[Qemu-devel] [PATCH] riscv: rv32: Root page table address can be larger than 32-bit

Bin Meng posted 1 patch 4 years, 8 months ago
Test asan passed
Test docker-mingw@fedora passed
Test docker-clang@ubuntu passed
Test FreeBSD passed
Test checkpatch passed
Test s390x passed
Patches applied successfully (tree, apply log)
git fetch https://github.com/patchew-project/qemu tags/patchew/1564577101-29020-1-git-send-email-bmeng.cn@gmail.com
Maintainers: Alistair Francis <Alistair.Francis@wdc.com>, Bastian Koppelmann <kbastian@mail.uni-paderborn.de>, Sagar Karandikar <sagark@eecs.berkeley.edu>, Palmer Dabbelt <palmer@sifive.com>
There is a newer version of this series
target/riscv/cpu_helper.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
[Qemu-devel] [PATCH] riscv: rv32: Root page table address can be larger than 32-bit
Posted by Bin Meng 4 years, 8 months ago
For RV32, the root page table's PPN has 22 bits hence its address
bits could be larger than the maximum bits that target_ulong is
able to represent. Use hwaddr instead.

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
---

 target/riscv/cpu_helper.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
index e32b612..3150a6a 100644
--- a/target/riscv/cpu_helper.c
+++ b/target/riscv/cpu_helper.c
@@ -176,7 +176,7 @@ static int get_physical_address(CPURISCVState *env, hwaddr *physical,
 
     *prot = 0;
 
-    target_ulong base;
+    hwaddr base;
     int levels, ptidxbits, ptesize, vm, sum;
     int mxr = get_field(env->mstatus, MSTATUS_MXR);
 
@@ -239,7 +239,7 @@ restart:
                            ((1 << ptidxbits) - 1);
 
         /* check that physical address of PTE is legal */
-        target_ulong pte_addr = base + idx * ptesize;
+        hwaddr pte_addr = base + idx * ptesize;
 
         if (riscv_feature(env, RISCV_FEATURE_PMP) &&
             !pmp_hart_has_privs(env, pte_addr, sizeof(target_ulong),
-- 
2.7.4


Re: [Qemu-devel] [PATCH] riscv: rv32: Root page table address can be larger than 32-bit
Posted by Richard Henderson 4 years, 8 months ago
On 7/31/19 5:45 AM, Bin Meng wrote:
> -    target_ulong base;
> +    hwaddr base;
...
> -        target_ulong pte_addr = base + idx * ptesize;
> +        hwaddr pte_addr = base + idx * ptesize;

I believe that you either need

    base + (hwaddr)idx * ptesize

or change the type of idx to hwaddr above.

Otherwise the multiply overflows before it gets promoted with the add.


r~

Re: [Qemu-devel] [PATCH] riscv: rv32: Root page table address can be larger than 32-bit
Posted by Bin Meng 4 years, 7 months ago
Hi Richard,

On Thu, Aug 1, 2019 at 1:35 AM Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> On 7/31/19 5:45 AM, Bin Meng wrote:
> > -    target_ulong base;
> > +    hwaddr base;
> ...
> > -        target_ulong pte_addr = base + idx * ptesize;
> > +        hwaddr pte_addr = base + idx * ptesize;
>
> I believe that you either need
>
>     base + (hwaddr)idx * ptesize
>
> or change the type of idx to hwaddr above.
>
> Otherwise the multiply overflows before it gets promoted with the add.
>

I am not sure how (idx * ptesize) could overflow. It represents the
offset by a page table which is [0, 4096).

Regards,
Bin

Re: [Qemu-devel] [PATCH] riscv: rv32: Root page table address can be larger than 32-bit
Posted by Richard Henderson 4 years, 7 months ago
On 7/31/19 6:53 PM, Bin Meng wrote:
> I am not sure how (idx * ptesize) could overflow. It represents the
> offset by a page table which is [0, 4096).

You're right, I mis-read what was going on there.

However, lower down, "target_ulong ppn" needs to be promoted to hwaddr, so that

    ppn = pte >> PTE_PPN_SHIFT;
    ...
    base = ppn << PGSHIFT;

does not overflow.  (Which is the part of the page table walk that I thought I
had gleaned from the patch without actually reading the entire function.)


r~

Re: [Qemu-devel] [PATCH] riscv: rv32: Root page table address can be larger than 32-bit
Posted by Bin Meng 4 years, 7 months ago
On Thu, Aug 1, 2019 at 10:16 PM Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> On 7/31/19 6:53 PM, Bin Meng wrote:
> > I am not sure how (idx * ptesize) could overflow. It represents the
> > offset by a page table which is [0, 4096).
>
> You're right, I mis-read what was going on there.
>
> However, lower down, "target_ulong ppn" needs to be promoted to hwaddr, so that
>
>     ppn = pte >> PTE_PPN_SHIFT;
>     ...
>     base = ppn << PGSHIFT;
>
> does not overflow.  (Which is the part of the page table walk that I thought I
> had gleaned from the patch without actually reading the entire function.)

Ah, yes. ppn should be promoted. Thanks for the review!

Regards,
Bin

Re: [Qemu-devel] [PATCH] riscv: rv32: Root page table address can be larger than 32-bit
Posted by Palmer Dabbelt 4 years, 7 months ago
On Thu, Aug 1, 2019 at 7:58 AM Bin Meng <bmeng.cn@gmail.com> wrote:

> On Thu, Aug 1, 2019 at 10:16 PM Richard Henderson
> <richard.henderson@linaro.org> wrote:
> >
> > On 7/31/19 6:53 PM, Bin Meng wrote:
> > > I am not sure how (idx * ptesize) could overflow. It represents the
> > > offset by a page table which is [0, 4096).
> >
> > You're right, I mis-read what was going on there.
> >
> > However, lower down, "target_ulong ppn" needs to be promoted to hwaddr,
> so that
> >
> >     ppn = pte >> PTE_PPN_SHIFT;
> >     ...
> >     base = ppn << PGSHIFT;
> >
> > does not overflow.  (Which is the part of the page table walk that I
> thought I
> > had gleaned from the patch without actually reading the entire function.)
>
> Ah, yes. ppn should be promoted. Thanks for the review!
>

Did I miss a v2?
Re: [Qemu-devel] [PATCH] riscv: rv32: Root page table address can be larger than 32-bit
Posted by Bin Meng 4 years, 7 months ago
Hi Palmer,

On Thu, Aug 8, 2019 at 4:55 AM Palmer Dabbelt <palmer@sifive.com> wrote:
>
> On Thu, Aug 1, 2019 at 7:58 AM Bin Meng <bmeng.cn@gmail.com> wrote:
>>
>> On Thu, Aug 1, 2019 at 10:16 PM Richard Henderson
>> <richard.henderson@linaro.org> wrote:
>> >
>> > On 7/31/19 6:53 PM, Bin Meng wrote:
>> > > I am not sure how (idx * ptesize) could overflow. It represents the
>> > > offset by a page table which is [0, 4096).
>> >
>> > You're right, I mis-read what was going on there.
>> >
>> > However, lower down, "target_ulong ppn" needs to be promoted to hwaddr, so that
>> >
>> >     ppn = pte >> PTE_PPN_SHIFT;
>> >     ...
>> >     base = ppn << PGSHIFT;
>> >
>> > does not overflow.  (Which is the part of the page table walk that I thought I
>> > had gleaned from the patch without actually reading the entire function.)
>>
>> Ah, yes. ppn should be promoted. Thanks for the review!
>
>
> Did I miss a v2?

No, I will send a v2 soon.

Regards,
Bin

[Qemu-devel] [PATCH v2] riscv: rv32: Root page table address can be larger than 32-bit
Posted by Bin Meng 4 years, 7 months ago
For RV32, the root page table's PPN has 22 bits hence its address
bits could be larger than the maximum bits that target_ulong is
able to represent. Use hwaddr instead.

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>

---

Changes in v2:
- promote ppn, env->satp/env->sptbl to hwaddr otherwise the page
  table base will not be correctly calculated

 target/riscv/cpu_helper.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
index e32b612..b2b4f3a 100644
--- a/target/riscv/cpu_helper.c
+++ b/target/riscv/cpu_helper.c
@@ -176,12 +176,12 @@ static int get_physical_address(CPURISCVState *env, hwaddr *physical,
 
     *prot = 0;
 
-    target_ulong base;
+    hwaddr base;
     int levels, ptidxbits, ptesize, vm, sum;
     int mxr = get_field(env->mstatus, MSTATUS_MXR);
 
     if (env->priv_ver >= PRIV_VERSION_1_10_0) {
-        base = get_field(env->satp, SATP_PPN) << PGSHIFT;
+        base = (hwaddr)get_field(env->satp, SATP_PPN) << PGSHIFT;
         sum = get_field(env->mstatus, MSTATUS_SUM);
         vm = get_field(env->satp, SATP_MODE);
         switch (vm) {
@@ -201,7 +201,7 @@ static int get_physical_address(CPURISCVState *env, hwaddr *physical,
           g_assert_not_reached();
         }
     } else {
-        base = env->sptbr << PGSHIFT;
+        base = (hwaddr)(env->sptbr) << PGSHIFT;
         sum = !get_field(env->mstatus, MSTATUS_PUM);
         vm = get_field(env->mstatus, MSTATUS_VM);
         switch (vm) {
@@ -239,7 +239,7 @@ restart:
                            ((1 << ptidxbits) - 1);
 
         /* check that physical address of PTE is legal */
-        target_ulong pte_addr = base + idx * ptesize;
+        hwaddr pte_addr = base + idx * ptesize;
 
         if (riscv_feature(env, RISCV_FEATURE_PMP) &&
             !pmp_hart_has_privs(env, pte_addr, sizeof(target_ulong),
@@ -251,7 +251,7 @@ restart:
 #elif defined(TARGET_RISCV64)
         target_ulong pte = ldq_phys(cs->as, pte_addr);
 #endif
-        target_ulong ppn = pte >> PTE_PPN_SHIFT;
+        hwaddr ppn = pte >> PTE_PPN_SHIFT;
 
         if (!(pte & PTE_V)) {
             /* Invalid PTE */
-- 
2.7.4


Re: [Qemu-devel] [PATCH v2] riscv: rv32: Root page table address can be larger than 32-bit
Posted by Alistair Francis 4 years, 7 months ago
On Wed, Aug 7, 2019 at 7:50 PM Bin Meng <bmeng.cn@gmail.com> wrote:
>
> For RV32, the root page table's PPN has 22 bits hence its address
> bits could be larger than the maximum bits that target_ulong is
> able to represent. Use hwaddr instead.
>
> Signed-off-by: Bin Meng <bmeng.cn@gmail.com>

Reviewed-by: Alistair Francis <alistair.francis@wdc.com>

Alistair

>
> ---
>
> Changes in v2:
> - promote ppn, env->satp/env->sptbl to hwaddr otherwise the page
>   table base will not be correctly calculated
>
>  target/riscv/cpu_helper.c | 10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)
>
> diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
> index e32b612..b2b4f3a 100644
> --- a/target/riscv/cpu_helper.c
> +++ b/target/riscv/cpu_helper.c
> @@ -176,12 +176,12 @@ static int get_physical_address(CPURISCVState *env, hwaddr *physical,
>
>      *prot = 0;
>
> -    target_ulong base;
> +    hwaddr base;
>      int levels, ptidxbits, ptesize, vm, sum;
>      int mxr = get_field(env->mstatus, MSTATUS_MXR);
>
>      if (env->priv_ver >= PRIV_VERSION_1_10_0) {
> -        base = get_field(env->satp, SATP_PPN) << PGSHIFT;
> +        base = (hwaddr)get_field(env->satp, SATP_PPN) << PGSHIFT;
>          sum = get_field(env->mstatus, MSTATUS_SUM);
>          vm = get_field(env->satp, SATP_MODE);
>          switch (vm) {
> @@ -201,7 +201,7 @@ static int get_physical_address(CPURISCVState *env, hwaddr *physical,
>            g_assert_not_reached();
>          }
>      } else {
> -        base = env->sptbr << PGSHIFT;
> +        base = (hwaddr)(env->sptbr) << PGSHIFT;
>          sum = !get_field(env->mstatus, MSTATUS_PUM);
>          vm = get_field(env->mstatus, MSTATUS_VM);
>          switch (vm) {
> @@ -239,7 +239,7 @@ restart:
>                             ((1 << ptidxbits) - 1);
>
>          /* check that physical address of PTE is legal */
> -        target_ulong pte_addr = base + idx * ptesize;
> +        hwaddr pte_addr = base + idx * ptesize;
>
>          if (riscv_feature(env, RISCV_FEATURE_PMP) &&
>              !pmp_hart_has_privs(env, pte_addr, sizeof(target_ulong),
> @@ -251,7 +251,7 @@ restart:
>  #elif defined(TARGET_RISCV64)
>          target_ulong pte = ldq_phys(cs->as, pte_addr);
>  #endif
> -        target_ulong ppn = pte >> PTE_PPN_SHIFT;
> +        hwaddr ppn = pte >> PTE_PPN_SHIFT;
>
>          if (!(pte & PTE_V)) {
>              /* Invalid PTE */
> --
> 2.7.4
>
>

Re: [Qemu-devel] [PATCH v2] riscv: rv32: Root page table address can be larger than 32-bit
Posted by Bin Meng 4 years, 7 months ago
Hi Palmer,

On Sat, Aug 10, 2019 at 9:49 AM Alistair Francis <alistair23@gmail.com> wrote:
>
> On Wed, Aug 7, 2019 at 7:50 PM Bin Meng <bmeng.cn@gmail.com> wrote:
> >
> > For RV32, the root page table's PPN has 22 bits hence its address
> > bits could be larger than the maximum bits that target_ulong is
> > able to represent. Use hwaddr instead.
> >
> > Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
>
> Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
>

Would you take this one too?

Regards,
Bin

Re: [Qemu-devel] [PATCH v2] riscv: rv32: Root page table address can be larger than 32-bit
Posted by Bin Meng 4 years, 7 months ago
On Wed, Aug 14, 2019 at 5:46 PM Bin Meng <bmeng.cn@gmail.com> wrote:
>
> Hi Palmer,
>
> On Sat, Aug 10, 2019 at 9:49 AM Alistair Francis <alistair23@gmail.com> wrote:
> >
> > On Wed, Aug 7, 2019 at 7:50 PM Bin Meng <bmeng.cn@gmail.com> wrote:
> > >
> > > For RV32, the root page table's PPN has 22 bits hence its address
> > > bits could be larger than the maximum bits that target_ulong is
> > > able to represent. Use hwaddr instead.
> > >
> > > Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
> >
> > Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
> >
>
> Would you take this one too?
>

Ping?

What's the status of this patch?

Regards,
Bin

Re: [Qemu-devel] [PATCH v2] riscv: rv32: Root page table address can be larger than 32-bit
Posted by Palmer Dabbelt 4 years, 7 months ago
On Sun, 18 Aug 2019 23:00:40 PDT (-0700), bmeng.cn@gmail.com wrote:
> On Wed, Aug 14, 2019 at 5:46 PM Bin Meng <bmeng.cn@gmail.com> wrote:
>>
>> Hi Palmer,
>>
>> On Sat, Aug 10, 2019 at 9:49 AM Alistair Francis <alistair23@gmail.com> wrote:
>> >
>> > On Wed, Aug 7, 2019 at 7:50 PM Bin Meng <bmeng.cn@gmail.com> wrote:
>> > >
>> > > For RV32, the root page table's PPN has 22 bits hence its address
>> > > bits could be larger than the maximum bits that target_ulong is
>> > > able to represent. Use hwaddr instead.
>> > >
>> > > Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
>> >
>> > Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
>> >
>>
>> Would you take this one too?
>>
>
> Ping?
>
> What's the status of this patch?

Also in the patch queue.