From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-qt1-x82d.google.com (mail-qt1-x82d.google.com [IPv6:2607:f8b0:4864:20::82d]) by sourceware.org (Postfix) with ESMTPS id 7517E3857823 for ; Tue, 6 Apr 2021 17:47:42 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.3.2 sourceware.org 7517E3857823 Received: by mail-qt1-x82d.google.com with SMTP id g24so11796927qts.6 for ; Tue, 06 Apr 2021 10:47:42 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:subject:to:cc:references:from:message-id:date :user-agent:mime-version:in-reply-to:content-language :content-transfer-encoding; bh=quMxPm634fPbrQ+FQbvZ5ev03Vf3uppxKTCn+MfziXU=; b=Miz5uIyAnRYct52TjX1fSzp/cUMi6noptTxBI3TBFEfIH0f1PxqNxp4874Hiol2TfO KKvUzOXrXsTt91jMRnBY8qvrrhMj3n7rbXwY1YecjbvtygiGmjgrEowXtFMiOR3VX/KD aH0hJuoQKwDOpDVVSN/JSdXVwJBuDhQUYZOfUr34b5tTtvHJT78P/rxeAAmoi47NObyO HSqJrXS82XJ1HZVCvBIc7KAALosHQ8bRxyog65HhpAnQoPXH+k17mg1QvlLVsUgdGz2S uySkTN1H2fq4/reLavB27VyT3QSciJqlXJC227AHg3GLui9KF66VKIQ1kFBRpPFHmYZd Gy3A== X-Gm-Message-State: AOAM532qOGn4hCqW5fvhfXiF7gT47F7K5R5gIfrbIV6UK8ko0+aJYUS4 PSfhF4XPYlE0jDHX95UcFHJ/WPISvjc77at3 X-Google-Smtp-Source: ABdhPJz87th4A27FgqO7jvTnFrgvDzID8Ij9CN4BlDSk6yKuMdmfUw/A4uhE9P8WoQBlTYSi15AlPA== X-Received: by 2002:a05:622a:48d:: with SMTP id p13mr28632449qtx.21.1617731261806; Tue, 06 Apr 2021 10:47:41 -0700 (PDT) Received: from [192.168.1.132] ([177.194.41.149]) by smtp.gmail.com with ESMTPSA id z188sm16636930qkb.40.2021.04.06.10.47.40 (version=TLS1_3 cipher=TLS_AES_128_GCM_SHA256 bits=128/128); Tue, 06 Apr 2021 10:47:41 -0700 (PDT) Subject: Re: [PATCH 06/15] elf: Fix comments and logic in _dl_add_to_slotinfo To: Szabolcs Nagy Cc: libc-alpha@sourceware.org References: <068770faf123b7c227f5f1e130812f7976e74cef.1613390045.git.szabolcs.nagy@arm.com> <20210406154828.GM23289@arm.com> From: Adhemerval Zanella Message-ID: Date: Tue, 6 Apr 2021 14:47:38 -0300 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101 Thunderbird/78.7.1 MIME-Version: 1.0 In-Reply-To: <20210406154828.GM23289@arm.com> Content-Type: text/plain; charset=utf-8 Content-Language: en-US Content-Transfer-Encoding: 7bit X-Spam-Status: No, score=-12.6 required=5.0 tests=BAYES_00, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, DKIM_VALID_EF, GIT_PATCH_0, NICE_REPLY_A, RCVD_IN_DNSWL_NONE, SPF_HELO_NONE, SPF_PASS, TXREP autolearn=ham autolearn_force=no version=3.4.2 X-Spam-Checker-Version: SpamAssassin 3.4.2 (2018-09-13) on server2.sourceware.org X-BeenThere: libc-alpha@sourceware.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Libc-alpha mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 06 Apr 2021 17:47:43 -0000 On 06/04/2021 12:48, Szabolcs Nagy wrote: > The 04/02/2021 17:50, Adhemerval Zanella via Libc-alpha wrote: >> On 15/02/2021 08:59, Szabolcs Nagy via Libc-alpha wrote: >>> From: Szabolcs Nagy >>> >>> Since >>> >>> commit a509eb117fac1d764b15eba64993f4bdb63d7f3c >>> Avoid late dlopen failure due to scope, TLS slotinfo updates [BZ #25112] >>> >>> the generation counter update is not needed in the failure path. >> >> It is not clear to me from just the commit reference why it would >> be safe to remove the GL(dl_tls_generation) update on >> _dl_add_to_slotinfo. >> >> The dl_open_worker calls update_tls_slotinfo which in turn call >> might call _dl_add_to_slotinfo *after* the demarcation point. Will >> it terminate the process? > > in that commit the logic got changed such that allocations > happen before the demarcation point in resize_tls_slotinfo. > > this is the reason for the do_add bool argument in > _dl_add_to_slotinfo: it's called twice and the first call > with do_add==false is only there to ensure everything is > allocated before the demarcation point (so module loading > can be reverted, no need to bump the generation count). > > i guess this is not visible by just looking at the > _dl_add_to_slotinfo code. Right, so if I reading correctly once _dl_add_to_slotinfo (..., true) is called by update_tls_slotinfo, the malloc that create a new dtv_slotinfo_list won't be called (since it was already allocated previously) since the entry was already pre-allocated and thus the search part at line 978-987 will find. Is that correct? > > note that adding some asserts to ensure there is no allocation > when do_add==true does not work: rtld uses the same api, but > without the do_add==false preallocation step since at startup > time allocation failure is fatal anyway. Right, the _dl_signal_error will trigger a fatal_error since lcatch won't be override yet. Thanks for the explanation. > >>> --- >>> elf/dl-tls.c | 11 +---------- >>> 1 file changed, 1 insertion(+), 10 deletions(-) >>> >>> diff --git a/elf/dl-tls.c b/elf/dl-tls.c >>> index 79b93ad91b..24d00c14ef 100644 >>> --- a/elf/dl-tls.c >>> +++ b/elf/dl-tls.c >>> @@ -998,16 +998,7 @@ _dl_add_to_slotinfo (struct link_map *l, bool do_add) >>> + TLS_SLOTINFO_SURPLUS * sizeof (struct dtv_slotinfo)); >>> if (listp == NULL) >>> { >>> - /* We ran out of memory. We will simply fail this >>> - call but don't undo anything we did so far. The >>> - application will crash or be terminated anyway very >>> - soon. */ >>> - >>> - /* We have to do this since some entries in the dtv >>> - slotinfo array might already point to this >>> - generation. */ >>> - ++GL(dl_tls_generation); >>> - >>> + /* We ran out of memory while resizing the dtv slotinfo list. */ >>> _dl_signal_error (ENOMEM, "dlopen", NULL, N_("\ >>> cannot create TLS data structures")); >>> }