2023-07-18 16:47:50

by Celeste Liu

Subject: [PATCH v2] riscv: entry: set a0 prior to syscall_enter_from_user_mode

When we test seccomp with 6.4 kernel, we found errno has wrong value.
If we deny NETLINK_AUDIT with EAFNOSUPPORT, after f0bddf50586d, we will
get ENOSYS instead. We got same result with 9c2598d43510 ("riscv: entry: Save a0
prior syscall_enter_from_user_mode()").

After analysing code, we think that regs->a0 = -ENOSYS should be advanced before
syscall_enter_from_user_mode to fix this problem. In __seccomp_filter, when
seccomp rejected this syscall with specified errno, they will set a0 to return
number as syscall ABI, and then return -1. This return number is finally pass
as return number of syscall_enter_from_user_mode, and then is compared
with NR_syscalls after converted to ulong (so it will be ULONG_MAX).
The condition syscall < NR_syscalls will always be false, so regs->a0 = -ENOSYS
is always executable. It covered a0 set by seccomp, so we always get
ENOSYS when match seccomp RET_ERRNO rule.

Fixes: f0bddf50586d ("riscv: entry: Convert to generic entry")
Reported-by: Felix Yan <[email protected]>
Co-developed-by: Ruizhe Pan <[email protected]>
Signed-off-by: Ruizhe Pan <[email protected]>
Co-developed-by: Shiqi Zhang <[email protected]>
Signed-off-by: Shiqi Zhang <[email protected]>
Signed-off-by: Celeste Liu <[email protected]>
Tested-by: Felix Yan <[email protected]>
---

v1 -> v2: added explanation on why always got ENOSYS

arch/riscv/kernel/traps.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/arch/riscv/kernel/traps.c b/arch/riscv/kernel/traps.c
index f910dfccbf5d2..ccadb5ffd063c 100644
--- a/arch/riscv/kernel/traps.c
+++ b/arch/riscv/kernel/traps.c
@@ -301,6 +301,7 @@ asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)

regs->epc += 4;
regs->orig_a0 = regs->a0;
+ regs->a0 = -ENOSYS;

riscv_v_vstate_discard(regs);

@@ -308,8 +309,6 @@ asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)

if (syscall < NR_syscalls)
syscall_handler(regs, syscall);
- else
- regs->a0 = -ENOSYS;

syscall_exit_to_user_mode(regs);
} else {
--
2.41.0

2023-07-18 20:44:11

by Björn Töpel

Subject: Re: [PATCH v2] riscv: entry: set a0 prior to syscall_enter_from_user_mode

Celeste Liu <[email protected]> writes:

> When we test seccomp with 6.4 kernel, we found errno has wrong value.
> If we deny NETLINK_AUDIT with EAFNOSUPPORT, after f0bddf50586d, we will
> get ENOSYS instead. We got same result with 9c2598d43510 ("riscv: entry: Save a0
> prior syscall_enter_from_user_mode()").
>
> After analysing code, we think that regs->a0 = -ENOSYS should be advanced before
> syscall_enter_from_user_mode to fix this problem. In __seccomp_filter, when
> seccomp rejected this syscall with specified errno, they will set a0 to return
> number as syscall ABI, and then return -1. This return number is finally pass
> as return number of syscall_enter_from_user_mode, and then is compared
> with NR_syscalls after converted to ulong (so it will be ULONG_MAX).
> The condition syscall < NR_syscalls will always be false, so regs->a0 = -ENOSYS
> is always executable. It covered a0 set by seccomp, so we always get
> ENOSYS when match seccomp RET_ERRNO rule.

Isn't something like...

diff --git a/arch/riscv/kernel/traps.c b/arch/riscv/kernel/traps.c
index f910dfccbf5d..15a8b4898a6c 100644
--- a/arch/riscv/kernel/traps.c
+++ b/arch/riscv/kernel/traps.c
@@ -297,7 +297,7 @@ asmlinkage __visible __trap_section void do_trap_break(struct pt_regs *regs)
asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)
{
if (user_mode(regs)) {
- ulong syscall = regs->a7;
+ long syscall = regs->a7;

regs->epc += 4;
regs->orig_a0 = regs->a0;
@@ -306,7 +306,7 @@ asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)

syscall = syscall_enter_from_user_mode(regs, syscall);

- if (syscall < NR_syscalls)
+ if (syscall > -1 && syscall < NR_syscalls)
syscall_handler(regs, syscall);
else
regs->a0 = -ENOSYS;


...easier to read?


Björn

2023-07-18 21:05:10

by Björn Töpel

Subject: Re: [PATCH v2] riscv: entry: set a0 prior to syscall_enter_from_user_mode

Celeste Liu <[email protected]> writes:

> On 2023/7/19 03:35, Björn Töpel wrote:
>> Celeste Liu <[email protected]> writes:
>>
>>> When we test seccomp with 6.4 kernel, we found errno has wrong value.
>>> If we deny NETLINK_AUDIT with EAFNOSUPPORT, after f0bddf50586d, we will
>>> get ENOSYS instead. We got same result with 9c2598d43510 ("riscv: entry: Save a0
>>> prior syscall_enter_from_user_mode()").
>>>
>>> After analysing code, we think that regs->a0 = -ENOSYS should be advanced before
>>> syscall_enter_from_user_mode to fix this problem. In __seccomp_filter, when
>>> seccomp rejected this syscall with specified errno, they will set a0 to return
>>> number as syscall ABI, and then return -1. This return number is finally pass
>>> as return number of syscall_enter_from_user_mode, and then is compared
>>> with NR_syscalls after converted to ulong (so it will be ULONG_MAX).
>>> The condition syscall < NR_syscalls will always be false, so regs->a0 = -ENOSYS
>>> is always executable. It covered a0 set by seccomp, so we always get
>>> ENOSYS when match seccomp RET_ERRNO rule.
>>
>> Isn't something like...
>>
>> diff --git a/arch/riscv/kernel/traps.c b/arch/riscv/kernel/traps.c
>> index f910dfccbf5d..15a8b4898a6c 100644
>> --- a/arch/riscv/kernel/traps.c
>> +++ b/arch/riscv/kernel/traps.c
>> @@ -297,7 +297,7 @@ asmlinkage __visible __trap_section void do_trap_break(struct pt_regs *regs)
>> asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)
>> {
>> if (user_mode(regs)) {
>> - ulong syscall = regs->a7;
>> + long syscall = regs->a7;
>>
>> regs->epc += 4;
>> regs->orig_a0 = regs->a0;
>> @@ -306,7 +306,7 @@ asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)
>>
>> syscall = syscall_enter_from_user_mode(regs, syscall);
>>
>> - if (syscall < NR_syscalls)
>> + if (syscall > -1 && syscall < NR_syscalls)
>> syscall_handler(regs, syscall);
>> else
>> regs->a0 = -ENOSYS;
>>
>>
>> ...easier to read?
>>
>>
>> Björn
>
> It seems that your change turn it back to the beginning. If syscall == -1,
> it is supposed to go neither first nor else branch. It should do NOTHING.
> However it was still a great idea. It may be better to use a set of if-statement
> to clarify the logic.

Ah, gotcha! (Notice that arch/x86/entry/common.c has

| if (!do_syscall_x64(regs, nr) && !do_syscall_x32(regs, nr) && nr != -1) {

and in do_syscall_x64()

| /*
| * Convert negative numbers to very high and thus out of range
| * numbers for comparisons.
| */
| unsigned int unr = nr;


> diff --git a/arch/riscv/kernel/traps.c b/arch/riscv/kernel/traps.c
> index f910dfccbf5d2..d0bd725244594 100644
> --- a/arch/riscv/kernel/traps.c
> +++ b/arch/riscv/kernel/traps.c
> @@ -306,7 +306,9 @@ asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)
>
> syscall = syscall_enter_from_user_mode(regs, syscall);
>
> - if (syscall < NR_syscalls)
> + if (syscall == -1)
> + // Do nothing
> + else if (syscall < NR_syscalls)
> syscall_handler(regs, syscall);
> else
> regs->a0 = -ENOSYS;

Maybe something a bit more explicit?

diff --git a/arch/riscv/kernel/traps.c b/arch/riscv/kernel/traps.c
index f910dfccbf5d..5cef72874542 100644
--- a/arch/riscv/kernel/traps.c
+++ b/arch/riscv/kernel/traps.c
@@ -297,6 +297,10 @@ asmlinkage __visible __trap_section void do_trap_break(struct pt_regs *regs)
asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)
{
if (user_mode(regs)) {
+ /*
+ * Convert negative numbers to very high and thus out of range
+ * numbers for comparisons.
+ */
ulong syscall = regs->a7;

regs->epc += 4;
@@ -308,7 +312,7 @@ asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)

if (syscall < NR_syscalls)
syscall_handler(regs, syscall);
- else
+ else if ((long)syscall != -1L)
regs->a0 = -ENOSYS;

syscall_exit_to_user_mode(regs);


Björn

2023-07-18 21:09:39

by Celeste Liu

Subject: Re: [PATCH v2] riscv: entry: set a0 prior to syscall_enter_from_user_mode

On 2023/7/19 03:35, Björn Töpel wrote:
> Celeste Liu <[email protected]> writes:
>
>> When we test seccomp with 6.4 kernel, we found errno has wrong value.
>> If we deny NETLINK_AUDIT with EAFNOSUPPORT, after f0bddf50586d, we will
>> get ENOSYS instead. We got same result with 9c2598d43510 ("riscv: entry: Save a0
>> prior syscall_enter_from_user_mode()").
>>
>> After analysing code, we think that regs->a0 = -ENOSYS should be advanced before
>> syscall_enter_from_user_mode to fix this problem. In __seccomp_filter, when
>> seccomp rejected this syscall with specified errno, they will set a0 to return
>> number as syscall ABI, and then return -1. This return number is finally pass
>> as return number of syscall_enter_from_user_mode, and then is compared
>> with NR_syscalls after converted to ulong (so it will be ULONG_MAX).
>> The condition syscall < NR_syscalls will always be false, so regs->a0 = -ENOSYS
>> is always executable. It covered a0 set by seccomp, so we always get
>> ENOSYS when match seccomp RET_ERRNO rule.
>
> Isn't something like...
>
> diff --git a/arch/riscv/kernel/traps.c b/arch/riscv/kernel/traps.c
> index f910dfccbf5d..15a8b4898a6c 100644
> --- a/arch/riscv/kernel/traps.c
> +++ b/arch/riscv/kernel/traps.c
> @@ -297,7 +297,7 @@ asmlinkage __visible __trap_section void do_trap_break(struct pt_regs *regs)
> asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)
> {
> if (user_mode(regs)) {
> - ulong syscall = regs->a7;
> + long syscall = regs->a7;
>
> regs->epc += 4;
> regs->orig_a0 = regs->a0;
> @@ -306,7 +306,7 @@ asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)
>
> syscall = syscall_enter_from_user_mode(regs, syscall);
>
> - if (syscall < NR_syscalls)
> + if (syscall > -1 && syscall < NR_syscalls)
> syscall_handler(regs, syscall);
> else
> regs->a0 = -ENOSYS;
>
>
> ...easier to read?
>
>
> Björn

It seems that your change turn it back to the beginning. If syscall == -1,
it is supposed to go neither first nor else branch. It should do NOTHING.
However it was still a great idea. It may be better to use a set of if-statement
to clarify the logic.

diff --git a/arch/riscv/kernel/traps.c b/arch/riscv/kernel/traps.c
index f910dfccbf5d2..d0bd725244594 100644
--- a/arch/riscv/kernel/traps.c
+++ b/arch/riscv/kernel/traps.c
@@ -306,7 +306,9 @@ asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)

syscall = syscall_enter_from_user_mode(regs, syscall);

- if (syscall < NR_syscalls)
+ if (syscall == -1)
+ // Do nothing
+ else if (syscall < NR_syscalls)
syscall_handler(regs, syscall);
else
regs->a0 = -ENOSYS;

2023-07-18 21:10:08

by Celeste Liu

Subject: Re: [PATCH v2] riscv: entry: set a0 prior to syscall_enter_from_user_mode

On 2023/7/19 04:35, Celeste Liu wrote:
>
> On 2023/7/19 03:35, Björn Töpel wrote:
>> Celeste Liu <[email protected]> writes:
>>
>>> When we test seccomp with 6.4 kernel, we found errno has wrong value.
>>> If we deny NETLINK_AUDIT with EAFNOSUPPORT, after f0bddf50586d, we will
>>> get ENOSYS instead. We got same result with 9c2598d43510 ("riscv: entry: Save a0
>>> prior syscall_enter_from_user_mode()").
>>>
>>> After analysing code, we think that regs->a0 = -ENOSYS should be advanced before
>>> syscall_enter_from_user_mode to fix this problem. In __seccomp_filter, when
>>> seccomp rejected this syscall with specified errno, they will set a0 to return
>>> number as syscall ABI, and then return -1. This return number is finally pass
>>> as return number of syscall_enter_from_user_mode, and then is compared
>>> with NR_syscalls after converted to ulong (so it will be ULONG_MAX).
>>> The condition syscall < NR_syscalls will always be false, so regs->a0 = -ENOSYS
>>> is always executable. It covered a0 set by seccomp, so we always get
>>> ENOSYS when match seccomp RET_ERRNO rule.
>>
>> Isn't something like...
>>
>> diff --git a/arch/riscv/kernel/traps.c b/arch/riscv/kernel/traps.c
>> index f910dfccbf5d..15a8b4898a6c 100644
>> --- a/arch/riscv/kernel/traps.c
>> +++ b/arch/riscv/kernel/traps.c
>> @@ -297,7 +297,7 @@ asmlinkage __visible __trap_section void do_trap_break(struct pt_regs *regs)
>> asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)
>> {
>> if (user_mode(regs)) {
>> - ulong syscall = regs->a7;
>> + long syscall = regs->a7;
>>
>> regs->epc += 4;
>> regs->orig_a0 = regs->a0;
>> @@ -306,7 +306,7 @@ asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)
>>
>> syscall = syscall_enter_from_user_mode(regs, syscall);
>>
>> - if (syscall < NR_syscalls)
>> + if (syscall > -1 && syscall < NR_syscalls)
>> syscall_handler(regs, syscall);
>> else
>> regs->a0 = -ENOSYS;
>>
>>
>> ...easier to read?
>>
>>
>> Björn
>
> It seems that your change turn it back to the beginning. If syscall == -1,
> it is supposed to go neither first nor else branch. It should do NOTHING.
> However it was still a great idea. It may be better to use a set of if-statement
> to clarify the logic.
>
> diff --git a/arch/riscv/kernel/traps.c b/arch/riscv/kernel/traps.c
> index f910dfccbf5d2..d0bd725244594 100644
> --- a/arch/riscv/kernel/traps.c
> +++ b/arch/riscv/kernel/traps.c
> @@ -306,7 +306,9 @@ asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)
>
> syscall = syscall_enter_from_user_mode(regs, syscall);
>
> - if (syscall < NR_syscalls)
> + if (syscall == -1)
> + // Do nothing
> + else if (syscall < NR_syscalls)
> syscall_handler(regs, syscall);
> else
> regs->a0 = -ENOSYS;
>

Oh, it's better to reorder to:

diff --git a/arch/riscv/kernel/traps.c b/arch/riscv/kernel/traps.c
index f910dfccbf5d2..932814e34ac37 100644
--- a/arch/riscv/kernel/traps.c
+++ b/arch/riscv/kernel/traps.c
@@ -308,7 +308,7 @@ asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)

if (syscall < NR_syscalls)
syscall_handler(regs, syscall);
- else
+ else if (syscall != -1)
regs->a0 = -ENOSYS;

syscall_exit_to_user_mode(regs);


I will send v3 later.

2023-07-18 21:25:44

by Celeste Liu

Subject: Re: [PATCH v2] riscv: entry: set a0 prior to syscall_enter_from_user_mode

On 2023/7/19 04:50, Björn Töpel wrote:
> Celeste Liu <[email protected]> writes:
>
>> On 2023/7/19 03:35, Björn Töpel wrote:
>>> Celeste Liu <[email protected]> writes:
>>>
>>>> When we test seccomp with 6.4 kernel, we found errno has wrong value.
>>>> If we deny NETLINK_AUDIT with EAFNOSUPPORT, after f0bddf50586d, we will
>>>> get ENOSYS instead. We got same result with 9c2598d43510 ("riscv: entry: Save a0
>>>> prior syscall_enter_from_user_mode()").
>>>>
>>>> After analysing code, we think that regs->a0 = -ENOSYS should be advanced before
>>>> syscall_enter_from_user_mode to fix this problem. In __seccomp_filter, when
>>>> seccomp rejected this syscall with specified errno, they will set a0 to return
>>>> number as syscall ABI, and then return -1. This return number is finally pass
>>>> as return number of syscall_enter_from_user_mode, and then is compared
>>>> with NR_syscalls after converted to ulong (so it will be ULONG_MAX).
>>>> The condition syscall < NR_syscalls will always be false, so regs->a0 = -ENOSYS
>>>> is always executable. It covered a0 set by seccomp, so we always get
>>>> ENOSYS when match seccomp RET_ERRNO rule.
>>>
>>> Isn't something like...
>>>
>>> diff --git a/arch/riscv/kernel/traps.c b/arch/riscv/kernel/traps.c
>>> index f910dfccbf5d..15a8b4898a6c 100644
>>> --- a/arch/riscv/kernel/traps.c
>>> +++ b/arch/riscv/kernel/traps.c
>>> @@ -297,7 +297,7 @@ asmlinkage __visible __trap_section void do_trap_break(struct pt_regs *regs)
>>> asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)
>>> {
>>> if (user_mode(regs)) {
>>> - ulong syscall = regs->a7;
>>> + long syscall = regs->a7;
>>>
>>> regs->epc += 4;
>>> regs->orig_a0 = regs->a0;
>>> @@ -306,7 +306,7 @@ asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)
>>>
>>> syscall = syscall_enter_from_user_mode(regs, syscall);
>>>
>>> - if (syscall < NR_syscalls)
>>> + if (syscall > -1 && syscall < NR_syscalls)
>>> syscall_handler(regs, syscall);
>>> else
>>> regs->a0 = -ENOSYS;
>>>
>>>
>>> ...easier to read?
>>>
>>>
>>> Björn
>>
>> It seems that your change turn it back to the beginning. If syscall == -1,
>> it is supposed to go neither first nor else branch. It should do NOTHING.
>> However it was still a great idea. It may be better to use a set of if-statement
>> to clarify the logic.
>
> Ah, gotcha! (Notice that arch/x86/entry/common.c has
>
> | if (!do_syscall_x64(regs, nr) && !do_syscall_x32(regs, nr) && nr != -1) {
>
> and in do_syscall_x64()
>
> | /*
> | * Convert negative numbers to very high and thus out of range
> | * numbers for comparisons.
> | */
> | unsigned int unr = nr;
>
>
>> diff --git a/arch/riscv/kernel/traps.c b/arch/riscv/kernel/traps.c
>> index f910dfccbf5d2..d0bd725244594 100644
>> --- a/arch/riscv/kernel/traps.c
>> +++ b/arch/riscv/kernel/traps.c
>> @@ -306,7 +306,9 @@ asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)
>>
>> syscall = syscall_enter_from_user_mode(regs, syscall);
>>
>> - if (syscall < NR_syscalls)
>> + if (syscall == -1)
>> + // Do nothing
>> + else if (syscall < NR_syscalls)
>> syscall_handler(regs, syscall);
>> else
>> regs->a0 = -ENOSYS;
>
> Maybe something a bit more explicit?
>
> diff --git a/arch/riscv/kernel/traps.c b/arch/riscv/kernel/traps.c
> index f910dfccbf5d..5cef72874542 100644
> --- a/arch/riscv/kernel/traps.c
> +++ b/arch/riscv/kernel/traps.c
> @@ -297,6 +297,10 @@ asmlinkage __visible __trap_section void do_trap_break(struct pt_regs *regs)
> asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)
> {
> if (user_mode(regs)) {
> + /*
> + * Convert negative numbers to very high and thus out of range
> + * numbers for comparisons.
> + */
> ulong syscall = regs->a7;
>
> regs->epc += 4;
> @@ -308,7 +312,7 @@ asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)
>
> if (syscall < NR_syscalls)
> syscall_handler(regs, syscall);
> - else
> + else if ((long)syscall != -1L)
> regs->a0 = -ENOSYS;
>
> syscall_exit_to_user_mode(regs);
>
>
> Björn

Ok, I will send v3.