system/physmem: Remove cpu_physical_memory _is_io() and _rw()
Subsystems (2)
v3 cover system/physmem: Remove cpu_physical_memory _is_io() and _rw()
Since v1:
- Removed extra 'len' arg in address_space_is_io (rth) Since v2:
- Fixed vhost change
- Better describe cpu_physical_memory_rw() removal (thuth)
---
The cpu_physical_memory API is legacy (see commit b7ecba0f6f6):
``cpu_physical_memory_*``
~~~~~~~~~~~~~~~~~~~~~~~~~
These are convenience functions which are identical to ``address_space_*`` but operate specifically on the system address space, always pass a ``MEMTXATTRS_UNSPECIFIED`` set of memory attributes and ignore whether the memory transaction succeeded or failed. For new code they are better avoided:
...
This series removes:
- cpu_physical_memory_is_io()
- cpu_physical_memory_rw()
and start converting some
- cpu_physical_memory_map()
- cpu_physical_memory_unmap()
calls.
Based-on: <[email protected]>
"system/memory: Split address_space_write_rom_internal"
Philippe Mathieu-Daudé (18):
docs/devel/loads-stores: Stop mentioning
cpu_physical_memory_write_rom()
system/memory: Better describe @plen argument of flatview_translate() system/memory: Factor address_space_is_io() out target/i386/arch_memory_mapping: Use address_space_memory_is_io() hw/s390x/sclp: Use address_space_memory_is_io() in sclp_service_call() system/physmem: Remove cpu_physical_memory_is_io() system/physmem: Pass address space argument to
cpu_flush_icache_range()
hw/s390x/sclp: Replace [cpu_physical_memory -> address_space]_r/w() target/s390x/mmu: Replace [cpu_physical_memory -> address_space]_rw() target/i386/whpx: Replace legacy cpu_physical_memory_rw() call target/i386/kvm: Replace legacy cpu_physical_memory_rw() call target/i386/nvmm: Inline cpu_physical_memory_rw() in nvmm_mem_callback hw/xen/hvm: Inline cpu_physical_memory_rw() in rw_phys_req_item() system/physmem: Un-inline cpu_physical_memory_read/write() system/physmem: Avoid cpu_physical_memory_rw when is_write is constant system/physmem: Remove legacy cpu_physical_memory_rw() hw/virtio/vhost: Replace legacy cpu_physical_memory_*map() calls hw/virtio/virtio: Replace legacy cpu_physical_memory_map() call
docs/devel/loads-stores.rst | 6 +-- scripts/coccinelle/exec_rw_const.cocci | 22 ----------- include/exec/cpu-common.h | 18 +-------- include/system/memory.h | 16 +++++++- hw/core/loader.c | 2 +- hw/s390x/sclp.c | 14 ++++--- hw/virtio/vhost.c | 7 +++- hw/virtio/virtio.c | 10 +++-- hw/xen/xen-hvm-common.c | 8 ++-- system/physmem.c | 51 ++++++++++++++------------ target/i386/arch_memory_mapping.c | 10 ++--- target/i386/kvm/xen-emu.c | 4 +- target/i386/nvmm/nvmm-all.c | 5 ++- target/i386/whpx/whpx-all.c | 7 +++- target/s390x/mmu_helper.c | 6 ++- 15 files changed, 92 insertions(+), 94 deletions(-)
Signature · 1 line
2.51.0
PATCH v3 1/18 docs/devel/loads-stores: Stop mentioning cpu_physical_memory_write_rom()
Update the documentation after commit 3c8133f9737 ("Rename cpu_physical_memory_write_rom() to address_space_write_rom()").
--- docs/devel/loads-stores.rst | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
docs/devel/loads-stores.rst +1 -1
@@ -474,7 +474,7 @@ This function is intended for use by the GDB stub and similar code. It takes a virtual address, converts it to a physical address via an MMU lookup using the current settings of the specified CPU, and then performs the access (using ``address_space_rw`` for -reads or ``cpu_physical_memory_write_rom`` for writes). +reads or ``address_space_write_rom`` for writes). This means that if the access is a write to a ROM then this function will modify the contents (whereas a normal guest CPU access would ignore the write attempt).
Signature · 1 line
2.51.0
PATCH v3 2/18 system/memory: Better describe @plen argument of flatview_translate()
In reply to Philippe Mathieu-Daudéflatview_translate()'s @plen argument is output-only and can be NULL.
When Xen is enabled, only update @plen_out when non-NULL.
--- include/system/memory.h | 5 +++-- system/physmem.c | 9 +++++---- 2 files changed, 8 insertions(+), 6 deletions(-)
include/system/memory.h +3 -2
@@ -2992,13 +2992,14 @@ IOMMUTLBEntry address_space_get_iotlb_entry(AddressSpace *as, hwaddr addr,* @addr: address within that address space * @xlat: pointer to address within the returned memory region section's * #MemoryRegion. -* @len: pointer to length +* @plen_out: pointer to valid read/write length of the translated address. +* It can be @NULL when we don't care about it. * @is_write: indicates the transfer direction * @attrs: memory attributes */ MemoryRegion *flatview_translate (FlatView *fv,hwaddr addr,hwaddr *xlat, -hwaddr *len,bool is_write, +hwaddr *plen_out,bool is_write,MemTxAttrs attrs); static inline MemoryRegion *address_space_translate(AddressSpace *as,
system/physmem.c +5 -4
@@ -566,7 +566,7 @@ iotlb_fail:/* Called from RCU critical section */ MemoryRegion *flatview_translate (FlatView *fv,hwaddr addr,hwaddr *xlat, -hwaddr *plen,bool is_write, +hwaddr *plen_out,bool is_write,MemTxAttrs attrs) {MemoryRegion *mr;
@@ -574,13 +574,14 @@ MemoryRegion *flatview_translate(FlatView *fv, hwaddr addr, hwaddr *xlat,AddressSpace *as =NULL ;/* This can be MMIO, so setup MMIO bit. */ - section =flatview_do_translate (fv, addr, xlat, plen,NULL , + section =flatview_do_translate (fv, addr, xlat, plen_out,NULL , is_write, true, &as, attrs); mr = section.mr; -if (xen_enabled () &&memory_access_is_direct (mr, is_write, attrs)) { +if (xen_enabled () && plen_out &&memory_access_is_direct (mr, is_write, + attrs)) {hwaddr page = ((addr &TARGET_PAGE_MASK ) +TARGET_PAGE_SIZE ) - addr; - *plen =MIN (page, *plen); + *plen_out =MIN (page, *plen_out); }return mr;
Signature · 1 line
2.51.0
Re: [PATCH v3 02/18] system/memory: Better describe @plen argument of flatview_translate()
Quoted 50 lines · Philippe Mathieu-Daudé
flatview_translate()'s @plen argument is output-only and can be NULL.
When Xen is enabled, only update @plen_out when non-NULL.
Signed-off-by Philippe Mathieu-Daudé <[email protected]>--- include/system/memory.h | 5 +++-- system/physmem.c | 9 +++++---- 2 files changed, 8 insertions(+), 6 deletions(-)include/system/memory.h +3
-2@@ -2992,13 +2992,14 @@ IOMMUTLBEntry address_space_get_iotlb_entry(AddressSpace *as, hwaddr addr,* @addr: address within that address space * @xlat: pointer to address within the returned memory region section's * #MemoryRegion. -* @len: pointer to length +* @plen_out: pointer to valid read/write length of the translated address. +* It can be @NULL when we don't care about it. * @is_write: indicates the transfer direction * @attrs: memory attributes */ MemoryRegion *flatview_translate (FlatView *fv,hwaddr addr,hwaddr *xlat, -hwaddr *len,bool is_write, +hwaddr *plen_out,bool is_write,MemTxAttrs attrs); static inline MemoryRegion *address_space_translate(AddressSpace *as,system/physmem.c +2
-2@@ -566,7 +566,7 @@ iotlb_fail:/* Called from RCU critical section */ MemoryRegion *flatview_translate (FlatView *fv,hwaddr addr,hwaddr *xlat, -hwaddr *plen,bool is_write, +hwaddr *plen_out,bool is_write,MemTxAttrs attrs) {MemoryRegion *mr;@@ -574,13 +574,14 @@ MemoryRegion *flatview_translate(FlatView *fv, hwaddr addr, hwaddr *xlat,AddressSpace *as =NULL ;/* This can be MMIO, so setup MMIO bit. */ - section = flatview_do_translate(fv, addr, xlat, plen, NULL, + section = flatview_do_translate(fv, addr, xlat, plen_out, NULL,
system/physmem.c MemoryRegion *flatview_translate(FlatView *fv, hwaddr addr, hwaddr *xlat, · line 585
Quoted 10 lines
system/physmem.c +3
-2is_write, true, &as, attrs); mr = section.mr; -if (xen_enabled () &&memory_access_is_direct (mr, is_write, attrs)) { +if (xen_enabled () && plen_out &&memory_access_is_direct (mr, is_write, + attrs)) {hwaddr page = ((addr &TARGET_PAGE_MASK ) +TARGET_PAGE_SIZE ) - addr; - *plen =MIN (page, *plen); + *plen_out =MIN (page, *plen_out); }
My question from the previous version is still unanswered:
https://lore.kernel.org/qemu-devel/[email protected]/
Thomas
Re: [PATCH v3 02/18] system/memory: Better describe @plen argument of flatview_translate()
Hi Thomas,
Quoted 65 lines · Thomas Huth
Quoted 51 lines · Philippe Mathieu-Daudé
flatview_translate()'s @plen argument is output-only and can be NULL.
When Xen is enabled, only update @plen_out when non-NULL.
Signed-off-by Philippe Mathieu-Daudé <[email protected]>---
include/system/memory.h| 5 +++--
system/physmem.c | 9 +++++----2 files changed, 8 insertions(+), 6 deletions(-)include/system/memory.h +1
-1@@ -2992,13 +2992,14 @@ IOMMUTLBEntry address_space_get_iotlb_entry(AddressSpace *as, hwaddr addr, * @addr: address within that address space * @xlat: pointer to address within the returned memory region section's * #MemoryRegion. - * @len: pointer to length + * @plen_out: pointer to valid read/write length of the translatedaddress.
+ * It can be @NULL when we don't care about it.
* @is_write: indicates the transfer direction
* @attrs: memory attributes
*/
MemoryRegion *flatview_translate(FlatView *fv,
hwaddr addr, hwaddr *xlat, - hwaddr *len, boolis_write, + hwaddr*plen_out, boolis_write, MemTxAttrs attrs);
static inline MemoryRegion *address_space_translate(AddressSpace *as,system/physmem.c +2
-2@@ -566,7 +566,7 @@ iotlb_fail: /* Called from RCU critical section */ MemoryRegion *flatview_translate(FlatView *fv, hwaddr addr, hwaddr *xlat, - hwaddr *plen, bool is_write, + hwaddr *plen_out, bool is_write, MemTxAttrs attrs) { MemoryRegion *mr;@@ -574,13 +574,14 @@ MemoryRegion *flatview_translate(FlatView *fv, hwaddr addr, hwaddr *xlat, AddressSpace *as = NULL; /* This can be MMIO, so setup MMIO bit. */ - section = flatview_do_translate(fv, addr, xlat, plen, NULL, + section = flatview_do_translate(fv, addr, xlat, plen_out, NULL, is_write, true, &as, attrs);system/physmem.c MemoryRegion *flatview_translate(FlatView *fv, · line 585
Quoted 10 lines
system/physmem.c +3
-3mr = section.mr; - if (xen_enabled() && memory_access_is_direct(mr, is_write, attrs)) { + if (xen_enabled() && plen_out && memory_access_is_direct(mr, is_write, + attrs)) { hwaddr page = ((addr & TARGET_PAGE_MASK) + TARGET_PAGE_SIZE) - addr; - *plen = MIN(page, *plen); + *plen_out = MIN(page, *plen_out); }My question from the previous version is still unanswered:
Quoted 1 line
This patches
- checks for plen not being NULL
- describes it as
"When Xen is enabled, only update @plen_out when non-NULL."
- mention that in the updated flatview_translate() documentation "It can be @NULL when we don't care about it." as documented for the flatview_do_translate() callee in commit d5e5fafd11b ("exec:
add page_mask for flatview_do_translate")
before:
it was not clear whether we can pass plen=NULL without having to look at the code.
after:
no change when plen is not NULL, we can pass plen=NULL safely (it is documented).
I shouldn't be understanding your original question, do you mind rewording it?
Thanks,
Phil.
Thomas Huth
Reviewed-byRe: [PATCH v3 02/18] system/memory: Better describe @plen argument of flatview_translate()
Quoted 81 lines · Philippe Mathieu-Daudé
Hi Thomas,
Quoted 62 lines · Thomas Huth
Quoted 48 lines · Philippe Mathieu-Daudé
flatview_translate()'s @plen argument is output-only and can be NULL.
When Xen is enabled, only update @plen_out when non-NULL.
Signed-off-by Philippe Mathieu-Daudé <[email protected]>---
include/system/memory.h| 5 +++--
system/physmem.c | 9 +++++----2 files changed, 8 insertions(+), 6 deletions(-)include/system/memory.h +3
-2@@ -2992,13 +2992,14 @@ IOMMUTLBEntryaddress_space_get_iotlb_entry(AddressSpace *as, hwaddr addr, * @addr: address within that address space * @xlat: pointer to address within the returned memory region section's * #MemoryRegion. -* @len: pointer to length +* @plen_out: pointer to valid read/write length of the translated address. +* It can be @NULL when we don't care about it. * @is_write: indicates the transfer direction * @attrs: memory attributes */ MemoryRegion *flatview_translate(FlatView *fv, hwaddr addr, hwaddr *xlat, - hwaddr *len, bool is_write, + hwaddr *plen_out, bool is_write, MemTxAttrs attrs); static inline MemoryRegion *address_space_translate(AddressSpace *as,system/physmem.c +1
-2@@ -566,7 +566,7 @@ iotlb_fail: /* Called from RCU critical section */ MemoryRegion *flatview_translate(FlatView *fv, hwaddr addr, hwaddr *xlat, - hwaddr *plen, bool is_write, + hwaddr *plen_out, bool is_write, MemTxAttrs attrs) { MemoryRegion *mr;@@ -574,13 +574,14 @@ MemoryRegion *flatview_translate(FlatView *fv, hwaddr addr, hwaddr *xlat, AddressSpace *as = NULL; /* This can be MMIO, so setup MMIO bit. */ - section = flatview_do_translate(fv, addr, xlat, plen, NULL,system/physmem.c MemoryRegion *flatview_translate(FlatView *fv, · line 584
Quoted 10 lines
system/physmem.c +4
-2+ section = flatview_do_translate(fv, addr, xlat, plen_out, NULL, is_write, true, &as, attrs); mr = section.mr; - if (xen_enabled() && memory_access_is_direct(mr, is_write, attrs)) { + if (xen_enabled() && plen_out && memory_access_is_direct(mr, is_write, + attrs)) { hwaddr page = ((addr & TARGET_PAGE_MASK) + TARGET_PAGE_SIZE) - addr; - *plen = MIN(page, *plen); + *plen_out = MIN(page, *plen_out); }My question from the previous version is still unanswered:
Quoted 1 line
This patches
- checks for plen not being NULL
- describes it as
"When Xen is enabled, only update @plen_out when non-NULL."
- mention that in the updatedflatview_translate()documentation "It can be @NULL when we don't care about it." as documented for theflatview_do_translate()callee in commitd5e5fafd11b("exec:
addpage_maskforflatview_do_translate")before:
it was not clear whether we can pass plen=NULL without having to look at the code.
Quoted 7 lines
after:
no change when plen is not NULL, we can pass plen=NULL safely (it is documented).
I shouldn't be understanding your original question, do you mind rewording it?
Ah, you've updated the patch in v3 to include a check for plen_out in the if-statement! It was not there in v2. Ok, this should be fine now:
I just re-complained since you did not respond to my mail in v2, and when I looked at the changelog in your v3 cover letter, you did not mention the modification here, so I blindly assumed that this patch was unchanged.
Thomas
Re: [PATCH v3 02/18] system/memory: Better describe @plen argument of flatview_translate()
Quoted 90 lines · Thomas Huth
Quoted 89 lines · Philippe Mathieu-Daudé
Hi Thomas,
Quoted 65 lines · Thomas Huth
Quoted 51 lines · Philippe Mathieu-Daudé
flatview_translate()'s @plen argument is output-only and can be NULL.
When Xen is enabled, only update @plen_out when non-NULL.
Signed-off-by Philippe Mathieu-Daudé <[email protected]>---
include/system/memory.h| 5 +++--
system/physmem.c | 9 +++++----2 files changed, 8 insertions(+), 6 deletions(-)include/system/memory.h +1
-1@@ -2992,13 +2992,14 @@ IOMMUTLBEntry address_space_get_iotlb_entry(AddressSpace *as, hwaddr addr, * @addr: address within that address space * @xlat: pointer to address within the returned memory region section's * #MemoryRegion. - * @len: pointer to length + * @plen_out: pointer to valid read/write length of the translatedaddress.
+ * It can be @NULL when we don't care about it.
* @is_write: indicates the transfer direction
* @attrs: memory attributes
*/
MemoryRegion *flatview_translate(FlatView *fv,
hwaddr addr, hwaddr *xlat, - hwaddr *len, boolis_write, + hwaddr*plen_out, boolis_write, MemTxAttrs attrs);
static inline MemoryRegion *address_space_translate(AddressSpace *as,system/physmem.c +2
-2@@ -566,7 +566,7 @@ iotlb_fail: /* Called from RCU critical section */ MemoryRegion *flatview_translate(FlatView *fv, hwaddr addr, hwaddr *xlat, - hwaddr *plen, bool is_write, + hwaddr *plen_out, bool is_write, MemTxAttrs attrs) { MemoryRegion *mr;@@ -574,13 +574,14 @@ MemoryRegion *flatview_translate(FlatView *fv, hwaddr addr, hwaddr *xlat, AddressSpace *as = NULL; /* This can be MMIO, so setup MMIO bit. */ - section = flatview_do_translate(fv, addr, xlat, plen, NULL, + section = flatview_do_translate(fv, addr, xlat, plen_out, NULL, is_write, true, &as, attrs);system/physmem.c MemoryRegion *flatview_translate(FlatView *fv, · line 586
Quoted 10 lines
system/physmem.c +3
-2mr = section.mr; - if (xen_enabled() && memory_access_is_direct(mr, is_write, attrs)) { + if (xen_enabled() && plen_out && memory_access_is_direct(mr, is_write, + attrs)) { hwaddr page = ((addr & TARGET_PAGE_MASK) + TARGET_PAGE_SIZE) - addr; - *plen = MIN(page, *plen); + *plen_out = MIN(page, *plen_out); }My question from the previous version is still unanswered:
Quoted 1 line
This patches
- checks for plen not being NULL
- describes it as
"When Xen is enabled, only update @plen_out when non-NULL."
- mention that in the updatedflatview_translate()documentation "It can be @NULL when we don't care about it." as documented for theflatview_do_translate()callee in commitd5e5fafd11b("exec:
addpage_maskforflatview_do_translate")before:
it was not clear whether we can pass plen=NULL without having to look at the code.
after:
no change when plen is not NULL, we can pass plen=NULL safely (it is documented).
Quoted 8 lines
Quoted 2 lines
I shouldn't be understanding your original question, do you mind rewording it?
Ah, you've updated the patch in v3 to include a check for
plen_outin the if-statement! It was not there in v2. Ok, this should be fine now:Reviewed-by Thomas Huth <[email protected]>I just re-complained since you did not respond to my mail in v2, and when I looked at the changelog in your v3 cover letter, you did not mention the modification here, so I blindly assumed that this patch was unchanged.
Ah I see... OK I'll try to be more explicit in my respins.
Thanks for your review!
Phil.
Re: [PATCH v3 02/18] system/memory: Better describe @plen argument of flatview_translate()
In reply to Philippe Mathieu-DaudéQuoted 19 lines · Philippe Mathieu-Daudé
flatview_translate()'s @plen argument is output-only and can be NULL.
When Xen is enabled, only update @plen_out when non-NULL.
Signed-off-by Philippe Mathieu-Daudé <[email protected]>--- include/system/memory.h | 5 +++-- system/physmem.c | 9 +++++---- 2 files changed, 8 insertions(+), 6 deletions(-)include/system/memory.h +0
-1@@ -2992,13 +2992,14 @@ IOMMUTLBEntry address_space_get_iotlb_entry(AddressSpace *as, hwaddr addr,* @addr: address within that address space * @xlat: pointer to address within the returned memory region section's * #MemoryRegion. -* @len: pointer to length
include/system/memory.h IOMMUTLBEntry address_space_get_iotlb_entry(AddressSpace *as, hwaddr addr, · line 3003
Quoted 10 lines
include/system/memory.h +3
-1+* @plen_out: pointer to valid read/write length of the translated address. +* It can be @NULL when we don't care about it. * @is_write: indicates the transfer direction * @attrs: memory attributes */ MemoryRegion *flatview_translate (FlatView *fv,hwaddr addr,hwaddr *xlat, -hwaddr *len,bool is_write, +hwaddr *plen_out,bool is_write,MemTxAttrs attrs);
Renaming to plen_out is misleading, because it is used as an input as well:
include/system/memory.h IOMMUTLBEntry address_space_get_iotlb_entry(AddressSpace *as, hwaddr addr, · line 3004
Quoted 2 lines
include/system/memory.h +1
-1- *plen =MIN (page, *plen); + *plen_out =MIN (page, *plen_out);
right here. If non-null, the input should be the access length of the caller.
r~
PATCH v3 3/18 system/memory: Factor address_space_is_io() out
In reply to Philippe Mathieu-DaudéFactor address_space_is_io() out of cpu_physical_memory_is_io().
--- include/system/memory.h | 9 +++++++++ system/physmem.c | 21 ++++++++++++--------- 2 files changed, 21 insertions(+), 9 deletions(-)
include/system/memory.h +9 -0
@@ -3030,6 +3030,15 @@ static inline MemoryRegion *address_space_translate(AddressSpace *as,bool address_space_access_valid (AddressSpace *as,hwaddr addr,hwaddr len,bool is_write,MemTxAttrs attrs); +/** +* address_space_is_io: check whether an guest physical addresses +* whithin an address space is I/O memory. +* +* @as: #AddressSpace to be accessed +* @addr: address within that address space +*/ +bool address_space_is_io (AddressSpace *as,hwaddr addr); +/* address_space_map: map a physical memory region into a host virtual address * * May map a subset of the requested range, given by and returned in @plen.
system/physmem.c +12 -9
@@ -3359,6 +3359,17 @@ bool address_space_access_valid(AddressSpace *as, hwaddr addr,return flatview_access_valid (fv, addr, len, is_write, attrs); } +bool address_space_is_io (AddressSpace *as,hwaddr addr) +{ +MemoryRegion *mr; + +RCU_READ_LOCK_GUARD (); + mr =address_space_translate (as, addr, &addr,NULL , false, +MEMTXATTRS_UNSPECIFIED ); + +return !(memory_region_is_ram (mr) ||memory_region_is_romd (mr)); +} + static hwaddr flatview_extend_translation(FlatView *fv, hwaddr addr, hwaddr target_len,
@@ -3755,15 +3766,7 @@ int cpu_memory_rw_debug(CPUState *cpu, vaddr addr,bool cpu_physical_memory_is_io (hwaddr phys_addr) { -MemoryRegion *mr; -hwaddr l =1 ; - -RCU_READ_LOCK_GUARD (); - mr =address_space_translate (&address_space_memory, - phys_addr, &phys_addr, &l, false, -MEMTXATTRS_UNSPECIFIED ); - -return !(memory_region_is_ram (mr) ||memory_region_is_romd (mr)); +return address_space_is_io (&address_space_memory, phys_addr); }int qemu_ram_foreach_block (RAMBlockIterFunc func,void *opaque)
Signature · 1 line
2.51.0
Re: [PATCH v3 03/18] system/memory: Factor address_space_is_io() out
Quoted 8 lines · Philippe Mathieu-Daudé
Factor
address_space_is_io()out ofcpu_physical_memory_is_io().Signed-off-by Philippe Mathieu-Daudé <[email protected]>Reviewed-by Thomas Huth <[email protected]>--- include/system/memory.h | 9 +++++++++ system/physmem.c | 21 ++++++++++++--------- 2 files changed, 21 insertions(+), 9 deletions(-)
r~
PATCH v3 4/18 target/i386/arch_memory_mapping: Use address_space_memory_is_io()
In reply to Philippe Mathieu-DaudéSince all functions have an address space argument, it is trivial to replace cpu_physical_memory_is_io() by address_space_memory_is_io().
--- target/i386/arch_memory_mapping.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
target/i386/arch_memory_mapping.c +5 -5
@@ -35,7 +35,7 @@ static void walk_pte(MemoryMappingList *list, AddressSpace *as, } start_paddr = (pte & ~0xfff ) & ~(0x1ULL <<63 ); -if (cpu_physical_memory_is_io (start_paddr)) { +if (address_space_is_io (as, start_paddr)) {/* I/O region */ continue ; }
@@ -65,7 +65,7 @@ static void walk_pte2(MemoryMappingList *list, AddressSpace *as, } start_paddr = pte & ~0xfff ; -if (cpu_physical_memory_is_io (start_paddr)) { +if (address_space_is_io (as, start_paddr)) {/* I/O region */ continue ; }
@@ -100,7 +100,7 @@ static void walk_pde(MemoryMappingList *list, AddressSpace *as,if (pde &PG_PSE_MASK ) {/* 2 MB page */ start_paddr = (pde & ~0x1fffff ) & ~(0x1ULL <<63 ); -if (cpu_physical_memory_is_io (start_paddr)) { +if (address_space_is_io (as, start_paddr)) {/* I/O region */ continue ; }
@@ -142,7 +142,7 @@ static void walk_pde2(MemoryMappingList *list, AddressSpace *as,*/ high_paddr = ((hwaddr)(pde &0x1fe000 ) <<19 ); start_paddr = (pde & ~0x3fffff ) | high_paddr; -if (cpu_physical_memory_is_io (start_paddr)) { +if (address_space_is_io (as, start_paddr)) {/* I/O region */ continue ; }
@@ -203,7 +203,7 @@ static void walk_pdpe(MemoryMappingList *list, AddressSpace *as,if (pdpe &PG_PSE_MASK ) {/* 1 GB page */ start_paddr = (pdpe & ~0x3fffffff ) & ~(0x1ULL <<63 ); -if (cpu_physical_memory_is_io (start_paddr)) { +if (address_space_is_io (as, start_paddr)) {/* I/O region */ continue ; }
Signature · 1 line
2.51.0
Re: [PATCH v3 04/18] target/i386/arch_memory_mapping: Use address_space_memory_is_io()
Quoted 8 lines · Philippe Mathieu-Daudé
Since all functions have an address space argument, it is trivial to replace
cpu_physical_memory_is_io()byaddress_space_memory_is_io().Signed-off-by Philippe Mathieu-Daudé <[email protected]>--- target/i386/arch_memory_mapping.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
r~
PATCH v3 5/18 hw/s390x/sclp: Use address_space_memory_is_io() in sclp_service_call()
In reply to Philippe Mathieu-DaudéWhen cpu_address_space_init() isn't called during vCPU creation, its single address space is the global &address_space_memory.
As s390x boards don't call cpu_address_space_init(), cpu_get_address_space(CPU(cpu), 0) returns &address_space_memory.
We can then replace cpu_physical_memory_is_io() by the semantically equivalent address_space_memory_is_io() call.
--- hw/s390x/sclp.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
hw/s390x/sclp.c +3 -1
@@ -16,6 +16,7 @@#include "qemu/units.h" #include "qapi/error.h" #include "hw/boards.h" +#include "system/memory.h" #include "hw/s390x/sclp.h" #include "hw/s390x/event-facility.h" #include "hw/s390x/s390-pci-bus.h"
@@ -301,6 +302,7 @@ int sclp_service_call(S390CPU *cpu, uint64_t sccb, uint32_t code)CPUS390XState *env = &cpu->env;SCLPDevice *sclp =get_sclp_device ();SCLPDeviceClass *sclp_c =SCLP_GET_CLASS (sclp); +AddressSpace *as =cpu_get_address_space (CPU (cpu),0 );SCCBHeader header; g_autofree SCCB *work_sccb = NULL;
@@ -308,7 +310,7 @@ int sclp_service_call(S390CPU *cpu, uint64_t sccb, uint32_t code)if (env->psw.mask &PSW_MASK_PSTATE ) {return -PGM_PRIVILEGED ; } -if (cpu_physical_memory_is_io (sccb)) { +if (address_space_is_io (as, sccb)) {return -PGM_ADDRESSING ; } if ((sccb & ~0x1fffUL) == 0 || (sccb & ~0x1fffUL) == env->psa
Signature · 1 line
2.51.0
Re: [PATCH v3 05/18] hw/s390x/sclp: Use address_space_memory_is_io() in sclp_service_call()
Quoted 6 lines · Philippe Mathieu-Daudé
When
cpu_address_space_init()isn't called during vCPU creation, its single address space is the global&address_space_memory.As s390x boards don't call
cpu_address_space_init(),cpu_get_address_space(CPU(cpu), 0)returns&address_space_memory.
Quoted 8 lines
We can then replace
cpu_physical_memory_is_io()by the semantically equivalentaddress_space_memory_is_io()call.Signed-off-by Philippe Mathieu-Daudé <[email protected]>Reviewed-by Thomas Huth <[email protected]>--- hw/s390x/sclp.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
r~
PATCH v3 6/18 system/physmem: Remove cpu_physical_memory_is_io()
In reply to Philippe Mathieu-DaudéThere are no more uses of the legacy cpu_physical_memory_is_io() method. Remove it.
--- include/exec/cpu-common.h | 2 -- system/physmem.c | 5 ----- 2 files changed, 7 deletions(-)
include/exec/cpu-common.h +0 -2
@@ -149,8 +149,6 @@ void *cpu_physical_memory_map(hwaddr addr,void cpu_physical_memory_unmap (void *buffer,hwaddr len,bool is_write,hwaddr access_len); -bool cpu_physical_memory_is_io (hwaddr phys_addr); -/* Coalesced MMIO regions are areas where write operations can be reordered. * This usually implies that write operations are side-effect free. This allows * batching which can make a major impact on performance when using
system/physmem.c +0 -5
@@ -3764,11 +3764,6 @@ int cpu_memory_rw_debug(CPUState *cpu, vaddr addr,return 0 ; } -bool cpu_physical_memory_is_io (hwaddr phys_addr) -{ -return address_space_is_io (&address_space_memory, phys_addr); -} -int qemu_ram_foreach_block (RAMBlockIterFunc func,void *opaque) {RAMBlock *block;
Signature · 1 line
2.51.0
PATCH v3 7/18 system/physmem: Pass address space argument to cpu_flush_icache_range()
In reply to Philippe Mathieu-DaudéRename cpu_flush_icache_range() as address_space_flush_icache_range(), passing an address space by argument. The single caller, rom_reset(), already operates on an address space. Use it.
--- include/exec/cpu-common.h | 2 -- include/system/memory.h | 2 ++ hw/core/loader.c | 2 +- system/physmem.c | 5 ++--- 4 files changed, 5 insertions(+), 6 deletions(-)
include/exec/cpu-common.h +0 -2
@@ -156,8 +156,6 @@ void cpu_physical_memory_unmap(void *buffer, hwaddr len,*/ void qemu_flush_coalesced_mmio_buffer (void ); -void cpu_flush_icache_range (hwaddr start,hwaddr len); -typedef int (RAMBlockIterFunc )(RAMBlock *rb,void *opaque);int qemu_ram_foreach_block (RAMBlockIterFunc func,void *opaque);
include/system/memory.h +2 -0
@@ -2977,6 +2977,8 @@ void address_space_cache_invalidate(MemoryRegionCache *cache,*/ void address_space_cache_destroy (MemoryRegionCache *cache); +void address_space_flush_icache_range (AddressSpace *as,hwaddr addr,hwaddr len); +/* address_space_get_iotlb_entry: translate an address into an IOTLB * entry. Should be called from an RCU critical section. */
hw/core/loader.c +1 -1
@@ -1242,7 +1242,7 @@ static void rom_reset(void *unused)* that the instruction cache for that new region is clear, so that the * CPU definitely fetches its instructions from the just written data. */ -cpu_flush_icache_range (rom->addr, rom->datasize); +address_space_flush_icache_range (rom->as, rom->addr, rom->datasize);trace_loader_write_rom (rom->name, rom->addr, rom->datasize, rom->isrom); }
system/physmem.c +2 -3
@@ -3215,7 +3215,7 @@ MemTxResult address_space_write_rom(AddressSpace *as, hwaddr addr,return MEMTX_OK ; } -void cpu_flush_icache_range (hwaddr addr,hwaddr len) +void address_space_flush_icache_range (AddressSpace *as,hwaddr addr,hwaddr len) {/* * This function should do the same thing as an icache flush that was
@@ -3230,8 +3230,7 @@ void cpu_flush_icache_range(hwaddr addr, hwaddr len)RCU_READ_LOCK_GUARD ();while (len >0 ) {hwaddr addr1, l = len; -MemoryRegion *mr =address_space_translate (&address_space_memory, - addr, &addr1, &l, true, +MemoryRegion *mr =address_space_translate (as, addr, &addr1, &l, true,MEMTXATTRS_UNSPECIFIED );if (!memory_region_supports_direct_access (mr)) {
Signature · 1 line
2.51.0
PATCH v3 8/18 hw/s390x/sclp: Replace [cpu_physical_memory -> address_space]_r/w()
In reply to Philippe Mathieu-Daudécpu_physical_memory_read() and cpu_physical_memory_write() are legacy (see commit b7ecba0f6f6), replace by address_space_read() and address_space_write().
--- hw/s390x/sclp.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-)
hw/s390x/sclp.c +6 -4
@@ -319,7 +319,8 @@ int sclp_service_call(S390CPU *cpu, uint64_t sccb, uint32_t code) }/* the header contains the actual length of the sccb */ -cpu_physical_memory_read (sccb, &header,sizeof (SCCBHeader)); +address_space_read (as, sccb,MEMTXATTRS_UNSPECIFIED , + &header,sizeof (SCCBHeader));/* Valid sccb sizes */ if (be16_to_cpu (header.length) <sizeof (SCCBHeader)) {
@@ -332,7 +333,8 @@ int sclp_service_call(S390CPU *cpu, uint64_t sccb, uint32_t code)* the host has checked the values */ work_sccb =g_malloc0 (be16_to_cpu (header.length)); -cpu_physical_memory_read (sccb, work_sccb,be16_to_cpu (header.length)); +address_space_read (as, sccb,MEMTXATTRS_UNSPECIFIED , + work_sccb,be16_to_cpu (header.length));if (!sclp_command_code_valid (code)) { work_sccb->h.response_code =cpu_to_be16 (SCLP_RC_INVALID_SCLP_COMMAND );
@@ -346,8 +348,8 @@ int sclp_service_call(S390CPU *cpu, uint64_t sccb, uint32_t code) sclp_c->execute (sclp, work_sccb, code); out_write: -cpu_physical_memory_write (sccb, work_sccb, -be16_to_cpu (work_sccb->h.length)); +address_space_write (as, sccb,MEMTXATTRS_UNSPECIFIED , + work_sccb,be16_to_cpu (header.length)); sclp_c->service_interrupt (sclp, sccb);
Signature · 1 line
2.51.0
Re: [PATCH v3 08/18] hw/s390x/sclp: Replace [cpu_physical_memory -> address_space]_r/w()
Quoted 9 lines · Philippe Mathieu-Daudé
cpu_physical_memory_read()andcpu_physical_memory_write()are legacy (see commitb7ecba0f6f6), replace byaddress_space_read()andaddress_space_write().Signed-off-by Philippe Mathieu-Daudé <[email protected]>Reviewed-by Thomas Huth <[email protected]>--- hw/s390x/sclp.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-)
r~
PATCH v3 9/18 target/s390x/mmu: Replace [cpu_physical_memory -> address_space]_rw()
In reply to Philippe Mathieu-DaudéWhen cpu_address_space_init() isn't called during vCPU creation, its single address space is the global &address_space_memory.
As s390x boards don't call cpu_address_space_init(), cpu_get_address_space(CPU(cpu), 0) returns &address_space_memory.
We can then replace cpu_physical_memory_rw() by the semantically equivalent address_space_rw() call.
--- target/s390x/mmu_helper.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
target/s390x/mmu_helper.c +4 -2
@@ -23,6 +23,7 @@#include "kvm/kvm_s390x.h" #include "system/kvm.h" #include "system/tcg.h" +#include "system/memory.h" #include "exec/page-protection.h" #include "exec/target_page.h" #include "hw/hw.h"
@@ -522,6 +523,7 @@ int s390_cpu_pv_mem_rw(S390CPU *cpu, unsigned int offset, void *hostbuf,int s390_cpu_virt_mem_rw (S390CPU *cpu,vaddr laddr,uint8_t ar,void *hostbuf,int len,bool is_write) { +AddressSpace *as =cpu_get_address_space (CPU (cpu),0 );int currlen, nr_pages, i;target_ulong *pages;uint64_t tec;
@@ -545,8 +547,8 @@ int s390_cpu_virt_mem_rw(S390CPU *cpu, vaddr laddr, uint8_t ar, void *hostbuf,/* Copy data by stepping through the area page by page */ for (i =0 ; i < nr_pages; i++) { currlen =MIN (len,TARGET_PAGE_SIZE - (laddr %TARGET_PAGE_SIZE )); -cpu_physical_memory_rw (pages[i] | (laddr & ~TARGET_PAGE_MASK ), - hostbuf, currlen, is_write); +address_space_rw (as, pages[i] | (laddr & ~TARGET_PAGE_MASK ), +MEMTXATTRS_UNSPECIFIED , hostbuf, currlen, is_write); laddr += currlen; hostbuf += currlen; len -= currlen;
Signature · 1 line
2.51.0
Re: [PATCH v3 09/18] target/s390x/mmu: Replace [cpu_physical_memory -> address_space]_rw()
Quoted 6 lines · Philippe Mathieu-Daudé
When
cpu_address_space_init()isn't called during vCPU creation, its single address space is the global&address_space_memory.As s390x boards don't call
cpu_address_space_init(),cpu_get_address_space(CPU(cpu), 0)returns&address_space_memory.
Quoted 8 lines
We can then replace
cpu_physical_memory_rw()by the semantically equivalentaddress_space_rw()call.Signed-off-by Philippe Mathieu-Daudé<[email protected]>Reviewed-by Thomas Huth<[email protected]>--- target/s390x/mmu_helper.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
r~
PATCH v3 10/18 target/i386/whpx: Replace legacy cpu_physical_memory_rw() call
In reply to Philippe Mathieu-DaudéGet the vCPU address space and convert the legacycpu_physical_memory_rw() by address_space_rw().
--- target/i386/whpx/whpx-all.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
target/i386/whpx/whpx-all.c +5 -2
@@ -788,8 +788,11 @@ static HRESULT CALLBACK whpx_emu_mmio_callback(void *ctx, WHV_EMULATOR_MEMORY_ACCESS_INFO *ma) { - cpu_physical_memory_rw(ma->GpaAddress, ma->Data, ma->AccessSize, - ma->Direction); + CPUState *cpu = (CPUState *)ctx; +AddressSpace *as =cpu_addressspace (cs,MEMTXATTRS_UNSPECIFIED ); + +address_space_rw (as, ma->GpaAddress,MEMTXATTRS_UNSPECIFIED , + ma->Data, ma->AccessSize, ma->Direction);return S_OK ; }
Signature · 1 line
2.51.0
Re: [PATCH v3 10/18] target/i386/whpx: Replace legacy cpu_physical_memory_rw() call
Quoted 7 lines · Philippe Mathieu-Daudé
Get the vCPU address space and convert the legacy
cpu_physical_memory_rw()byaddress_space_rw().Signed-off-by Philippe Mathieu-Daudé<[email protected]>--- target/i386/whpx/whpx-all.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
r~
PATCH v3 11/18 target/i386/kvm: Replace legacy cpu_physical_memory_rw() call
In reply to Philippe Mathieu-DaudéGet the vCPU address space and convert the legacycpu_physical_memory_rw() by address_space_rw().
--- target/i386/kvm/xen-emu.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
target/i386/kvm/xen-emu.c +3 -1
@@ -21,6 +21,7 @@#include "system/address-spaces.h" #include "xen-emu.h" #include "trace.h" +#include "system/memory.h" #include "system/runstate.h" #include "hw/pci/msi.h"
@@ -75,6 +76,7 @@ static bool kvm_gva_to_gpa(CPUState *cs, uint64_t gva, uint64_t *gpa,static int kvm_gva_rw (CPUState *cs,uint64_t gva,void *_buf,size_t sz,bool is_write) { +AddressSpace *as =cpu_addressspace (cs,MEMTXATTRS_UNSPECIFIED );uint8_t *buf = (uint8_t *)_buf;uint64_t gpa;size_t len;
@@ -87,7 +89,7 @@ static int kvm_gva_rw(CPUState *cs, uint64_t gva, void *_buf, size_t sz, len = sz; } -cpu_physical_memory_rw (gpa, buf, len, is_write); +address_space_rw (as, gpa,MEMTXATTRS_UNSPECIFIED , buf, len, is_write); buf += len; sz -= len;
Signature · 1 line
2.51.0
Re: [PATCH v3 11/18] target/i386/kvm: Replace legacy cpu_physical_memory_rw() call
Quoted 7 lines · Philippe Mathieu-Daudé
Get the vCPU address space and convert the legacy
cpu_physical_memory_rw()byaddress_space_rw().Signed-off-by Philippe Mathieu-Daudé<[email protected]>--- target/i386/kvm/xen-emu.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
r~
PATCH v3 12/18 target/i386/nvmm: Inline cpu_physical_memory_rw() in nvmm_mem_callback
In reply to Philippe Mathieu-Daudé--- target/i386/nvmm/nvmm-all.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
target/i386/nvmm/nvmm-all.c +4 -1
@@ -15,6 +15,7 @@#include "accel/accel-ops.h" #include "system/nvmm.h" #include "system/cpus.h" +#include "system/memory.h" #include "system/runstate.h" #include "qemu/main-loop.h" #include "qemu/error-report.h"
@@ -516,7 +517,9 @@ nvmm_io_callback(struct nvmm_io *io)static void nvmm_mem_callback (struct nvmm_mem *mem) { -cpu_physical_memory_rw (mem->gpa, mem->data, mem->size, mem->write); +/* TODO: Get CPUState via mem->vcpu? */ +address_space_rw (&address_space_memory, mem->gpa,MEMTXATTRS_UNSPECIFIED , + mem->data, mem->size, mem->write);/* Needed, otherwise infinite loop. */ current_cpu->vcpu_dirty = false;
Signature · 1 line
2.51.0
Re: [PATCH v3 12/18] target/i386/nvmm: Inline cpu_physical_memory_rw() in nvmm_mem_callback
Quoted 4 lines · Philippe Mathieu-Daudé
Signed-off-by Philippe Mathieu-Daudé <[email protected]>--- target/i386/nvmm/nvmm-all.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
r~
PATCH v3 13/18 hw/xen/hvm: Inline cpu_physical_memory_rw() in rw_phys_req_item()
In reply to Philippe Mathieu-Daudécpu_physical_memory_rw() is legacy, replace by address_space_rw().
--- hw/xen/xen-hvm-common.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-)
hw/xen/xen-hvm-common.c +5 -3
@@ -12,6 +12,7 @@#include "hw/xen/xen-bus.h" #include "hw/boards.h" #include "hw/xen/arch_hvm.h" +#include "system/memory.h" #include "system/runstate.h" #include "system/system.h" #include "system/xen.h"
@@ -279,8 +280,8 @@ static void do_outp(uint32_t addr,* memory, as part of the implementation of an ioreq. * * Equivalent to -* cpu_physical_memory_rw(addr + (req->df ? -1 : +1) * req->size * i, -* val, req->size, 0/1) +* address_space_rw(as, addr + (req->df ? -1 : +1) * req->size * i, +* attrs, val, req->size, 0/1) * except without the integer overflow problems. */ static void rw_phys_req_item(hwaddr addr,
@@ -295,7 +296,8 @@ static void rw_phys_req_item(hwaddr addr, }else { addr += offset; } -cpu_physical_memory_rw (addr, val, req->size, rw); +address_space_rw (&address_space_memory, addr,MEMTXATTRS_UNSPECIFIED , + val, req->size, rw); } static inline void read_phys_req_item(hwaddr addr,
Signature · 1 line
2.51.0
Re: [PATCH v3 13/18] hw/xen/hvm: Inline cpu_physical_memory_rw() in rw_phys_req_item()
Quoted 6 lines · Philippe Mathieu-Daudé
cpu_physical_memory_rw()is legacy, replace byaddress_space_rw().Signed-off-by Philippe Mathieu-Daudé <[email protected]>--- hw/xen/xen-hvm-common.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-)
r~
PATCH v3 14/18 system/physmem: Un-inline cpu_physical_memory_read/write()
In reply to Philippe Mathieu-DaudéIn order to remove cpu_physical_memory_rw() in a pair of commits, and due to a cyclic dependency between "exec/cpu-common.h" and "system/memory.h", un-inline cpu_physical_memory_read() and cpu_physical_memory_write() as a prerequired step.
--- include/exec/cpu-common.h | 12 ++---------- system/physmem.c | 10 ++++++++++ 2 files changed, 12 insertions(+), 10 deletions(-)
include/exec/cpu-common.h +2 -10
@@ -133,16 +133,8 @@ void cpu_address_space_destroy(CPUState *cpu, int asidx);void cpu_physical_memory_rw (hwaddr addr,void *buf,hwaddr len,bool is_write); -static inline void cpu_physical_memory_read (hwaddr addr, -void *buf,hwaddr len) -{ -cpu_physical_memory_rw (addr, buf, len, false); -} -static inline void cpu_physical_memory_write (hwaddr addr, -const void *buf,hwaddr len) -{ -cpu_physical_memory_rw (addr, (void *)buf, len, true); -} +void cpu_physical_memory_read (hwaddr addr,void *buf,hwaddr len); +void cpu_physical_memory_write (hwaddr addr,const void *buf,hwaddr len);void *cpu_physical_memory_map (hwaddr addr,hwaddr *plen,bool is_write);
system/physmem.c +10 -0
@@ -3189,6 +3189,16 @@ void cpu_physical_memory_rw(hwaddr addr, void *buf, buf, len, is_write); } +void cpu_physical_memory_read (hwaddr addr,void *buf,hwaddr len) +{ +cpu_physical_memory_rw (addr, buf, len, false); +} + +void cpu_physical_memory_write (hwaddr addr,const void *buf,hwaddr len) +{ +cpu_physical_memory_rw (addr, (void *)buf, len, true); +} +/* used for ROM loading : can write in RAM and ROM */ MemTxResult address_space_write_rom(AddressSpace *as, hwaddr addr, MemTxAttrs attrs,
Signature · 1 line
2.51.0
Re: [PATCH v3 14/18] system/physmem: Un-inline cpu_physical_memory_read/write()
Quoted 10 lines · Philippe Mathieu-Daudé
In order to remove
cpu_physical_memory_rw()in a pair of commits, and due to a cyclic dependency between "exec/cpu-common.h" and "system/memory.h", un-inlinecpu_physical_memory_read()andcpu_physical_memory_write()as a prerequired step.Signed-off-by Philippe Mathieu-Daudé <[email protected]>--- include/exec/cpu-common.h | 12 ++---------- system/physmem.c | 10 ++++++++++ 2 files changed, 12 insertions(+), 10 deletions(-)
r~
PATCH v3 15/18 system/physmem: Avoid cpu_physical_memory_rw when is_write is constant
In reply to Philippe Mathieu-DaudéFollowing the mechanical changes of commit adeefe01671 ("Avoid cpu_physical_memory_rw() with a constant is_write argument"), replace:
- cpu_physical_memory_rw(, is_write=false) -> address_space_read()
- cpu_physical_memory_rw(, is_write=true) -> address_space_write()
--- scripts/coccinelle/exec_rw_const.cocci | 12 ------------ system/physmem.c | 6 ++++-- 2 files changed, 4 insertions(+), 14 deletions(-)
scripts/coccinelle/exec_rw_const.cocci +0 -12
@@ -62,18 +62,6 @@ symbol true, false; + address_space_write(E1, E2, E3, E4, E5) ) -// Avoid uses of cpu_physical_memory_rw() with a constant is_write argument. -@@ -expression E1, E2, E3; -@@ -( -- cpu_physical_memory_rw(E1, E2, E3, false) -+ cpu_physical_memory_read(E1, E2, E3) -| -- cpu_physical_memory_rw(E1, E2, E3, true) -+ cpu_physical_memory_write(E1, E2, E3) -) - // Remove useless cast @@ expression E1, E2, E3, E4, E5, E6;
system/physmem.c +4 -2
@@ -3191,12 +3191,14 @@ void cpu_physical_memory_rw(hwaddr addr, void *buf,void cpu_physical_memory_read (hwaddr addr,void *buf,hwaddr len) { -cpu_physical_memory_rw (addr, buf, len, false); +address_space_read (&address_space_memory, addr, +MEMTXATTRS_UNSPECIFIED , buf, len); }void cpu_physical_memory_write (hwaddr addr,const void *buf,hwaddr len) { -cpu_physical_memory_rw (addr, (void *)buf, len, true); +address_space_write (&address_space_memory, addr, +MEMTXATTRS_UNSPECIFIED , buf, len); }/* used for ROM loading : can write in RAM and ROM */
Signature · 1 line
2.51.0
Re: [PATCH v3 15/18] system/physmem: Avoid cpu_physical_memory_rw when is_write is constant
Quoted 4 lines · Philippe Mathieu-Daudé
Following the mechanical changes of commit
adeefe01671("Avoidcpu_physical_memory_rw()with a constantis_writeargument"), replace:
Quoted 8 lines
-
cpu_physical_memory_rw(, is_write=false)->address_space_read()
-cpu_physical_memory_rw(, is_write=true)->address_space_write()Signed-off-by Philippe Mathieu-Daudé <[email protected]>--- scripts/coccinelle/exec_rw_const.cocci | 12 ------------ system/physmem.c | 6 ++++-- 2 files changed, 4 insertions(+), 14 deletions(-)
r~
PATCH v3 16/18 system/physmem: Remove legacy cpu_physical_memory_rw()
In reply to Philippe Mathieu-DaudéThe legacy cpu_physical_memory_rw() method is no more used, remove it.
--- docs/devel/loads-stores.rst | 4 +--- scripts/coccinelle/exec_rw_const.cocci | 10 ---------- include/exec/cpu-common.h | 2 -- system/physmem.c | 7 ------- 4 files changed, 1 insertion(+), 22 deletions(-)
docs/devel/loads-stores.rst +1 -3
@@ -460,10 +460,8 @@ For new code they are better avoided: ``cpu_physical_memory_write`` -``cpu_physical_memory_rw`` - Regexes for git grep: - - ``\<cpu_physical_memory_\(read\|write\|rw\)\>`` + - ``\<cpu_physical_memory_\(read\|write\)\>`` ``cpu_memory_rw_debug`` ~~~~~~~~~~~~~~~~~~~~~~~
scripts/coccinelle/exec_rw_const.cocci +0 -10
@@ -21,13 +21,6 @@ expression E1, E2, E3, E4, E5; + address_space_rw(E1, E2, E3, E4, E5, true) | -- cpu_physical_memory_rw(E1, E2, E3, 0) -+ cpu_physical_memory_rw(E1, E2, E3, false) -| -- cpu_physical_memory_rw(E1, E2, E3, 1) -+ cpu_physical_memory_rw(E1, E2, E3, true) -| - - cpu_physical_memory_map(E1, E2, 0) + cpu_physical_memory_map(E1, E2, false) |
@@ -81,9 +74,6 @@ type T; + address_space_write_rom(E1, E2, E3, E4, E5) | -- cpu_physical_memory_rw(E1, (T *)(E2), E3, E4) -+ cpu_physical_memory_rw(E1, E2, E3, E4) -| - cpu_physical_memory_read(E1, (T *)(E2), E3) + cpu_physical_memory_read(E1, E2, E3) |
include/exec/cpu-common.h +0 -2
@@ -131,8 +131,6 @@ void cpu_address_space_init(CPUState *cpu, int asidx,*/ void cpu_address_space_destroy (CPUState *cpu,int asidx); -void cpu_physical_memory_rw (hwaddr addr,void *buf, -hwaddr len,bool is_write);void cpu_physical_memory_read (hwaddr addr,void *buf,hwaddr len);void cpu_physical_memory_write (hwaddr addr,const void *buf,hwaddr len); void *cpu_physical_memory_map(hwaddr addr,
system/physmem.c +0 -7
@@ -3182,13 +3182,6 @@ MemTxResult address_space_set(AddressSpace *as, hwaddr addr,return error; } -void cpu_physical_memory_rw (hwaddr addr,void *buf, -hwaddr len,bool is_write) -{ -address_space_rw (&address_space_memory, addr,MEMTXATTRS_UNSPECIFIED , - buf, len, is_write); -} -void cpu_physical_memory_read (hwaddr addr,void *buf,hwaddr len) { address_space_read(&address_space_memory, addr,
Signature · 1 line
2.51.0
Re: [PATCH v3 16/18] system/physmem: Remove legacy cpu_physical_memory_rw()
Quoted 10 lines · Philippe Mathieu-Daudé
The legacy
cpu_physical_memory_rw()method is no more used, remove it.Signed-off-by Philippe Mathieu-Daudé<[email protected]>--- docs/devel/loads-stores.rst | 4 +--- scripts/coccinelle/exec_rw_const.cocci | 10 ---------- include/exec/cpu-common.h | 2 -- system/physmem.c | 7 ------- 4 files changed, 1 insertion(+), 22 deletions(-)
r~
PATCH v3 17/18 hw/virtio/vhost: Replace legacy cpu_physical_memory_*map() calls
In reply to Philippe Mathieu-DaudéUse VirtIODevice::dma_as address space to convert the legacy cpu_physical_memory_[un]map() calls to address_space_[un]map().
--- hw/virtio/vhost.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
hw/virtio/vhost.c +5 -2
@@ -27,6 +27,7 @@#include "migration/blocker.h" #include "migration/qemu-file-types.h" #include "system/dma.h" +#include "system/memory.h" #include "trace.h" /* enabled until disconnected backend stabilizes */
@@ -455,7 +456,8 @@ static void *vhost_memory_map(struct vhost_dev *dev, hwaddr addr, hwaddr *plen, bool is_write) {if (!vhost_dev_has_iommu (dev)) { -return cpu_physical_memory_map (addr, plen, is_write); +return address_space_map (dev->vdev->dma_as, addr, plen, is_write, +MEMTXATTRS_UNSPECIFIED ); }else {return (void *)(uintptr_t )addr; }
@@ -466,7 +468,8 @@ static void vhost_memory_unmap(struct vhost_dev *dev, void *buffer, hwaddr access_len) {if (!vhost_dev_has_iommu (dev)) { -cpu_physical_memory_unmap (buffer, len, is_write, access_len); +address_space_unmap (dev->vdev->dma_as, buffer, len, is_write, + access_len); } }
Signature · 1 line
2.51.0
Re: [PATCH v3 17/18] hw/virtio/vhost: Replace legacy cpu_physical_memory_*map() calls
Quoted 7 lines · Philippe Mathieu-Daudé
Use
VirtIODevice::dma_asaddress space to convert the legacy cpu_physical_memory_[un]map() calls to address_space_[un]map().Signed-off-by Philippe Mathieu-Daudé <[email protected]>--- hw/virtio/vhost.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
r~
PATCH v3 18/18 hw/virtio/virtio: Replace legacy cpu_physical_memory_map() call
In reply to Philippe Mathieu-DaudéPropagate VirtIODevice::dma_as to virtqueue_undo_map_desc() in order to replace the legacy cpu_physical_memory_unmap() call by address_space_unmap().
--- hw/virtio/virtio.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-)
hw/virtio/virtio.c +6 -4
@@ -31,6 +31,7 @@#include "hw/qdev-properties.h" #include "hw/virtio/virtio-access.h" #include "system/dma.h" +#include "system/memory.h" #include "system/runstate.h" #include "virtio-qmp.h"
@@ -1622,7 +1623,8 @@ out:* virtqueue_unmap_sg() can't be used). Assumes buffers weren't written to * yet. */ -static void virtqueue_undo_map_desc (unsigned int out_num,unsigned int in_num, +static void virtqueue_undo_map_desc (AddressSpace *as, +unsigned int out_num,unsigned int in_num,struct iovec *iov) {unsigned int i;
@@ -1630,7 +1632,7 @@ static void virtqueue_undo_map_desc(unsigned int out_num, unsigned int in_num,for (i =0 ; i < out_num + in_num; i++) {int is_write = i >= out_num; -cpu_physical_memory_unmap (iov->iov_base, iov->iov_len, is_write,0 ); +address_space_unmap (as, iov->iov_base, iov->iov_len, is_write,0 ); iov++; } }
@@ -1832,7 +1834,7 @@ done:return elem; err_undo_map: -virtqueue_undo_map_desc (out_num, in_num, iov); +virtqueue_undo_map_desc (vdev->dma_as, out_num, in_num, iov); goto done; }
@@ -1982,7 +1984,7 @@ done:return elem; err_undo_map: -virtqueue_undo_map_desc (out_num, in_num, iov); +virtqueue_undo_map_desc (vdev->dma_as, out_num, in_num, iov); goto done; }
Signature · 1 line
2.51.0
Re: [PATCH v3 18/18] hw/virtio/virtio: Replace legacy cpu_physical_memory_map() call
Quoted 8 lines · Philippe Mathieu-Daudé
Propagate
VirtIODevice::dma_astovirtqueue_undo_map_desc()in order to replace the legacycpu_physical_memory_unmap()call byaddress_space_unmap().Signed-off-by Philippe Mathieu-Daudé <[email protected]>--- hw/virtio/virtio.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-)
r~