From: Vladislav Shpilevoy <v.shpilevoy@tarantool.org> To: Cyrill Gorcunov <gorcunov@gmail.com>, tml <tarantool-patches@dev.tarantool.org> Subject: Re: [Tarantool-patches] [PATCH v7 1/2] fiber: set diagnostics at madvise/mprotect failure Date: Fri, 14 Feb 2020 00:26:04 +0100 [thread overview] Message-ID: <46c82f5e-5e4c-ec4a-b8ce-b999e6a6230c@tarantool.org> (raw) In-Reply-To: <20200213205618.7982-2-gorcunov@gmail.com> Thanks for the fixes! > @@ -1007,6 +1035,8 @@ fiber_stack_watermark_create(struct fiber *fiber) > static void > fiber_stack_destroy(struct fiber *fiber, struct slab_cache *slabc) > { > + static const int mprotect_flags = PROT_READ | PROT_WRITE; > + Why is it static? From what I know, when it is static we don't have a guarantee, that it won't occupy memory in the .data section. Even though here it is clearly not necessary. Wouldn't just const be enough? Why was not it possible to leave these flags inlined in fiber_mprotect() call? They are not used anywhere else. PROT_NONE, for example, is left inlined. > if (fiber->stack != NULL) { > VALGRIND_STACK_DEREGISTER(fiber->stack_id); > #if ENABLE_ASAN > @@ -1017,7 +1047,22 @@ fiber_stack_destroy(struct fiber *fiber, struct slab_cache *slabc) > guard = page_align_down(fiber->stack - page_size); > else > guard = page_align_up(fiber->stack + fiber->stack_size); > - fiber_mprotect(guard, page_size, PROT_READ | PROT_WRITE); > + > + if (fiber_mprotect(guard, page_size, mprotect_flags) != 0) { > + /* > + * FIXME: We need some intelligent handling: > + * say put this slab into a queue and retry > + * to setup the original protection back in > + * background. > + * > + * Note that in case if we're called from > + * fiber_stack_create() the @a mprotect_flags is > + * the same as the slab been created with, so > + * calling mprotect for VMA with same flags > + * won't fail. > + */ > + diag_log(); > + } > slab_put(slabc, fiber->stack_slab); > } > } > @@ -1064,6 +1109,11 @@ fiber_stack_create(struct fiber *fiber, struct slab_cache *slabc, > fiber->stack_size); > > if (fiber_mprotect(guard, page_size, PROT_NONE)) { > + /* > + * Write an error into the log since a guard > + * page is critical for functionality. > + */ > + diag_log(); > fiber_stack_destroy(fiber, slabc); > return -1; > }
next prev parent reply other threads:[~2020-02-13 23:26 UTC|newest] Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top 2020-02-13 20:56 [Tarantool-patches] [PATCH v7 0/2] fiber: Handle stack madvise/mprotect errors Cyrill Gorcunov 2020-02-13 20:56 ` [Tarantool-patches] [PATCH v7 1/2] fiber: set diagnostics at madvise/mprotect failure Cyrill Gorcunov 2020-02-13 23:26 ` Vladislav Shpilevoy [this message] 2020-02-14 7:54 ` Cyrill Gorcunov 2020-02-14 22:27 ` Vladislav Shpilevoy 2020-02-15 6:57 ` Cyrill Gorcunov 2020-02-15 15:41 ` Vladislav Shpilevoy 2020-02-15 17:55 ` Cyrill Gorcunov 2020-02-13 20:56 ` [Tarantool-patches] [PATCH v7 2/2] fiber: leak slab if unable to bring prots back Cyrill Gorcunov 2020-02-13 23:26 ` Vladislav Shpilevoy 2020-02-14 8:25 ` Cyrill Gorcunov 2020-02-13 20:57 ` [Tarantool-patches] [PATCH v7 0/2] fiber: Handle stack madvise/mprotect errors Cyrill Gorcunov 2020-02-13 23:26 ` Vladislav Shpilevoy 2020-02-14 7:07 ` Cyrill Gorcunov
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=46c82f5e-5e4c-ec4a-b8ce-b999e6a6230c@tarantool.org \ --to=v.shpilevoy@tarantool.org \ --cc=gorcunov@gmail.com \ --cc=tarantool-patches@dev.tarantool.org \ --subject='Re: [Tarantool-patches] [PATCH v7 1/2] fiber: set diagnostics at madvise/mprotect failure' \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: link
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox