Re: [Qemu-devel] [PATCH v2] target-cris: update CPU state save/load to use VMStateDescription
On 12 August 2015 at 15:15, Edgar E. Iglesiaswrote: > On Fri, Aug 07, 2015 at 05:02:14PM +0100, Peter Maydell wrote: >> From: Juan Quintela >> >> Update the CRIS CPU state save/load to use a VMStateDescription struct >> rather than cpu_save/cpu_load functions. >> >> Have to define TLBSet struct. >> Multidimensional arrays in C are a mess, just unroll them. >> >> Signed-off-by: Juan Quintela > > > Acked-by: Edgar E. Iglesias So what's the process for target-cris changes to get into the tree? thanks -- PMM
Re: [Qemu-devel] [PATCH v2] target-cris: update CPU state save/load to use VMStateDescription
On Thu, Sep 17, 2015 at 01:39:17PM +0100, Peter Maydell wrote: > On 12 August 2015 at 15:15, Edgar E. Iglesias> wrote: > > On Fri, Aug 07, 2015 at 05:02:14PM +0100, Peter Maydell wrote: > >> From: Juan Quintela > >> > >> Update the CRIS CPU state save/load to use a VMStateDescription struct > >> rather than cpu_save/cpu_load functions. > >> > >> Have to define TLBSet struct. > >> Multidimensional arrays in C are a mess, just unroll them. > >> > >> Signed-off-by: Juan Quintela > > > > > > Acked-by: Edgar E. Iglesias > > So what's the process for target-cris changes to get into the tree? > Sorry, I thought this was maybe going through migration tree. Feel free to push it. Best regards, Edgar
Re: [Qemu-devel] [PATCH v2] target-cris: update CPU state save/load to use VMStateDescription
On 17 September 2015 at 14:04, Edgar E. Iglesiaswrote: > On Thu, Sep 17, 2015 at 01:39:17PM +0100, Peter Maydell wrote: >> On 12 August 2015 at 15:15, Edgar E. Iglesias >> wrote: >> > On Fri, Aug 07, 2015 at 05:02:14PM +0100, Peter Maydell wrote: >> >> From: Juan Quintela >> >> >> >> Update the CRIS CPU state save/load to use a VMStateDescription struct >> >> rather than cpu_save/cpu_load functions. >> >> >> >> Have to define TLBSet struct. >> >> Multidimensional arrays in C are a mess, just unroll them. >> >> >> >> Signed-off-by: Juan Quintela >> > >> > >> > Acked-by: Edgar E. Iglesias >> >> So what's the process for target-cris changes to get into the tree? >> > > Sorry, I thought this was maybe going through migration tree. Feel free > to push it. OK, applied to master. thanks -- PMM
Re: [Qemu-devel] [PATCH v2] target-cris: update CPU state save/load to use VMStateDescription
On Fri, Aug 07, 2015 at 05:02:14PM +0100, Peter Maydell wrote: From: Juan Quintela quint...@redhat.com Update the CRIS CPU state save/load to use a VMStateDescription struct rather than cpu_save/cpu_load functions. Have to define TLBSet struct. Multidimensional arrays in C are a mess, just unroll them. Signed-off-by: Juan Quintela quint...@redhat.com Acked-by: Edgar E. Iglesias edgar.igles...@xilinx.com [PMM: * expand commit message a little since it's no longer one patch in a 35-patch series * add header/copyright comment to machine.c; credited copyright is Red Hat and author is Juan, since this commit gives the file all-new contents; license is LGPL-2-or-later, to match other target-cris code * remove hardcoded tab * add fields for locked_irq, interrupt_vector, fault_vector, trap_vector * drop minimum_version_id_old fields * bump version_id to 2 as we are not compatible with old state format * remove unnecessary hw/boards.h include * update to register via dc-vmsd] Signed-off-by: Peter Maydell peter.mayd...@linaro.org --- This is a patch of Juan's from way back in 2012, which I am resurrecting because we now have only two CPUs which still use old-style non-VMState save/load (CRIS and SPARC). If we can update them both we can drop the machinery in the common code which supports this. Notes: * CPUCRISState indent style is somewhat mismatched (cf load_info); I took the minimal make checkpatch happy approach, but am happy to do something else with the line changed here * I have added a copyright header to target-cris/machine.c, because it did not have one at all, and this commit effectively gives the file all-new contents. I have set it up as LGPLv2-or-later, copyright Red Hat, author Juan. Please let me know if you would prefer something else! * I added vmstate entries for four fields which did not previously get saved and restored, which is presumably a bug fix... * vmsave/vmload on the axis-dev88 board does not currently seem to work (among other obvious problems, there is no vmstate support in the interrupt controller), so we're limited to looks good on code review here. Changes v1-v2: * actually register the vmstate struct, by setting dc-vmsd (discussion on IRC with Andreas a few weeks back indicated that dc-vmsd is preferred over cc-vmsd for new CPUs) * update it to be the right format for a dc-vmsd struct Testing: I checked with a host gdb that we seem to be filling in the registers with the right values. Since none of the devices in the axis model (including the interrupt controller) support vmstate save/load, the run after load from snapshot goes wrong pretty quickly, obviously. target-cris/cpu-qom.h | 4 ++ target-cris/cpu.c | 1 + target-cris/cpu.h | 13 ++-- target-cris/machine.c | 167 +- 4 files changed, 95 insertions(+), 90 deletions(-) diff --git a/target-cris/cpu-qom.h b/target-cris/cpu-qom.h index 6fc30c2..df4c0b5 100644 --- a/target-cris/cpu-qom.h +++ b/target-cris/cpu-qom.h @@ -73,6 +73,10 @@ static inline CRISCPU *cris_env_get_cpu(CPUCRISState *env) #define ENV_OFFSET offsetof(CRISCPU, env) +#ifndef CONFIG_USER_ONLY +extern const struct VMStateDescription vmstate_cris_cpu; +#endif + void cris_cpu_do_interrupt(CPUState *cpu); void crisv10_cpu_do_interrupt(CPUState *cpu); bool cris_cpu_exec_interrupt(CPUState *cpu, int int_req); diff --git a/target-cris/cpu.c b/target-cris/cpu.c index b17e849..d461e07 100644 --- a/target-cris/cpu.c +++ b/target-cris/cpu.c @@ -302,6 +302,7 @@ static void cris_cpu_class_init(ObjectClass *oc, void *data) cc-handle_mmu_fault = cris_cpu_handle_mmu_fault; #else cc-get_phys_page_debug = cris_cpu_get_phys_page_debug; +dc-vmsd = vmstate_cris_cpu; #endif cc-gdb_num_core_regs = 49; diff --git a/target-cris/cpu.h b/target-cris/cpu.h index d422e35..6760dc6 100644 --- a/target-cris/cpu.h +++ b/target-cris/cpu.h @@ -108,6 +108,11 @@ #define NB_MMU_MODES 2 +typedef struct { +uint32_t hi; +uint32_t lo; +} TLBSet; + typedef struct CPUCRISState { uint32_t regs[16]; /* P0 - P15 are referred to as special registers in the docs. */ @@ -161,11 +166,7 @@ typedef struct CPUCRISState { * * One for I and another for D. */ - struct - { - uint32_t hi; - uint32_t lo; - } tlbsets[2][4][16]; +TLBSet tlbsets[2][4][16]; CPU_COMMON @@ -227,8 +228,6 @@ enum { #define cpu_gen_code cpu_cris_gen_code #define cpu_signal_handler cpu_cris_signal_handler -#define CPU_SAVE_VERSION 1 - /* MMU modes definitions */ #define MMU_MODE0_SUFFIX _kernel #define MMU_MODE1_SUFFIX _user diff --git a/target-cris/machine.c b/target-cris/machine.c index 8f9c0dd..983b67c 100644 --- a/target-cris/machine.c +++
[Qemu-devel] [PATCH v2] target-cris: update CPU state save/load to use VMStateDescription
From: Juan Quintela quint...@redhat.com Update the CRIS CPU state save/load to use a VMStateDescription struct rather than cpu_save/cpu_load functions. Have to define TLBSet struct. Multidimensional arrays in C are a mess, just unroll them. Signed-off-by: Juan Quintela quint...@redhat.com [PMM: * expand commit message a little since it's no longer one patch in a 35-patch series * add header/copyright comment to machine.c; credited copyright is Red Hat and author is Juan, since this commit gives the file all-new contents; license is LGPL-2-or-later, to match other target-cris code * remove hardcoded tab * add fields for locked_irq, interrupt_vector, fault_vector, trap_vector * drop minimum_version_id_old fields * bump version_id to 2 as we are not compatible with old state format * remove unnecessary hw/boards.h include * update to register via dc-vmsd] Signed-off-by: Peter Maydell peter.mayd...@linaro.org --- This is a patch of Juan's from way back in 2012, which I am resurrecting because we now have only two CPUs which still use old-style non-VMState save/load (CRIS and SPARC). If we can update them both we can drop the machinery in the common code which supports this. Notes: * CPUCRISState indent style is somewhat mismatched (cf load_info); I took the minimal make checkpatch happy approach, but am happy to do something else with the line changed here * I have added a copyright header to target-cris/machine.c, because it did not have one at all, and this commit effectively gives the file all-new contents. I have set it up as LGPLv2-or-later, copyright Red Hat, author Juan. Please let me know if you would prefer something else! * I added vmstate entries for four fields which did not previously get saved and restored, which is presumably a bug fix... * vmsave/vmload on the axis-dev88 board does not currently seem to work (among other obvious problems, there is no vmstate support in the interrupt controller), so we're limited to looks good on code review here. Changes v1-v2: * actually register the vmstate struct, by setting dc-vmsd (discussion on IRC with Andreas a few weeks back indicated that dc-vmsd is preferred over cc-vmsd for new CPUs) * update it to be the right format for a dc-vmsd struct Testing: I checked with a host gdb that we seem to be filling in the registers with the right values. Since none of the devices in the axis model (including the interrupt controller) support vmstate save/load, the run after load from snapshot goes wrong pretty quickly, obviously. target-cris/cpu-qom.h | 4 ++ target-cris/cpu.c | 1 + target-cris/cpu.h | 13 ++-- target-cris/machine.c | 167 +- 4 files changed, 95 insertions(+), 90 deletions(-) diff --git a/target-cris/cpu-qom.h b/target-cris/cpu-qom.h index 6fc30c2..df4c0b5 100644 --- a/target-cris/cpu-qom.h +++ b/target-cris/cpu-qom.h @@ -73,6 +73,10 @@ static inline CRISCPU *cris_env_get_cpu(CPUCRISState *env) #define ENV_OFFSET offsetof(CRISCPU, env) +#ifndef CONFIG_USER_ONLY +extern const struct VMStateDescription vmstate_cris_cpu; +#endif + void cris_cpu_do_interrupt(CPUState *cpu); void crisv10_cpu_do_interrupt(CPUState *cpu); bool cris_cpu_exec_interrupt(CPUState *cpu, int int_req); diff --git a/target-cris/cpu.c b/target-cris/cpu.c index b17e849..d461e07 100644 --- a/target-cris/cpu.c +++ b/target-cris/cpu.c @@ -302,6 +302,7 @@ static void cris_cpu_class_init(ObjectClass *oc, void *data) cc-handle_mmu_fault = cris_cpu_handle_mmu_fault; #else cc-get_phys_page_debug = cris_cpu_get_phys_page_debug; +dc-vmsd = vmstate_cris_cpu; #endif cc-gdb_num_core_regs = 49; diff --git a/target-cris/cpu.h b/target-cris/cpu.h index d422e35..6760dc6 100644 --- a/target-cris/cpu.h +++ b/target-cris/cpu.h @@ -108,6 +108,11 @@ #define NB_MMU_MODES 2 +typedef struct { +uint32_t hi; +uint32_t lo; +} TLBSet; + typedef struct CPUCRISState { uint32_t regs[16]; /* P0 - P15 are referred to as special registers in the docs. */ @@ -161,11 +166,7 @@ typedef struct CPUCRISState { * * One for I and another for D. */ - struct - { - uint32_t hi; - uint32_t lo; - } tlbsets[2][4][16]; +TLBSet tlbsets[2][4][16]; CPU_COMMON @@ -227,8 +228,6 @@ enum { #define cpu_gen_code cpu_cris_gen_code #define cpu_signal_handler cpu_cris_signal_handler -#define CPU_SAVE_VERSION 1 - /* MMU modes definitions */ #define MMU_MODE0_SUFFIX _kernel #define MMU_MODE1_SUFFIX _user diff --git a/target-cris/machine.c b/target-cris/machine.c index 8f9c0dd..983b67c 100644 --- a/target-cris/machine.c +++ b/target-cris/machine.c @@ -1,90 +1,91 @@ -#include hw/hw.h -#include hw/boards.h - -void cpu_save(QEMUFile *f, void *opaque) -{ -CPUCRISState *env = opaque; -int i; -int s; -int mmu; - -for (i = 0; i 16; i++) -