From patchwork Thu Dec 5 15:20:25 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Florian Weimer X-Patchwork-Id: 1204644 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Authentication-Results: ozlabs.org; spf=pass (sender SPF authorized) smtp.mailfrom=sourceware.org (client-ip=209.132.180.131; helo=sourceware.org; envelope-from=libc-alpha-return-107745-incoming=patchwork.ozlabs.org@sourceware.org; receiver=) Authentication-Results: ozlabs.org; dmarc=pass (p=none dis=none) header.from=redhat.com Authentication-Results: ozlabs.org; dkim=pass (1024-bit key; secure) header.d=sourceware.org header.i=@sourceware.org header.b="iE90wDa2"; dkim=pass (1024-bit key; unprotected) header.d=redhat.com header.i=@redhat.com header.b="Rw7UWhLy"; dkim-atps=neutral Received: from sourceware.org (server1.sourceware.org [209.132.180.131]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ozlabs.org (Postfix) with ESMTPS id 47TKD62qWzz9sP3 for ; Fri, 6 Dec 2019 02:20:38 +1100 (AEDT) DomainKey-Signature: a=rsa-sha1; c=nofws; d=sourceware.org; h=list-id :list-unsubscribe:list-subscribe:list-archive:list-post :list-help:sender:from:to:subject:in-reply-to:references :message-id:date:mime-version:content-type :content-transfer-encoding; q=dns; s=default; b=B7ml7ywYfoU32XhT pYijkTdK9lxzBdfwEoRsQLQlnhluBnImOm8yMjuULCsGFDSesa41MBQDWnMRtesx jeerS/F/cdlPgpWdPFr5Xuo8z3FZeszzimXCToT+5y7BT/TP+5cYfchveAS+X456 dkOupSmAmWffeYVT9xhRC6Ft5BQ= DKIM-Signature: v=1; a=rsa-sha1; c=relaxed; d=sourceware.org; h=list-id :list-unsubscribe:list-subscribe:list-archive:list-post :list-help:sender:from:to:subject:in-reply-to:references :message-id:date:mime-version:content-type :content-transfer-encoding; s=default; bh=qRkaaPjwUQ2ZZS0iCMyad1 muCWI=; b=iE90wDa26sJ2W/1TSnd05ddEPcpNajmGtwuSpbxw2yeyqa9oX8TWyp vGXn40vV67wLw9KVrh+hFsjSf3WH2bhZnidqCoRcFBpyV1Qar6+RVSdLAI9Wqymd 6w2n91Y4NyFlvnRBqnQRr5dNhap8YmkVUo/PpzCEcj+rhWh0+c+IY= Received: (qmail 33529 invoked by alias); 5 Dec 2019 15:20:32 -0000 Mailing-List: contact libc-alpha-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Unsubscribe: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: libc-alpha-owner@sourceware.org Delivered-To: mailing list libc-alpha@sourceware.org Received: (qmail 33520 invoked by uid 89); 5 Dec 2019 15:20:32 -0000 Authentication-Results: sourceware.org; auth=none X-Spam-SWARE-Status: No, score=-18.5 required=5.0 tests=AWL, BAYES_00, GIT_PATCH_0, GIT_PATCH_1, GIT_PATCH_2, GIT_PATCH_3 autolearn=ham version=3.3.1 spammy= X-HELO: us-smtp-1.mimecast.com DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=mimecast20190719; t=1575559229; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:mime-version:mime-version:content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references; bh=9bCW045r+wWJbdKj6CDSVyqP4TJ7tNeKbTTzlZt+iyg=; b=Rw7UWhLy2iqyXw1xADh1e7t9UPg2Wor71ci5mUtQ/dOt6ZHc49DSqTRGw9TBsTqZiqMeYa h0/bfVRYJCpa0A9+EnJQZsizZxxLlqI832ePfKoL+SCemTJ/mTaK9e8B5rrJPmz4RUSBUZ jKOwxB1gaFUN8eSY36Puwr0slLGZDD8= From: Florian Weimer To: libc-alpha@sourceware.org Subject: [PATCH 2/2] dlopen: Do not block signals In-Reply-To: References: Message-Id: <09f5540ab05de0a14992d8071f6feb0137775156.1575558989.git.fweimer@redhat.com> Date: Thu, 05 Dec 2019 16:20:25 +0100 User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/26.2 (gnu/linux) MIME-Version: 1.0 X-Mimecast-Spam-Score: 0 Blocking signals causes issues with certain anti-malware solutions which rely on an unblocked SIGSYS signal for system calls they intercept. This reverts commit a2e8aa0d9ea648068d8be52dd7b15f1b6a008e23 ("Block signals during the initial part of dlopen") and adds comments related to async signal safety to active_nodelete and its caller. Note that this does not make lazy binding async-signal-safe with regards to dlopen. It merely avoids introducing new async-signal-safety hazards as part of the NODELETE changes. Reviewed-by: Adhemerval Zanella Reviewed-by: Carlos O'Donell --- elf/dl-open.c | 37 +++++++++++-------------------------- 1 file changed, 11 insertions(+), 26 deletions(-) diff --git a/elf/dl-open.c b/elf/dl-open.c index c23341be58..5a1c5b5326 100644 --- a/elf/dl-open.c +++ b/elf/dl-open.c @@ -34,7 +34,6 @@ #include #include #include -#include #include #include @@ -53,10 +52,6 @@ struct dl_open_args /* Namespace ID. */ Lmid_t nsid; - /* Original signal mask. Used for unblocking signal handlers before - running ELF constructors. */ - sigset_t original_signal_mask; - /* Original value of _ns_global_scope_pending_adds. Set by dl_open_worker. Only valid if nsid is a real namespace (non-negative). */ @@ -446,6 +441,9 @@ activate_nodelete (struct link_map *new) _dl_debug_printf ("activating NODELETE for %s [%lu]\n", l->l_name, l->l_ns); + /* The flag can already be true at this point, e.g. a signal + handler may have triggered lazy binding and set NODELETE + status immediately. */ l->l_nodelete_active = true; /* This is just a debugging aid, to indicate that @@ -520,16 +518,12 @@ dl_open_worker (void *a) if (new == NULL) { assert (mode & RTLD_NOLOAD); - __libc_signal_restore_set (&args->original_signal_mask); return; } if (__glibc_unlikely (mode & __RTLD_SPROF)) - { - /* This happens only if we load a DSO for 'sprof'. */ - __libc_signal_restore_set (&args->original_signal_mask); - return; - } + /* This happens only if we load a DSO for 'sprof'. */ + return; /* This object is directly loaded. */ ++new->l_direct_opencount; @@ -565,7 +559,6 @@ dl_open_worker (void *a) assert (_dl_debug_initialize (0, args->nsid)->r_state == RT_CONSISTENT); - __libc_signal_restore_set (&args->original_signal_mask); return; } @@ -712,6 +705,12 @@ dl_open_worker (void *a) All memory allocations for new objects must have happened before. */ + /* Finalize the NODELETE status first. This comes before + update_scopes, so that lazy binding will not see pending NODELETE + state for newly loaded objects. There is a compiler barrier in + update_scopes which ensures that the changes from + activate_nodelete are visible before new objects show up in the + local scope. */ activate_nodelete (new); /* Second stage after resize_scopes: Actually perform the scope @@ -745,10 +744,6 @@ dl_open_worker (void *a) if (mode & RTLD_GLOBAL) add_to_global_resize (new); - /* Unblock signals. Data structures are now consistent, and - application code may run. */ - __libc_signal_restore_set (&args->original_signal_mask); - /* Run the initializer functions of new objects. Temporarily disable the exception handler, so that lazy binding failures are fatal. */ @@ -838,10 +833,6 @@ no more namespaces available for dlmopen()")); args.argv = argv; args.env = env; - /* Recursive lazy binding during manipulation of the dynamic loader - structures may result in incorrect behavior. */ - __libc_signal_block_all (&args.original_signal_mask); - struct dl_exception exception; int errcode = _dl_catch_exception (&exception, dl_open_worker, &args); @@ -882,16 +873,10 @@ no more namespaces available for dlmopen()")); _dl_close_worker (args.map, true); - /* Restore the signal mask. In the success case, this - happens inside dl_open_worker. */ - __libc_signal_restore_set (&args.original_signal_mask); - /* All l_nodelete_pending objects should have been deleted at this point, which is why it is not necessary to reset the flag here. */ } - else - __libc_signal_restore_set (&args.original_signal_mask); assert (_dl_debug_initialize (0, args.nsid)->r_state == RT_CONSISTENT);