* Re: [PATCH v4] memory: Directly dispatch alias accesses on origin memory region
2021-04-18 5:57 [PATCH v4] memory: Directly dispatch alias accesses on origin memory region Philippe Mathieu-Daudé
@ 2021-04-20 21:08 ` Peter Xu
2021-07-07 13:58 ` Philippe Mathieu-Daudé
` (2 subsequent siblings)
3 siblings, 0 replies; 6+ messages in thread
From: Peter Xu @ 2021-04-20 21:08 UTC (permalink / raw)
To: Philippe Mathieu-Daudé; +Cc: Paolo Bonzini, Richard Henderson, qemu-devel
On Sun, Apr 18, 2021 at 07:57:08AM +0200, Philippe Mathieu-Daudé wrote:
> Since commit 2cdfcf272d ("memory: assign MemoryRegionOps to all
> regions"), all newly created regions are assigned with
> unassigned_mem_ops (which might be then overwritten).
>
> When using aliased container regions, and there is no region mapped
> at address 0 in the container, the memory_region_dispatch_read()
> and memory_region_dispatch_write() calls incorrectly return the
> container unassigned_mem_ops, because the alias offset is not used.
>
> Consider the following setup:
>
> +--------------------+ < - - - - - - - - - - - +
> | Container | mr
> | (unassigned_mem) | |
> | |
> | | |
> | | alias_offset
> + + <- - - - - - +----------+---------+
> | +----------------+ | | |
> | | MemoryRegion0 | | | |
> | +----------------+ | | Alias | addr1
> | | MemoryRegion1 | | <~ ~ ~ ~ ~ | | <~~~~~~
> | +----------------+ | | |
> | | +--------------------+
> | |
> | |
> | |
> | |
> | +----------------+ |
> | | MemoryRegionX | |
> | +----------------+ |
> | | MemoryRegionY | |
> | +----------------+ |
> | | MemoryRegionZ | |
> | +----------------+ |
> +--------------------+
>
> The memory_region_init_alias() flow is:
>
> memory_region_init_alias()
> -> memory_region_init()
> -> object_initialize(TYPE_MEMORY_REGION)
> -> memory_region_initfn()
> -> mr->ops = &unassigned_mem_ops;
>
> Later when accessing offset=addr1 via the alias, we expect to hit
> MemoryRegion1. The memory_region_dispatch_read() flow is:
>
> memory_region_dispatch_read(addr1)
> -> memory_region_access_valid(mr) <- addr1 offset is ignored
> -> mr->ops->valid.accepts()
> -> unassigned_mem_accepts()
> <- false
> <- false
> <- MEMTX_DECODE_ERROR
>
> The caller gets a MEMTX_DECODE_ERROR while the access is OK.
>
> Fix by dispatching aliases recursively, accessing its origin region
> after adding the alias offset.
>
> Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Peter Xu <peterx@redhat.com>
--
Peter Xu
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH v4] memory: Directly dispatch alias accesses on origin memory region
2021-04-18 5:57 [PATCH v4] memory: Directly dispatch alias accesses on origin memory region Philippe Mathieu-Daudé
2021-04-20 21:08 ` Peter Xu
@ 2021-07-07 13:58 ` Philippe Mathieu-Daudé
2021-08-04 21:11 ` [PATCH-for-6.2 " Philippe Mathieu-Daudé
2021-10-14 23:46 ` [PATCH " Philippe Mathieu-Daudé
2022-01-18 9:35 ` Philippe Mathieu-Daudé via
3 siblings, 1 reply; 6+ messages in thread
From: Philippe Mathieu-Daudé @ 2021-07-07 13:58 UTC (permalink / raw)
To: qemu-devel; +Cc: Paolo Bonzini, Richard Henderson, Peter Xu
Peter Xu already reviewed, but Cc'ing Peter Maydell too due to
his last comment on v3:
https://www.mail-archive.com/qemu-devel@nongnu.org/msg800482.html
On 4/18/21 7:57 AM, Philippe Mathieu-Daudé wrote:
> Since commit 2cdfcf272d ("memory: assign MemoryRegionOps to all
> regions"), all newly created regions are assigned with
> unassigned_mem_ops (which might be then overwritten).
>
> When using aliased container regions, and there is no region mapped
> at address 0 in the container, the memory_region_dispatch_read()
> and memory_region_dispatch_write() calls incorrectly return the
> container unassigned_mem_ops, because the alias offset is not used.
>
> Consider the following setup:
>
> +--------------------+ < - - - - - - - - - - - +
> | Container | mr
> | (unassigned_mem) | |
> | |
> | | |
> | | alias_offset
> + + <- - - - - - +----------+---------+
> | +----------------+ | | |
> | | MemoryRegion0 | | | |
> | +----------------+ | | Alias | addr1
> | | MemoryRegion1 | | <~ ~ ~ ~ ~ | | <~~~~~~
> | +----------------+ | | |
> | | +--------------------+
> | |
> | |
> | |
> | |
> | +----------------+ |
> | | MemoryRegionX | |
> | +----------------+ |
> | | MemoryRegionY | |
> | +----------------+ |
> | | MemoryRegionZ | |
> | +----------------+ |
> +--------------------+
>
> The memory_region_init_alias() flow is:
>
> memory_region_init_alias()
> -> memory_region_init()
> -> object_initialize(TYPE_MEMORY_REGION)
> -> memory_region_initfn()
> -> mr->ops = &unassigned_mem_ops;
>
> Later when accessing offset=addr1 via the alias, we expect to hit
> MemoryRegion1. The memory_region_dispatch_read() flow is:
>
> memory_region_dispatch_read(addr1)
> -> memory_region_access_valid(mr) <- addr1 offset is ignored
> -> mr->ops->valid.accepts()
> -> unassigned_mem_accepts()
> <- false
> <- false
> <- MEMTX_DECODE_ERROR
>
> The caller gets a MEMTX_DECODE_ERROR while the access is OK.
>
> Fix by dispatching aliases recursively, accessing its origin region
> after adding the alias offset.
>
> Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
> ---
> v4:
> - added ASCII schema
> v3:
> - reworded, mentioning the "alias to container" case
> - use recursive call instead of while(), because easier when debugging
> therefore reset Richard R-b tag.
> v2:
> - use while()
> ---
> softmmu/memory.c | 10 ++++++++++
> 1 file changed, 10 insertions(+)
>
> diff --git a/softmmu/memory.c b/softmmu/memory.c
> index d4493ef9e43..b899ca6a6b7 100644
> --- a/softmmu/memory.c
> +++ b/softmmu/memory.c
> @@ -1442,6 +1442,11 @@ MemTxResult memory_region_dispatch_read(MemoryRegion *mr,
> unsigned size = memop_size(op);
> MemTxResult r;
>
> + if (mr->alias) {
> + return memory_region_dispatch_read(mr->alias,
> + mr->alias_offset + addr,
> + pval, op, attrs);
> + }
> if (!memory_region_access_valid(mr, addr, size, false, attrs)) {
> *pval = unassigned_mem_read(mr, addr, size);
> return MEMTX_DECODE_ERROR;
> @@ -1486,6 +1491,11 @@ MemTxResult memory_region_dispatch_write(MemoryRegion *mr,
> {
> unsigned size = memop_size(op);
>
> + if (mr->alias) {
> + return memory_region_dispatch_write(mr->alias,
> + mr->alias_offset + addr,
> + data, op, attrs);
> + }
> if (!memory_region_access_valid(mr, addr, size, true, attrs)) {
> unassigned_mem_write(mr, addr, data, size);
> return MEMTX_DECODE_ERROR;
>
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH-for-6.2 v4] memory: Directly dispatch alias accesses on origin memory region
2021-07-07 13:58 ` Philippe Mathieu-Daudé
@ 2021-08-04 21:11 ` Philippe Mathieu-Daudé
0 siblings, 0 replies; 6+ messages in thread
From: Philippe Mathieu-Daudé @ 2021-08-04 21:11 UTC (permalink / raw)
To: qemu-devel; +Cc: Paolo Bonzini, Richard Henderson, Peter Xu, David Hildenbrand
Let's be honest, this won't make 6.1, so update the subject to 6.2,
as some scan the list for "6.2" in the subject when 6.1 release
is out.
On 7/7/21 3:58 PM, Philippe Mathieu-Daudé wrote:
> Peter Xu already reviewed, but Cc'ing Peter Maydell too due to
> his last comment on v3:
> https://www.mail-archive.com/qemu-devel@nongnu.org/msg800482.html
>
> On 4/18/21 7:57 AM, Philippe Mathieu-Daudé wrote:
>> Since commit 2cdfcf272d ("memory: assign MemoryRegionOps to all
>> regions"), all newly created regions are assigned with
>> unassigned_mem_ops (which might be then overwritten).
>>
>> When using aliased container regions, and there is no region mapped
>> at address 0 in the container, the memory_region_dispatch_read()
>> and memory_region_dispatch_write() calls incorrectly return the
>> container unassigned_mem_ops, because the alias offset is not used.
>>
>> Consider the following setup:
>>
>> +--------------------+ < - - - - - - - - - - - +
>> | Container | mr
>> | (unassigned_mem) | |
>> | |
>> | | |
>> | | alias_offset
>> + + <- - - - - - +----------+---------+
>> | +----------------+ | | |
>> | | MemoryRegion0 | | | |
>> | +----------------+ | | Alias | addr1
>> | | MemoryRegion1 | | <~ ~ ~ ~ ~ | | <~~~~~~
>> | +----------------+ | | |
>> | | +--------------------+
>> | |
>> | |
>> | |
>> | |
>> | +----------------+ |
>> | | MemoryRegionX | |
>> | +----------------+ |
>> | | MemoryRegionY | |
>> | +----------------+ |
>> | | MemoryRegionZ | |
>> | +----------------+ |
>> +--------------------+
>>
>> The memory_region_init_alias() flow is:
>>
>> memory_region_init_alias()
>> -> memory_region_init()
>> -> object_initialize(TYPE_MEMORY_REGION)
>> -> memory_region_initfn()
>> -> mr->ops = &unassigned_mem_ops;
>>
>> Later when accessing offset=addr1 via the alias, we expect to hit
>> MemoryRegion1. The memory_region_dispatch_read() flow is:
>>
>> memory_region_dispatch_read(addr1)
>> -> memory_region_access_valid(mr) <- addr1 offset is ignored
>> -> mr->ops->valid.accepts()
>> -> unassigned_mem_accepts()
>> <- false
>> <- false
>> <- MEMTX_DECODE_ERROR
>>
>> The caller gets a MEMTX_DECODE_ERROR while the access is OK.
>>
>> Fix by dispatching aliases recursively, accessing its origin region
>> after adding the alias offset.
>>
>> Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
>> ---
>> v4:
>> - added ASCII schema
>> v3:
>> - reworded, mentioning the "alias to container" case
>> - use recursive call instead of while(), because easier when debugging
>> therefore reset Richard R-b tag.
>> v2:
>> - use while()
>> ---
>> softmmu/memory.c | 10 ++++++++++
>> 1 file changed, 10 insertions(+)
>>
>> diff --git a/softmmu/memory.c b/softmmu/memory.c
>> index d4493ef9e43..b899ca6a6b7 100644
>> --- a/softmmu/memory.c
>> +++ b/softmmu/memory.c
>> @@ -1442,6 +1442,11 @@ MemTxResult memory_region_dispatch_read(MemoryRegion *mr,
>> unsigned size = memop_size(op);
>> MemTxResult r;
>>
>> + if (mr->alias) {
>> + return memory_region_dispatch_read(mr->alias,
>> + mr->alias_offset + addr,
>> + pval, op, attrs);
>> + }
>> if (!memory_region_access_valid(mr, addr, size, false, attrs)) {
>> *pval = unassigned_mem_read(mr, addr, size);
>> return MEMTX_DECODE_ERROR;
>> @@ -1486,6 +1491,11 @@ MemTxResult memory_region_dispatch_write(MemoryRegion *mr,
>> {
>> unsigned size = memop_size(op);
>>
>> + if (mr->alias) {
>> + return memory_region_dispatch_write(mr->alias,
>> + mr->alias_offset + addr,
>> + data, op, attrs);
>> + }
>> if (!memory_region_access_valid(mr, addr, size, true, attrs)) {
>> unassigned_mem_write(mr, addr, data, size);
>> return MEMTX_DECODE_ERROR;
>>
>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH v4] memory: Directly dispatch alias accesses on origin memory region
2021-04-18 5:57 [PATCH v4] memory: Directly dispatch alias accesses on origin memory region Philippe Mathieu-Daudé
2021-04-20 21:08 ` Peter Xu
2021-07-07 13:58 ` Philippe Mathieu-Daudé
@ 2021-10-14 23:46 ` Philippe Mathieu-Daudé
2022-01-18 9:35 ` Philippe Mathieu-Daudé via
3 siblings, 0 replies; 6+ messages in thread
From: Philippe Mathieu-Daudé @ 2021-10-14 23:46 UTC (permalink / raw)
To: qemu-devel@nongnu.org Developers
Cc: Paolo Bonzini, Richard Henderson, Peter Xu
ping?
On Sun, Apr 18, 2021 at 7:57 AM Philippe Mathieu-Daudé <f4bug@amsat.org> wrote:
>
> Since commit 2cdfcf272d ("memory: assign MemoryRegionOps to all
> regions"), all newly created regions are assigned with
> unassigned_mem_ops (which might be then overwritten).
>
> When using aliased container regions, and there is no region mapped
> at address 0 in the container, the memory_region_dispatch_read()
> and memory_region_dispatch_write() calls incorrectly return the
> container unassigned_mem_ops, because the alias offset is not used.
>
> Consider the following setup:
>
> +--------------------+ < - - - - - - - - - - - +
> | Container | mr
> | (unassigned_mem) | |
> | |
> | | |
> | | alias_offset
> + + <- - - - - - +----------+---------+
> | +----------------+ | | |
> | | MemoryRegion0 | | | |
> | +----------------+ | | Alias | addr1
> | | MemoryRegion1 | | <~ ~ ~ ~ ~ | | <~~~~~~
> | +----------------+ | | |
> | | +--------------------+
> | |
> | |
> | |
> | |
> | +----------------+ |
> | | MemoryRegionX | |
> | +----------------+ |
> | | MemoryRegionY | |
> | +----------------+ |
> | | MemoryRegionZ | |
> | +----------------+ |
> +--------------------+
>
> The memory_region_init_alias() flow is:
>
> memory_region_init_alias()
> -> memory_region_init()
> -> object_initialize(TYPE_MEMORY_REGION)
> -> memory_region_initfn()
> -> mr->ops = &unassigned_mem_ops;
>
> Later when accessing offset=addr1 via the alias, we expect to hit
> MemoryRegion1. The memory_region_dispatch_read() flow is:
>
> memory_region_dispatch_read(addr1)
> -> memory_region_access_valid(mr) <- addr1 offset is ignored
> -> mr->ops->valid.accepts()
> -> unassigned_mem_accepts()
> <- false
> <- false
> <- MEMTX_DECODE_ERROR
>
> The caller gets a MEMTX_DECODE_ERROR while the access is OK.
>
> Fix by dispatching aliases recursively, accessing its origin region
> after adding the alias offset.
>
> Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
> ---
> v4:
> - added ASCII schema
> v3:
> - reworded, mentioning the "alias to container" case
> - use recursive call instead of while(), because easier when debugging
> therefore reset Richard R-b tag.
> v2:
> - use while()
> ---
> softmmu/memory.c | 10 ++++++++++
> 1 file changed, 10 insertions(+)
>
> diff --git a/softmmu/memory.c b/softmmu/memory.c
> index d4493ef9e43..b899ca6a6b7 100644
> --- a/softmmu/memory.c
> +++ b/softmmu/memory.c
> @@ -1442,6 +1442,11 @@ MemTxResult memory_region_dispatch_read(MemoryRegion *mr,
> unsigned size = memop_size(op);
> MemTxResult r;
>
> + if (mr->alias) {
> + return memory_region_dispatch_read(mr->alias,
> + mr->alias_offset + addr,
> + pval, op, attrs);
> + }
> if (!memory_region_access_valid(mr, addr, size, false, attrs)) {
> *pval = unassigned_mem_read(mr, addr, size);
> return MEMTX_DECODE_ERROR;
> @@ -1486,6 +1491,11 @@ MemTxResult memory_region_dispatch_write(MemoryRegion *mr,
> {
> unsigned size = memop_size(op);
>
> + if (mr->alias) {
> + return memory_region_dispatch_write(mr->alias,
> + mr->alias_offset + addr,
> + data, op, attrs);
> + }
> if (!memory_region_access_valid(mr, addr, size, true, attrs)) {
> unassigned_mem_write(mr, addr, data, size);
> return MEMTX_DECODE_ERROR;
> --
> 2.26.3
>
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH v4] memory: Directly dispatch alias accesses on origin memory region
2021-04-18 5:57 [PATCH v4] memory: Directly dispatch alias accesses on origin memory region Philippe Mathieu-Daudé
` (2 preceding siblings ...)
2021-10-14 23:46 ` [PATCH " Philippe Mathieu-Daudé
@ 2022-01-18 9:35 ` Philippe Mathieu-Daudé via
3 siblings, 0 replies; 6+ messages in thread
From: Philippe Mathieu-Daudé via @ 2022-01-18 9:35 UTC (permalink / raw)
To: qemu-devel; +Cc: Richard Henderson, Peter Xu, Paolo Bonzini
On 4/18/21 07:57, Philippe Mathieu-Daudé wrote:
> Since commit 2cdfcf272d ("memory: assign MemoryRegionOps to all
> regions"), all newly created regions are assigned with
> unassigned_mem_ops (which might be then overwritten).
>
> When using aliased container regions, and there is no region mapped
> at address 0 in the container, the memory_region_dispatch_read()
> and memory_region_dispatch_write() calls incorrectly return the
> container unassigned_mem_ops, because the alias offset is not used.
>
> Consider the following setup:
>
> +--------------------+ < - - - - - - - - - - - +
> | Container | mr
> | (unassigned_mem) | |
> | |
> | | |
> | | alias_offset
> + + <- - - - - - +----------+---------+
> | +----------------+ | | |
> | | MemoryRegion0 | | | |
> | +----------------+ | | Alias | addr1
> | | MemoryRegion1 | | <~ ~ ~ ~ ~ | | <~~~~~~
> | +----------------+ | | |
> | | +--------------------+
> | |
> | |
> | |
> | |
> | +----------------+ |
> | | MemoryRegionX | |
> | +----------------+ |
> | | MemoryRegionY | |
> | +----------------+ |
> | | MemoryRegionZ | |
> | +----------------+ |
> +--------------------+
>
> The memory_region_init_alias() flow is:
>
> memory_region_init_alias()
> -> memory_region_init()
> -> object_initialize(TYPE_MEMORY_REGION)
> -> memory_region_initfn()
> -> mr->ops = &unassigned_mem_ops;
>
> Later when accessing offset=addr1 via the alias, we expect to hit
> MemoryRegion1. The memory_region_dispatch_read() flow is:
>
> memory_region_dispatch_read(addr1)
> -> memory_region_access_valid(mr) <- addr1 offset is ignored
> -> mr->ops->valid.accepts()
> -> unassigned_mem_accepts()
> <- false
> <- false
> <- MEMTX_DECODE_ERROR
>
> The caller gets a MEMTX_DECODE_ERROR while the access is OK.
>
> Fix by dispatching aliases recursively, accessing its origin region
> after adding the alias offset.
>
> Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
> ---
> softmmu/memory.c | 10 ++++++++++
> 1 file changed, 10 insertions(+)
Queued via memory-api.
^ permalink raw reply [flat|nested] 6+ messages in thread