This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] powerpc: Only enable TLE with PPC_FEATURE2_HTM_NOSC
- From: Breno Leitao <leitao at debian dot org>
- To: Adhemerval Zanella <adhemerval dot zanella at linaro dot org>, Tulio Magno Quites Machado Filho <tuliom at ascii dot art dot br>, libc-alpha at sourceware dot org
- Cc: Gustavo Romero <gromero at linux dot vnet dot ibm dot com>
- Date: Thu, 13 Sep 2018 15:30:16 -0300
- Subject: Re: [PATCH] powerpc: Only enable TLE with PPC_FEATURE2_HTM_NOSC
- References: <1535403247-27306-1-git-send-email-adhemerval.zanella@linaro.org> <87mut7bfb9.fsf@linux.ibm.com> <e422bfc1-80dd-c56e-3867-8a3180bc8d1e@linaro.org>
Hi Adhemerval,
On 08/28/2018 08:48 AM, Adhemerval Zanella wrote:
>
>
> On 27/08/2018 19:44, Tulio Magno Quites Machado Filho wrote:
>> Adhemerval Zanella <adhemerval.zanella@linaro.org> writes:
<snip>
>> Can't ABORT_TRANSACTION check for PPC_FEATURE2_HTM_NOSC in tcbhead_t.hwcap and
>> continue to support older Linux versions?
>>
>
> I don't have a strong opinion here, but chatting with Breno I had the impression
> that TLE (and HTM in general) is only used on kernel with this recent fix. This
> patch is really a hack insertion that penalizes all syscall for an specific
> feature which may not be present in the kernel.
In fact, it is more than a penalization (which is also there). The current
design seems to be incoherent with the Kernel design[1] and limiting the
feature usage. By the kernel design, a syscall executed inside a suspended
transaction is allowed, as said by the kernel documentation[1]:
"Syscalls made from within a suspended transaction are performed as normal and
the transaction is not explicitly doomed by the kernel"
The statement above becomes invalid if the syscall is performed through glibc
nowadays, which I understand is an undesired behavior.
[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/powerpc/transactional_memory.txt#n81