linux-kernel - Re: [PATCH v1] ARM: A9: Add ARM ERRATA 764319 workaround

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <9da605e9-2cbd-6d8f-7425-fdfe2e744d3a@arm.com>
Date:   Tue, 10 May 2022 10:47:03 +0100
From:   Robin Murphy <robin.murphy@....com>
To:     Arnd Bergmann <arnd@...db.de>,
        "Hawkins, Nick" <nick.hawkins@....com>
Cc:     "Verdun, Jean-Marie" <verdun@....com>,
        Russell King - ARM Linux <linux@...linux.org.uk>,
        Will Deacon <will@...nel.org>,
        Mark Rutland <mark.rutland@....com>,
        Linux ARM <linux-arm-kernel@...ts.infradead.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Joel Stanley <joel@....id.au>
Subject: Re: [PATCH v1] ARM: A9: Add ARM ERRATA 764319 workaround

On 2022-05-10 08:19, Arnd Bergmann wrote:
> On Fri, May 6, 2022 at 9:29 PM <nick.hawkins@....com> wrote:
>>
>> From: Nick Hawkins <nick.hawkins@....com>
>>
>> Enable the workaround for the 764319 Cortex A-9 erratum.
>> CP14 read accesses to the DBGPRSR and DBGOSLSR registers generate an
>> unexpected Undefined Instruction exception when the DBGSWENABLE external
>> pin is set to 0, even when the CP14 accesses are performed from a
>> privileged mode. The work around catches the exception in a way
>> the kernel does not stop execution with the use of undef_hook. This
>> has been found to effect the HPE GXP SoC.
>>
>> Signed-off-by: Nick Hawkins <nick.hawkins@....com>
>> ---
>>   arch/arm/Kconfig                | 11 +++++++++++
>>   arch/arm/kernel/hw_breakpoint.c | 26 ++++++++++++++++++++++++++
>>   2 files changed, 37 insertions(+)
>>
>> diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
>> index 13f77eec7c40..6944adfb0fae 100644
>> --- a/arch/arm/Kconfig
>> +++ b/arch/arm/Kconfig
>> @@ -974,6 +974,17 @@ config ARM_ERRATA_764369
>>            relevant cache maintenance functions and sets a specific bit
>>            in the diagnostic control register of the SCU.
>>
>> +config ARM_ERRATA_764319
>> +       bool "ARM errata: Read to DBGPRSR and DBGOSLSR may generate Undefined instruction"
>> +       depends on CPU_V7
>> +       help
>> +         This option enables the workaround for the 764319 Cortex A-9 erratum.
>> +         CP14 read accesses to the DBGPRSR and DBGOSLSR registers generate an
>> +         unexpected Undefined Instruction exception when the DBGSWENABLE
>> +         external pin is set to 0, even when the CP14 accesses are performed
>> +         from a privileged mode. This work around catches the exception in a
>> +         way the kernel does not stop execution.
>> +
>>   config ARM_ERRATA_775420
>>          bool "ARM errata: A data cache maintenance operation which aborts, might lead to deadlock"
>>          depends on CPU_V7
>> diff --git a/arch/arm/kernel/hw_breakpoint.c b/arch/arm/kernel/hw_breakpoint.c
>> index b1423fb130ea..c41a8436a796 100644
>> --- a/arch/arm/kernel/hw_breakpoint.c
>> +++ b/arch/arm/kernel/hw_breakpoint.c
>> @@ -941,6 +941,23 @@ static int hw_breakpoint_pending(unsigned long addr, unsigned int fsr,
>>          return ret;
>>   }
>>
>> +#ifdef CONFIG_ARM_ERRATA_764319
>> +int oslsr_fault;
>> +
>> +static int debug_oslsr_trap(struct pt_regs *regs, unsigned int instr)
>> +{
>> +       oslsr_fault = 1;
>> +       instruction_pointer(regs) += 4;
>> +       return 0;
>> +}
>> +
>> +static struct undef_hook debug_oslsr_hook = {
>> +       .instr_mask  = 0xffffffff,
>> +       .instr_val = 0xee115e91,
>> +       .fn = debug_oslsr_trap,
>> +};
>> +#endif
>> +
> 
> Hi Nick,
> 
> This seems a bit more complex than necessary. Can't you just use a custom
> inline asm with an ex_table entry to catch the fault? Have a look at
> __get_user_asm() for an example.

Indeed, according to the Cortex-A9 documentation the register should 
always read as zero anyway, so all we should need to do is initialise 
the local oslsr variable to 0 and have the fault handler just return to 
the next instruction.

Robin.