x/helper.h | 9 ++++ target/s390x/insn-data.def | 15 +++++++ target/s390x/misc_helper.c | 109 +++++++++++++++++++++++++++++++++++++++++++++ target/s390x/translate.c | 101 +++++++++++++++++++++++++++++++++++++++++ 4 files changed, 234 insertions(+)target/s390x/helper.h +9
-0@@ -172,4 +172,13 @@ DEF_HELPER_2(stcrw, void, env, i64) DEF_HELPER_3(stsch, void, env, i64, i64) DEF_HELPER_3(tsch, void, env, i64, i64) DEF_HELPER_2(chsc, void, env, i64) + +DEF_HELPER_2(clp, void, env, i32) +DEF_HELPER_3(pcilg, void, env, i32, i32) +DEF_HELPER_3(pcistg, void, env, i32, i32) +DEF_HELPER_4(stpcifc, void, env, i32, i64, i32) +DEF_HELPER_3(sic, void, env, i64, i64) +DEF_HELPER_3(rpcit, void, env, i32, i32) +DEF_HELPER_5(pcistb, void, env, i32, i32, i64, i32) +DEF_HELPER_4(mpcifc, void, env, i32, i64, i32)#endif target/s390x/insn-data.def +15
-0@@ -1067,4 +1067,19 @@ /* ??? Not listed in PoO ninth edition, but there's a linux driver that uses it: "A CHSC subchannel is usually present on LPAR only." */ C(0xb25f, CHSC, RRE, Z, 0, insn, 0, 0, chsc, 0) + +/* zPCI Instructions */ + /* None of these instructions are documented in the PoP, so this is all + based upon target/s390x/kvm.c and Linux code and likely incomplete */ + C(0xebd0, PCISTB, RSY_a, PCI, la2, 0, 0, 0, pcistb, 0) + /* SIC does not really depend on pci, but it is unclear from the code + with which facility it becomes available */ + C(0xebd1, SIC, RSY_a, Z, r1, r3, 0, 0, sic, 0) + C(0xb9a0, CLP, RRF_c, PCI, 0, 0, 0, 0, clp, 0) + C(0xb9d0, PCISTG, RRE, PCI, 0, 0, 0, 0, pcistg, 0) + C(0xb9d2, PCILG, RRE, PCI, 0, 0, 0, 0, pcilg, 0) + C(0xb9d3, RPCIT, RRE, PCI, 0, 0, 0, 0, rpcit, 0) + C(0xe3d0, MPCIFC, RXY_a, PCI, la2, 0, 0, 0, mpcifc, 0) + C(0xe3d4, STPCIFC, RXY_a, PCI, la2, 0, 0, 0, stpcifc, 0) + #endif /* CONFIG_USER_ONLY */target/s390x/misc_helper.c +109
-0@@ -36,6 +36,7 @@#include "hw/s390x/ebcdic.h" #include "hw/s390x/s390-virtio-hcall.h" #include "hw/s390x/sclp.h" +#include "hw/s390x/s390-pci-inst.h" #endif /* #define DEBUG_HELPER */ @@ -560,3 +561,111 @@ uint32_t HELPER(stfle)(CPUS390XState *env, uint64_t addr) env->regs[0 ] =deposit64 (env->regs[0 ],0 ,8 , (max_bytes /8 ) -1 );return count_bytes >= max_bytes ?0 :3 ; } + +#ifndef CONFIG_USER_ONLY +void HELPER (clp )(CPUS390XState *env,uint32_t r2) +{ +S390CPU *cpu =s390_env_get_cpu (env); +int r; + +qemu_mutex_lock_iothread (); + r =clp_service_call (cpu, r2,GETPC ()); +qemu_mutex_unlock_iothread (); +if (r) { +s390_program_interrupt (env,PGM_OPERATION ,4 ,GETPC ()); + } +} + +void HELPER (pcilg )(CPUS390XState *env,uint32_t r1,uint32_t r2) +{ +S390CPU *cpu =s390_env_get_cpu (env); +int r; + +qemu_mutex_lock_iothread (); + r =pcilg_service_call (cpu, r1, r2,GETPC ()); +qemu_mutex_unlock_iothread (); +if (r) { +s390_program_interrupt (env,PGM_OPERATION ,4 ,GETPC ()); + } +} + +void HELPER (pcistg )(CPUS390XState *env,uint32_t r1,uint32_t r2) +{ +S390CPU *cpu =s390_env_get_cpu (env); +int r; + +qemu_mutex_lock_iothread (); + r =pcistg_service_call (cpu, r1, r2,GETPC ()); +qemu_mutex_unlock_iothread (); +if (r) { +s390_program_interrupt (env,PGM_OPERATION ,4 ,GETPC ()); + } +} + +void HELPER (stpcifc )(CPUS390XState *env,uint32_t r1,uint64_t fiba, +uint32_t ar) +{ +S390CPU *cpu =s390_env_get_cpu (env); +int r; + +qemu_mutex_lock_iothread (); + r =stpcifc_service_call (cpu, r1, fiba, ar,GETPC ()); +qemu_mutex_unlock_iothread (); +if (r) { +s390_program_interrupt (env,PGM_OPERATION ,4 ,GETPC ()); + } +} + +void HELPER (sic )(CPUS390XState *env,uint64_t r1,uint64_t r3) +{ +int r; + +qemu_mutex_lock_iothread (); + r =css_do_sic (env, r1 &0xffff , (r3 >>27 ) &0x7 ); +qemu_mutex_unlock_iothread (); +if (r) { +s390_program_interrupt (env,PGM_OPERATION ,4 ,GETPC ()); + } +} + +void HELPER (rpcit )(CPUS390XState *env,uint32_t r1,uint32_t r2) +{ +S390CPU *cpu =s390_env_get_cpu (env); +int r; + +qemu_mutex_lock_iothread (); + r =rpcit_service_call (cpu, r1, r2,GETPC ()); +qemu_mutex_unlock_iothread (); +if (r) { +s390_program_interrupt (env,PGM_OPERATION ,4 ,GETPC ()); + } +} + +void HELPER (pcistb )(CPUS390XState *env,uint32_t r1,uint32_t r3, +uint64_t gaddr,uint32_t ar) +{ +S390CPU *cpu =s390_env_get_cpu (env); +int r; + +qemu_mutex_lock_iothread (); + r =pcistb_service_call (cpu, r1, r3, gaddr, ar,GETPC ()); +qemu_mutex_unlock_iothread (); +if (r) { +s390_program_interrupt (env,PGM_OPERATION ,4 ,GETPC ()); + } +} + +void HELPER (mpcifc )(CPUS390XState *env,uint32_t r1,uint64_t fiba, +uint32_t ar) +{ +S390CPU *cpu =s390_env_get_cpu (env); +int r; + +qemu_mutex_lock_iothread (); + r =mpcifc_service_call (cpu, r1, fiba, ar,GETPC ()); +qemu_mutex_unlock_iothread (); +if (r) { +s390_program_interrupt (env,PGM_OPERATION ,4 ,GETPC ()); + } +} +#endif target/s390x/translate.c +100
-0@@ -4777,6 +4777,106 @@ static ExitStatus op_zero2(DisasContext *s, DisasOps *o)return NO_EXIT ; } +#ifndef CONFIG_USER_ONLY +static ExitStatus op_clp (DisasContext *s,DisasOps *o) +{ +TCGv_i32 r2 =tcg_const_i32 (get_field (s->fields, r2)); + +check_privileged (s); +gen_helper_clp (cpu_env, r2); +tcg_temp_free_i32 (r2); +set_cc_static (s); +return NO_EXIT ; +} + +static ExitStatus op_pcilg (DisasContext *s,DisasOps *o) +{ +TCGv_i32 r1 =tcg_const_i32 (get_field (s->fields, r1)); +TCGv_i32 r2 =tcg_const_i32 (get_field (s->fields, r2)); + +check_privileged (s); +gen_helper_pcilg (cpu_env, r1, r2); +tcg_temp_free_i32 (r1); +tcg_temp_free_i32 (r2); +set_cc_static (s); +return NO_EXIT ; +} + +static ExitStatus op_pcistg (DisasContext *s,DisasOps *o) +{ +TCGv_i32 r1 =tcg_const_i32 (get_field (s->fields, r1)); +TCGv_i32 r2 =tcg_const_i32 (get_field (s->fields, r2)); + +check_privileged (s); +gen_helper_pcistg (cpu_env, r1, r2); +tcg_temp_free_i32 (r1); +tcg_temp_free_i32 (r2); +set_cc_static (s); +return NO_EXIT ; +} + +static ExitStatus op_stpcifc (DisasContext *s,DisasOps *o) +{ +TCGv_i32 r1 =tcg_const_i32 (get_field (s->fields, r1)); +TCGv_i32 ar =tcg_const_i32 (get_field (s->fields, b2)); + +check_privileged (s); +gen_helper_stpcifc (cpu_env, r1, o->addr1, ar); +tcg_temp_free_i32 (ar); +tcg_temp_free_i32 (r1); +set_cc_static (s); +return NO_EXIT ; +} + +static ExitStatus op_sic (DisasContext *s,DisasOps *o) +{ +check_privileged (s); +gen_helper_sic (cpu_env, o->in1, o->in2); +return NO_EXIT ; +} + +static ExitStatus op_rpcit (DisasContext *s,DisasOps *o) +{ +TCGv_i32 r1 =tcg_const_i32 (get_field (s->fields, r1)); +TCGv_i32 r2 =tcg_const_i32 (get_field (s->fields, r2)); + +check_privileged (s); +gen_helper_rpcit (cpu_env, r1, r2); +tcg_temp_free_i32 (r1); +tcg_temp_free_i32 (r2); +set_cc_static (s); +return NO_EXIT ; +} + +static ExitStatus op_pcistb (DisasContext *s,DisasOps *o) +{ +TCGv_i32 r1 =tcg_const_i32 (get_field (s->fields, r1)); +TCGv_i32 r3 =tcg_const_i32 (get_field (s->fields, r3)); +TCGv_i32 ar =tcg_const_i32 (get_field (s->fields, b2)); + +check_privileged (s); +gen_helper_pcistb (cpu_env, r1, r3, o->addr1, ar); +tcg_temp_free_i32 (ar); +tcg_temp_free_i32 (r1); +tcg_temp_free_i32 (r3); +set_cc_static (s); +return NO_EXIT ; +} + +static ExitStatus op_mpcifc (DisasContext *s,DisasOps *o) +{ +TCGv_i32 r1 =tcg_const_i32 (get_field (s->fields, r1)); +TCGv_i32 ar =tcg_const_i32 (get_field (s->fields, b2)); + +check_privileged (s); +gen_helper_mpcifc (cpu_env, r1, o->addr1, ar); +tcg_temp_free_i32 (ar); +tcg_temp_free_i32 (r1); +set_cc_static (s); +return NO_EXIT ; +} +#endif +/* ====================================================================== */ /* The "Cc OUTput" generators. Given the generated output (and in some cases the original inputs), update the various cc data structures in order to
target/s390x/translate.c enum DisasInsnEnum { · line 5814
Quoted 7 lines
target/s390x/translate.c +1
-0@@ -5708,6 +5808,7 @@ enum DisasInsnEnum {#define FAC_MSA4 S390_FEAT_MSA_EXT_4/* msa-extension-4 facility */ #define FAC_MSA5 S390_FEAT_MSA_EXT_5/* msa-extension-5 facility */ #define FAC_ECT S390_FEAT_EXTRACT_CPU_TIME +#define FAC_PCI S390_FEAT_ZPCI/* z/PCI facility */ static const DisasInsn insn_info[] = { #include "insn-data.def"
Hi,
I am not an expert for TCG but the patches are looking clean to me.
And since I also rapidly tested with multibus, virtio_pci_block, and all went good so:
Regards,
Pierre
Signature · 2 lines
Pierre Morel Linux/KVM/QEMU in Böblingen - Germany
Re: [Qemu-devel] [PATCH v2 1/2] s390x/tcg: wire up pci instructions
Quoted 14 lines · On Thu, 1 Feb 2018 11:00:09 +0100 Pierre Morel <[email protected]>…
Quoted 5 lines · Cornelia Huck
On s390x, pci support is implemented via a set of instructions (no mmio). Unfortunately, none of them are documented in the PoP; the code is based upon the existing implementation for KVM and the Linux zpci driver.
Quoted 7 lines
Signed-off-by Cornelia Huck <[email protected]>--- target/s390x/helper.h | 9 ++++ target/s390x/insn-data.def | 15 +++++++ target/s390x/misc_helper.c | 109 +++++++++++++++++++++++++++++++++++++++++++++ target/s390x/translate.c | 101 +++++++++++++++++++++++++++++++++++++++++ 4 files changed, 234 insertions(+)
Quoted 8 lines
Hi,
I am not an expert for TCG but the patches are looking clean to me.
And since I also rapidly tested with multibus,
virtio_pci_block, and all went good so:Reviewed-by Pierre Morel <[email protected]>
Thanks!
Out of curiousity, do you have some scripts or instructions for testing you can share?
Re: [Qemu-devel] [PATCH v2 1/2] s390x/tcg: wire up pci instructions
In reply to Cornelia HuckQuoted 36 lines · Cornelia Huck
On s390x, pci support is implemented via a set of instructions (no mmio). Unfortunately, none of them are documented in the PoP; the code is based upon the existing implementation for KVM and the Linux zpci driver.
Signed-off-by Cornelia Huck <[email protected]>--- target/s390x/helper.h | 9 ++++ target/s390x/insn-data.def | 15 +++++++ target/s390x/misc_helper.c | 109 +++++++++++++++++++++++++++++++++++++++++++++ target/s390x/translate.c | 101 +++++++++++++++++++++++++++++++++++++++++ 4 files changed, 234 insertions(+)target/s390x/helper.h +9
-0@@ -172,4 +172,13 @@ DEF_HELPER_2(stcrw, void, env, i64) DEF_HELPER_3(stsch, void, env, i64, i64) DEF_HELPER_3(tsch, void, env, i64, i64) DEF_HELPER_2(chsc, void, env, i64) + +DEF_HELPER_2(clp, void, env, i32) +DEF_HELPER_3(pcilg, void, env, i32, i32) +DEF_HELPER_3(pcistg, void, env, i32, i32) +DEF_HELPER_4(stpcifc, void, env, i32, i64, i32) +DEF_HELPER_3(sic, void, env, i64, i64) +DEF_HELPER_3(rpcit, void, env, i32, i32) +DEF_HELPER_5(pcistb, void, env, i32, i32, i64, i32) +DEF_HELPER_4(mpcifc, void, env, i32, i64, i32)#endif target/s390x/insn-data.def +0
-0
target/s390x/insn-data.def line 1073
Quoted 7 lines
target/s390x/insn-data.def +4
-0@@ -1067,4 +1067,19 @@ /* ??? Not listed in PoO ninth edition, but there's a linux driver that uses it: "A CHSC subchannel is usually present on LPAR only." */ C(0xb25f, CHSC, RRE, Z, 0, insn, 0, 0, chsc, 0) + +/* zPCI Instructions */ + /* None of these instructions are documented in the PoP, so this is all + based upon target/s390x/kvm.c and Linux code and likely incomplete */
I'd drop this comment and rather move it to the patch description.
Quoted 30 lines
target/s390x/insn-data.def +10
-0+ C(0xebd0, PCISTB, RSY_a, PCI, la2, 0, 0, 0, pcistb, 0) + /* SIC does not really depend on pci, but it is unclear from the code + with which facility it becomes available */> + C(0xebd1, SIC, RSY_a, Z, r1, r3, 0, 0, sic, 0) + C(0xb9a0, CLP, RRF_c, PCI, 0, 0, 0, 0, clp, 0) + C(0xb9d0, PCISTG, RRE, PCI, 0, 0, 0, 0, pcistg, 0) + C(0xb9d2, PCILG, RRE, PCI, 0, 0, 0, 0, pcilg, 0) + C(0xb9d3, RPCIT, RRE, PCI, 0, 0, 0, 0, rpcit, 0) + C(0xe3d0, MPCIFC, RXY_a, PCI, la2, 0, 0, 0, mpcifc, 0) + C(0xe3d4, STPCIFC, RXY_a, PCI, la2, 0, 0, 0, stpcifc, 0) + #endif /* CONFIG_USER_ONLY */target/s390x/misc_helper.c +4
-0@@ -36,6 +36,7 @@#include "hw/s390x/ebcdic.h" #include "hw/s390x/s390-virtio-hcall.h" #include "hw/s390x/sclp.h" +#include "hw/s390x/s390-pci-inst.h" #endif /* #define DEBUG_HELPER */ @@ -560,3 +561,111 @@ uint32_t HELPER(stfle)(CPUS390XState *env, uint64_t addr) env->regs[0 ] =deposit64 (env->regs[0 ],0 ,8 , (max_bytes /8 ) -1 );return count_bytes >= max_bytes ?0 :3 ; } + +#ifndef CONFIG_USER_ONLY +void HELPER (clp )(CPUS390XState *env,uint32_t r2)
target/s390x/misc_helper.c uint32_t HELPER(stfle)(CPUS390XState *env, uint64_t addr) · line 576
Quoted 10 lines
target/s390x/misc_helper.c +10
-0+{ +S390CPU *cpu =s390_env_get_cpu (env); +int r; + +qemu_mutex_lock_iothread (); + r =clp_service_call (cpu, r2,GETPC ()); +qemu_mutex_unlock_iothread (); +if (r) { +s390_program_interrupt (env,PGM_OPERATION ,4 ,GETPC ()); + }
We don't need the if (r) ... so I suggest dropping all these. (as I said, will be handled later via the generic flag checking in translation code). We can ignore any error from these functions.
A sane guest will newer trigger this. (if we have no CONFIG_PCI, the also the ZPCI feature will not be available)
Makes the code even shorter :)
Signature · 3 lines
Thanks, David / dhildenb
Re: [Qemu-devel] [PATCH v2 1/2] s390x/tcg: wire up pci instructions
Quoted 18 lines · On Thu, 1 Feb 2018 13:42:52 +0100 David Hildenbrand <[email protected]> wrote:
target/s390x/misc_helper.c uint32_t HELPER(stfle)(CPUS390XState *env, uint64_t addr) · line 576
Quoted 12 lines · Cornelia Huck
target/s390x/misc_helper.c +12
-0+#ifndef CONFIG_USER_ONLY +void HELPER (clp )(CPUS390XState *env,uint32_t r2) +{ +S390CPU *cpu =s390_env_get_cpu (env); +int r; + +qemu_mutex_lock_iothread (); + r =clp_service_call (cpu, r2,GETPC ()); +qemu_mutex_unlock_iothread (); +if (r) { +s390_program_interrupt (env,PGM_OPERATION ,4 ,GETPC ()); + }We don't need the if (r) ... so I suggest dropping all these. (as I said, will be handled later via the generic flag checking in translation code). We can ignore any error from these functions.
I did not check the instruction implementations in detail... was the error really only for the !CONFIG_PCI case?
(I really should know that...)
Quoted 2 lines
A sane guest will newer trigger this. (if we have no
CONFIG_PCI, the also the ZPCI feature will not be available)
Hopefully we can also handle non-sane guests correctly...
Quoted 3 lines
Makes the code even shorter :)
Re: [Qemu-devel] [PATCH v2 1/2] s390x/tcg: wire up pci instructions
Quoted 22 lines · Cornelia Huck
Quoted 18 lines · On Thu, 1 Feb 2018 13:42:52 +0100 David Hildenbrand <[email protected]> wrote:
Review
Quoted 12 lines · Cornelia Huck
(quoted diff) +12
-0+#ifndef CONFIG_USER_ONLY +void HELPER(clp)(CPUS390XState *env, uint32_t r2) +{ + S390CPU *cpu = s390_env_get_cpu(env); + int r; + + qemu_mutex_lock_iothread(); + r = clp_service_call(cpu, r2, GETPC()); + qemu_mutex_unlock_iothread(); + if (r) { + s390_program_interrupt(env, PGM_OPERATION, 4, GETPC()); + }We don't need the if (r) ... so I suggest dropping all these. (as I said, will be handled later via the generic flag checking in translation code). We can ignore any error from these functions.
Quoted 2 lines
I did not check the instruction implementations in detail... was the error really only for the !CONFIG_PCI case?
!FEATURE_ZPCI (which includes !CONFIG_PCI)
Yes, that's how I remember.
Quoted 7 lines
(I really should know that...)
Quoted 2 lines
A sane guest will newer trigger this. (if we have no
CONFIG_PCI, the also the ZPCI feature will not be available)Hopefully we can also handle non-sane guests correctly...
Usually, if you call instructions that are not indicated via STFL, there is no guarantee what will happen. Some time in the future, we will handle this globally (but haven't done so to allow new applications to run with old CPU models - which was necessary before we bumped the CPU model to a z12).
If you don't want to wait until that support is added, you should be save with something like this:
target/s390x/translate.c +0 -0
@@ -5907,6 +5907,10 @@ static ExitStatus translate_one(CPUS390XState
*env, DisasContext *s)
gen_illegal_opcode(s);
return EXIT_NORETURN;
}+ if (insn->fac == FAC_PCI && !s390_has_feat(FAC_PCI)) {
+ gen_illegal_opcode(s);
+ return EXIT_NORETURN;
+ }
#ifndef CONFIG_USER_ONLY
if (s->tb->flags & FLAG_MASK_PER) {Signature · 3 lines
Thanks, David / dhildenb
Re: [Qemu-devel] [PATCH v2 1/2] s390x/tcg: wire up pci instructions
Quoted 63 lines · On Thu, 1 Feb 2018 13:59:13 +0100 David Hildenbrand <[email protected]> wrote:
Quoted 22 lines · Cornelia Huck
Quoted 18 lines · On Thu, 1 Feb 2018 13:42:52 +0100 David Hildenbrand <[email protected]> wrote:
Review
Quoted 12 lines · Cornelia Huck
(quoted diff) +12
-0+#ifndef CONFIG_USER_ONLY +void HELPER(clp)(CPUS390XState *env, uint32_t r2) +{ + S390CPU *cpu = s390_env_get_cpu(env); + int r; + + qemu_mutex_lock_iothread(); + r = clp_service_call(cpu, r2, GETPC()); + qemu_mutex_unlock_iothread(); + if (r) { + s390_program_interrupt(env, PGM_OPERATION, 4, GETPC()); + }We don't need the if (r) ... so I suggest dropping all these. (as I said, will be handled later via the generic flag checking in translation code). We can ignore any error from these functions.
Quoted 2 lines
I did not check the instruction implementations in detail... was the error really only for the !CONFIG_PCI case?
!FEATURE_ZPCI (which includes !CONFIG_PCI)
Yes, that's how I remember.
Quoted 7 lines
(I really should know that...)
Quoted 2 lines
A sane guest will newer trigger this. (if we have no
CONFIG_PCI, the also the ZPCI feature will not be available)Hopefully we can also handle non-sane guests correctly...
Usually, if you call instructions that are not indicated via STFL, there is no guarantee what will happen. Some time in the future, we will handle this globally (but haven't done so to allow new applications to run with old CPU models - which was necessary before we bumped the CPU model to a z12).
If you don't want to wait until that support is added, you should be save with something like this:
target/s390x/translate.c +4
-0@@ -5907,6 +5907,10 @@ static ExitStatus translate_one(CPUS390XState *env, DisasContext *s)gen_illegal_opcode (s);return EXIT_NORETURN ; } +if (insn->fac ==FAC_PCI && !s390_has_feat (FAC_PCI )) { +gen_illegal_opcode (s); +return EXIT_NORETURN ; + }#ifndef CONFIG_USER_ONLY
Quoted 2 lines
if (s->tb->flags & FLAG_MASK_PER) {I think I'll just keep it (the instructions won't do anything, as no pci devices can be added without FEATURE_ZPCI) and just add a comment.
Re: [Qemu-devel] [PATCH v2 1/2] s390x/tcg: wire up pci instructions
In reply to Cornelia Hucktarget/s390x/misc_helper.c uint32_t HELPER(stfle)(CPUS390XState *env, uint64_t addr) · line 624
Quoted 6 lines · On Wed, 31 Jan 2018 19:17:41 +0100 Cornelia Huck <[email protected]> wrote:
target/s390x/misc_helper.c +6
-0+void HELPER (sic )(CPUS390XState *env,uint64_t r1,uint64_t r3) +{ +int r; + +qemu_mutex_lock_iothread (); + r =css_do_sic (env, r1 &0xffff , (r3 >>27 ) &0x7 );
Here, the arguments have to be swapped around...
target/s390x/misc_helper.c uint32_t HELPER(stfle)(CPUS390XState *env, uint64_t addr) · line 575
Quoted 3 lines
target/s390x/misc_helper.c +3
-0+qemu_mutex_unlock_iothread (); +if (r) { +s390_program_interrupt (env,PGM_OPERATION ,4 ,GETPC ());
...and here we have to pass -r instead of PGM_OPERATION. (Thanks to David for spotting those!)
Review
Quoted 2 lines
(quoted diff) +2
-0+ } +}
With that fixed (and rebased upon current s390-next, which contains David's latest changes), I can also turn on ais for tcg (which means pre-4.15 kernels can see pci devices as well.)
PATCH v2 2/2 s390x/cpumodel: allow zpci features in qemu model
In reply to Cornelia HuckAEN can be provided unconditionally, ZPCI should be turned on manually.
With -cpu qemu,zpci=on, a 4.15 guest kernel can now successfully detect virtio-pci devices under tcg.
--- target/s390x/cpu_models.c | 6 ++++++ target/s390x/gen-features.c | 3 +++ 2 files changed, 9 insertions(+)
target/s390x/cpu_models.c +6 -0
@@ -23,6 +23,7 @@#include "qapi/qmp/qbool.h" #ifndef CONFIG_USER_ONLY #include "sysemu/arch_init.h" +#include "hw/pci/pci.h" #endif #define CPUDEF_INIT (_type, _gen, _ec_ga, _mha_pow, _hmfai, _name, _desc) \
@@ -1271,6 +1272,11 @@ static void register_types(void)/* init all bitmaps from gnerated data initially */ s390_init_feat_bitmap (qemu_max_cpu_feat_init, qemu_max_cpu_feat); +#ifndef CONFIG_USER_ONLY +if (!pci_available) { +clear_bit (S390_FEAT_ZPCI , qemu_max_cpu_feat); + } +#endif for (i =0 ; i <ARRAY_SIZE (s390_cpu_defs); i++) {s390_init_feat_bitmap (s390_cpu_defs[i].base_init, s390_cpu_defs[i].base_feat);
target/s390x/gen-features.c +3 -0
@@ -574,6 +574,7 @@ static uint16_t qemu_LATEST[] = {S390_FEAT_INTERLOCKED_ACCESS_2 ,S390_FEAT_MSA_EXT_4 ,S390_FEAT_MSA_EXT_3 , + S390_FEAT_ADAPTER_EVENT_NOTIFICATION, };/* add all new definitions before this point */
@@ -582,6 +583,8 @@ static uint16_t qemu_MAX[] = {S390_FEAT_STFLE_53 ,/* generates a dependency warning, leave it out for now */ S390_FEAT_MSA_EXT_5 , +/* only with CONFIG_PCI */ + S390_FEAT_ZPCI, };/****** END FEATURE DEFS ******/
Signature · 1 line
2.13.6
Re: [Qemu-devel] [PATCH v2 2/2] s390x/cpumodel: allow zpci features in qemu model
Quoted 41 lines · Cornelia Huck
AEN can be provided unconditionally, ZPCI should be turned on manually.
With -cpu qemu,zpci=on, a 4.15 guest kernel can now successfully detect virtio-pci devices under tcg.
Signed-off-by Cornelia Huck <[email protected]>--- target/s390x/cpu_models.c | 6 ++++++ target/s390x/gen-features.c | 3 +++ 2 files changed, 9 insertions(+)target/s390x/cpu_models.c +6
-0@@ -23,6 +23,7 @@#include "qapi/qmp/qbool.h" #ifndef CONFIG_USER_ONLY #include "sysemu/arch_init.h" +#include "hw/pci/pci.h" #endif #define CPUDEF_INIT (_type, _gen, _ec_ga, _mha_pow, _hmfai, _name, _desc) \@@ -1271,6 +1272,11 @@ static void register_types(void)/* init all bitmaps from gnerated data initially */ s390_init_feat_bitmap (qemu_max_cpu_feat_init, qemu_max_cpu_feat); +#ifndef CONFIG_USER_ONLY +if (!pci_available) { +clear_bit (S390_FEAT_ZPCI , qemu_max_cpu_feat); + } +#endif for (i =0 ; i <ARRAY_SIZE (s390_cpu_defs); i++) {s390_init_feat_bitmap (s390_cpu_defs[i].base_init, s390_cpu_defs[i].base_feat);target/s390x/gen-features.c +0
-0
target/s390x/gen-features.c static uint16_t qemu_LATEST[] = { · line 577
Quoted 4 lines
target/s390x/gen-features.c +1
-0@@ -574,6 +574,7 @@ static uint16_t qemu_LATEST[] = {S390_FEAT_INTERLOCKED_ACCESS_2 ,S390_FEAT_MSA_EXT_4 ,S390_FEAT_MSA_EXT_3 , + S390_FEAT_ADAPTER_EVENT_NOTIFICATION,
Minor nit, according to the order in target/s390x/cpu_features_def.h, this should go before FEAT_MSA_EXT_3 (yes, the order of EXT3/EXT4 is messed up)
target/s390x/gen-features.c static uint16_t qemu_MAX[] = { · line 590
Quoted 13 lines
target/s390x/gen-features.c +2
-0};/* add all new definitions before this point */ @@ -582,6 +583,8 @@ static uint16_t qemu_MAX[] = {S390_FEAT_STFLE_53 ,/* generates a dependency warning, leave it out for now */ S390_FEAT_MSA_EXT_5 , +/* only with CONFIG_PCI */ + S390_FEAT_ZPCI, };/****** END FEATURE DEFS ******/
Signature · 3 lines
Thanks, David / dhildenb
Re: [Qemu-devel] [PATCH v2 2/2] s390x/cpumodel: allow zpci features in qemu model
Quoted 14 lines · On Thu, 1 Feb 2018 13:31:46 +0100 David Hildenbrand <[email protected]> wrote:
Quoted 12 lines · Cornelia Huck
AEN can be provided unconditionally, ZPCI should be turned on manually.
With -cpu qemu,zpci=on, a 4.15 guest kernel can now successfully detect virtio-pci devices under tcg.
Signed-off-by Cornelia Huck <[email protected]>--- target/s390x/cpu_models.c | 6 ++++++ target/s390x/gen-features.c | 3 +++ 2 files changed, 9 insertions(+)
Quoted 13 lines
target/s390x/gen-features.c static uint16_t qemu_LATEST[] = { · line 577
Quoted 9 lines
target/s390x/gen-features.c +1
-0@@ -574,6 +574,7 @@ static uint16_t qemu_LATEST[] = {S390_FEAT_INTERLOCKED_ACCESS_2 ,S390_FEAT_MSA_EXT_4 ,S390_FEAT_MSA_EXT_3 , + S390_FEAT_ADAPTER_EVENT_NOTIFICATION,Minor nit, according to the order in
target/s390x/cpu_features_def.h, this should go beforeFEAT_MSA_EXT_3(yes, the order of EXT3/EXT4 is messed up)
Will reorder.
Quoted 12 lines
Quoted 12 lines
target/s390x/gen-features.c +2
-0};/* add all new definitions before this point */ @@ -582,6 +583,8 @@ static uint16_t qemu_MAX[] = {S390_FEAT_STFLE_53 ,/* generates a dependency warning, leave it out for now */ S390_FEAT_MSA_EXT_5 , +/* only with CONFIG_PCI */ + S390_FEAT_ZPCI, };
Quoted 5 lines
target/s390x/gen-features.c static uint16_t qemu_MAX[] = { · line 590
Quoted 2 lines
target/s390x/gen-features.c +0
-0/****** END FEATURE DEFS ******/ Reviewed-by David Hildenbrand <[email protected]>
Thanks!