From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mp0 ([2001:41d0:2:4a6f::]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits)) by ms11 with LMTPS id cPMFAE94L1/LJwAA0tVLHw (envelope-from ) for ; Sun, 09 Aug 2020 04:15:11 +0000 Received: from aspmx1.migadu.com ([2001:41d0:2:4a6f::]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits)) by mp0 with LMTPS id UNF+N054L1+SWwAA1q6Kng (envelope-from ) for ; Sun, 09 Aug 2020 04:15:10 +0000 Received: from lists.gnu.org (lists.gnu.org [209.51.188.17]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by aspmx1.migadu.com (Postfix) with ESMTPS id 700459403C9 for ; Sun, 9 Aug 2020 04:15:09 +0000 (UTC) Received: from localhost ([::1]:33192 helo=lists1p.gnu.org) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1k4cjj-00015I-Ix for larch@yhetil.org; Sun, 09 Aug 2020 00:15:07 -0400 Received: from eggs.gnu.org ([2001:470:142:3::10]:59480) by lists.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1k4cjf-00014b-5J for guix-patches@gnu.org; Sun, 09 Aug 2020 00:15:03 -0400 Received: from debbugs.gnu.org ([209.51.188.43]:48282) by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_128_GCM_SHA256:128) (Exim 4.90_1) (envelope-from ) id 1k4cje-00052k-Se for guix-patches@gnu.org; Sun, 09 Aug 2020 00:15:02 -0400 Received: from Debian-debbugs by debbugs.gnu.org with local (Exim 4.84_2) (envelope-from ) id 1k4cje-000087-I0 for guix-patches@gnu.org; Sun, 09 Aug 2020 00:15:02 -0400 X-Loop: help-debbugs@gnu.org Subject: [bug#42023] [PATCH] database: register-items: reduce transaction scope. Resent-From: Caleb Ristvedt Original-Sender: "Debbugs-submit" Resent-CC: guix-patches@gnu.org Resent-Date: Sun, 09 Aug 2020 04:15:02 +0000 Resent-Message-ID: Resent-Sender: help-debbugs@gnu.org X-GNU-PR-Message: followup 42023 X-GNU-PR-Package: guix-patches X-GNU-PR-Keywords: patch To: Ludovic =?UTF-8?Q?Court=C3=A8s?= Cc: 42023@debbugs.gnu.org, Christopher Baines Received: via spool by 42023-submit@debbugs.gnu.org id=B42023.1596946463436 (code B ref 42023); Sun, 09 Aug 2020 04:15:02 +0000 Received: (at 42023) by debbugs.gnu.org; 9 Aug 2020 04:14:23 +0000 Received: from localhost ([127.0.0.1]:59828 helo=debbugs.gnu.org) by debbugs.gnu.org with esmtp (Exim 4.84_2) (envelope-from ) id 1k4cj0-00006x-KT for submit@debbugs.gnu.org; Sun, 09 Aug 2020 00:14:23 -0400 Received: from mail-il1-f178.google.com ([209.85.166.178]:40404) by debbugs.gnu.org with esmtp (Exim 4.84_2) (envelope-from ) id 1k4civ-00006g-Uw for 42023@debbugs.gnu.org; Sun, 09 Aug 2020 00:14:20 -0400 Received: by mail-il1-f178.google.com with SMTP id x1so4980577ilp.7 for <42023@debbugs.gnu.org>; Sat, 08 Aug 2020 21:14:17 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cune-org.20150623.gappssmtp.com; s=20150623; h=from:to:cc:subject:references:date:in-reply-to:message-id :user-agent:mime-version; bh=u4Ih1rN7QVYP7F1CB2xXuvtuPRWLWtBm8PnxjEMBCrs=; b=ke+aKyIV0amP+RYey8a/OxDllry3fNTSwukeENeOKdnzTE+CfAES5stmjI9GoZMpDa oDSHcBmKCCJlYRcDC3tBoOosJeodb7l5LaEqlyw6noIx3PvQJdde9TzC8/B+unSiCrCl vHF5EPok/scAN+Pvnr/UIek6o8xN0khoaP6teIBSQAWLcvXhCEOKKt62BTKujWPCcoJJ D0fDSZZyywwXJg96GYpxDD8cXz/SrQLGAhO3lHugPx2Cy6cjc9zRbiRi5V8DEvwtUQZo Q6dE8GxVvv5X47LWDOAqKl1IHhw9CyLHboZZJWv8rANX8Qpfv34FDM1MJuFWocyLhgDZ 8hYA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:references:date:in-reply-to :message-id:user-agent:mime-version; bh=u4Ih1rN7QVYP7F1CB2xXuvtuPRWLWtBm8PnxjEMBCrs=; b=scIT7gYr8jsXx000oqRHSUc6D02yvLcnO/JjC7WL9qNR7Rkct3RFa/chIw95EiLhQA 17VNvSq1HrEFa6epYP+l2JP+pXHYemKZmDuQbaBLD7R+M08aLiZC/JQUMHxS8NYbjtRe FwH/7CnnoQWAQPuy3Yr0fK5J0sTA/urszaE4cfUfFwyV7MHTGa9hld+aCdWKnR4eFnXD afgUEVKSGKQAMqu+nKgoDgjr36VugGPrOgFk1uDqla7ioY6F4fIf9Gt7EeSzYBOg8Nk7 on0lqe0aAchfFEwpznzwD08VKQp/ky5wVir98E0zI5CnM4bZMDy52XIHDdEPsaKbhdCU biyQ== X-Gm-Message-State: AOAM531isYPWsYQ3+XPCLnSzHs4jgNUWWKj5OvbdT7+ZRh184KHuaBv6 i/9iCwH+yqtMWHZYK5fiJ3j0eQ== X-Google-Smtp-Source: ABdhPJzL5ypgU/kjrV5J0rCnettKK+eJFKCXMzbfQAFedH3ySrRQD5G4nHY4dt3gUIiLJDLBKB6avA== X-Received: by 2002:a92:d289:: with SMTP id p9mr12442930ilp.100.1596946452061; Sat, 08 Aug 2020 21:14:12 -0700 (PDT) Received: from GuixPotato ([208.89.170.4]) by smtp.gmail.com with ESMTPSA id 25sm9762150ilv.85.2020.08.08.21.14.10 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Sat, 08 Aug 2020 21:14:11 -0700 (PDT) From: Caleb Ristvedt References: <20200623163649.32444-1-mail@cbaines.net> <87366lxwcd.fsf@gnu.org> <87v9jg4aiz.fsf@cune.org> <87eeq3vbat.fsf@gnu.org> <87tuxu2sz6.fsf@gnu.org> Date: Sat, 08 Aug 2020 23:13:35 -0500 In-Reply-To: <87tuxu2sz6.fsf@gnu.org> ("Ludovic \=\?utf-8\?Q\?Court\=C3\=A8s\=22'\?\= \=\?utf-8\?Q\?s\?\= message of "Sun, 26 Jul 2020 17:31:25 +0200") Message-ID: <87bljka234.fsf@cune.org> User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/26.3 (gnu/linux) MIME-Version: 1.0 Content-Type: multipart/signed; boundary="==-=-="; micalg=pgp-sha256; protocol="application/pgp-signature" X-Spam-Score: -0.0 (/) X-BeenThere: debbugs-submit@debbugs.gnu.org X-Mailman-Version: 2.1.18 Precedence: list X-Spam-Score: -1.0 (-) X-BeenThere: guix-patches@gnu.org List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: guix-patches-bounces+larch=yhetil.org@gnu.org Sender: "Guix-patches" X-Scanner: scn0 Authentication-Results: aspmx1.migadu.com; dkim=fail (rsa verify failed) header.d=cune-org.20150623.gappssmtp.com header.s=20150623 header.b=ke+aKyIV; dmarc=none; spf=pass (aspmx1.migadu.com: domain of guix-patches-bounces@gnu.org designates 209.51.188.17 as permitted sender) smtp.mailfrom=guix-patches-bounces@gnu.org X-Spam-Score: -2.11 X-TUID: PV+2A8sotAUj --==-=-= Content-Type: multipart/mixed; boundary="=-=-=" --=-=-= Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable Ludovic Court=C3=A8s writes: > Hi reepca, > > Did you have time to look into this (see below)? I still see a lot of > contention on /var/guix/db/db.sqlite-shm on berlin and I feel like these > changes would be welcome. :-) Apologies for the long delay, I think I'm caught up on this issue now. Patch series detailed below. > > Thanks, > Ludo=E2=80=99. >>> So general "fiddling" can't happen, but GC can. The responsibility for >>> both locking and registering as temporary roots falls on the caller, >>> currently, as I believe it should. We may want to document this >>> responsibility in the docstring for register-items, though. >> >> Yes, it would be good to add a sentence to document it. Patch attached. >> >>> So currently finalize-store-file is safe from clobbering by another >>> process attempting to finalize to the same path, but not safe from >>> garbage collection between when the temporary store file is renamed and >>> when it is registered. It needs an add-temp-root prior to renaming. >> >> Ah, so we=E2=80=99d need to do that before applying the patch that reduc= es the >> scope of the transaction, right? AFAIK the only code path that actually uses finalize-store-file currently is from the build hook / offload thingy, and it turns out that the outputs of derivations should already be registered as temp roots by the daemon process that launched the offload process (specifically registered in haveDerivation() in nix/libstore/build.cc). So this is technically not currently necessary before or after the scope-reducing patch. But it makes sense to guarantee that the register-items invocation will only happen on items that are protected from garbage collection, so I've put a patch, right before the documenting of that requirement, that modifies finalize-store-file to always add the file being finalized as a temp root for the extent of the register-items invocation. >>> Also, replace-with-link doesn't check whether the "containing directory" >>> is the store like optimisePath_() does, so in theory if another process >>> tried to make a permanent change to the store's permissions it could be >>> clobbered when replace-with-link restores the permissions. I don't know >>> of any instance where this could happen currently, but it's something to >>> keep in mind. >> >> I guess we should also avoid changing permissions on /gnu/store, it >> would be wiser. While rebasing my changes I noticed that the current implementation of this uses (%store-directory) from (guix build utils), which may not correspond to the #:store keyword argument of 'deduplicate'. So I added a patch that propagates the store through to replace-with-link and from there to with-writable-file. >>> diff --git a/guix/store/deduplication.scm b/guix/store/deduplication.scm >>> index 6784ee0b92..b6d94e49c2 100644 >>> --- a/guix/store/deduplication.scm >>> +++ b/guix/store/deduplication.scm >>> @@ -161,26 +161,44 @@ under STORE." >>> (scandir* path)) >>> (let ((link-file (string-append links-directory "/" >>> (bytevector->nix-base32-string= hash)))) >>> - (if (file-exists? link-file) >>> - (replace-with-link link-file path >>> - #:swap-directory links-directory) >>> - (catch 'system-error >>> - (lambda () >>> - (link path link-file)) >>> - (lambda args >>> - (let ((errno (system-error-errno args))) >>> - (cond ((=3D errno EEXIST) >>> - ;; Someone else put an entry for PATH in >>> - ;; LINKS-DIRECTORY before we could. Let's = use it. >>> - (replace-with-link path link-file >>> - #:swap-directory links-d= irectory)) >>> - ((=3D errno ENOSPC) >>> - ;; There's not enough room in the directory= index for >>> - ;; more entries in .links, but that's fine:= we can >>> - ;; just stop. >>> - #f) >>> - ((=3D errno EMLINK) >>> - ;; PATH has reached the maximum number of l= inks, but >>> - ;; that's OK: we just can't deduplicate it = more. >>> - #f) >>> - (else (apply throw args))))))))))) >>> + (let retry () >>> + (if (file-exists? link-file) >>> + (catch 'system-error >>> + (lambda () >>> + (replace-with-link link-file path >>> + #:swap-directory links-director= y)) >>> + (lambda args >>> + (if (and (=3D (system-error-errno args) ENOENT) >>> + ;; ENOENT can be produced because: >>> + ;; - LINK-FILE has missing directory comp= onents >>> + ;; - LINKS-DIRECTORY has missing directory >>> + ;; components >>> + ;; - PATH has missing directory components >>> + ;; >>> + ;; the last two are errors, the first just >>> + ;; requires a retry. >>> + (file-exists? (dirname path)) >>> + (file-exists? links-directory)) >>> + (retry) >>> + (apply throw args)))) >> >> I feel like there are TOCTTOU issues here and redundant =E2=80=98stat=E2= =80=99 calls, >> plus the risk of catching a =E2=80=98system-error=E2=80=99 coming from = =E2=80=9Csomewhere else.=E2=80=9D >> >> What about baking this logic directly in =E2=80=98replace-with-link=E2= =80=99, and >> replacing =E2=80=98file-exists?=E2=80=99 calls by 'system-error handling? Patch attached. I've renamed replace-with-link to canonicalize-with-link since it may now have to create the target link. Unfortunately there are places where 'file-exists?' error handling is necessary, simply because of ambiguity in errnos. For example, link(oldpath, newpath) can return ENOENT, but there's no way of knowing from that alone whether this is because oldpath has missing directories or newpath has missing directories or the directory components of oldpath are all present but the file itself is missing (the case we need to be able to recognize and retry in case of). Also, I tried removing the first check for whether the canonical link exists in favor of error handling like you suggested, but this actually messes up the hard-coded number of link-invocations expected in tests/store-deduplication.scm - it expects a single link invocation when the canonical link already exists, but the error-handling approach uses two - one to discover it exists, and another to create the temporary link. I didn't know how to reconcile the testing strategy with this behavior, so I kept the behavior of first using a 'file-exists?' call to test for the existence of the canonical link. All tests pass except for tests/packages.scm, which failed even without the patches. =2D reepca --=-=-= Content-Type: text/x-patch Content-Disposition: attachment; filename=0001-.dir-locals.el-fix-call-with-retrying-transaction-in.patch Content-Transfer-Encoding: quoted-printable From=204c8f0cc50e2a1a33d9ce2f8e58cc426872676a7f Mon Sep 17 00:00:00 2001 From: Caleb Ristvedt Date: Wed, 24 Jun 2020 01:00:40 -0500 Subject: [PATCH 1/6] .dir-locals.el: fix call-with-{retrying-}transaction indenting. * .dir-locals.el (call-with-transaction, call-with-retrying-transaction): change scheme-indent-function property from 2 to 1. =2D-- .dir-locals.el | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/.dir-locals.el b/.dir-locals.el index 155255a770..d3ec2dd00d 100644 =2D-- a/.dir-locals.el +++ b/.dir-locals.el @@ -89,9 +89,9 @@ (eval . (put 'let-system 'scheme-indent-function 1)) =20 (eval . (put 'with-database 'scheme-indent-function 2)) =2D (eval . (put 'call-with-transaction 'scheme-indent-function 2)) + (eval . (put 'call-with-transaction 'scheme-indent-function 1)) (eval . (put 'with-statement 'scheme-indent-function 3)) =2D (eval . (put 'call-with-retrying-transaction 'scheme-indent-function = 2)) + (eval . (put 'call-with-retrying-transaction 'scheme-indent-function 1)) (eval . (put 'call-with-savepoint 'scheme-indent-function 1)) (eval . (put 'call-with-retrying-savepoint 'scheme-indent-function 1)) =20 =2D-=20 2.28.0 --=-=-= Content-Type: text/x-patch Content-Disposition: attachment; filename=0002-deduplication-pass-store-directory-to-replace-with-l.patch Content-Transfer-Encoding: quoted-printable From=209717568f922e0921e5fdc320cbe6689768d29a29 Mon Sep 17 00:00:00 2001 From: Caleb Ristvedt Date: Sat, 8 Aug 2020 10:05:22 -0500 Subject: [PATCH 2/6] deduplication: pass store directory to replace-with-li= nk. This causes with-writable-file to take into consideration the actual store being used, as passed to 'deduplicate', rather than whatever (%store-directory) may return. * guix/store/deduplication.scm (replace-with-link): new keyword argument 'store'. Pass to with-writable-file. (with-writable-file, call-with-writable-file): new store argument. (deduplicate): pass store to replace-with-link. =2D-- .dir-locals.el | 2 +- guix/store/deduplication.scm | 102 ++++++++++++++++++----------------- 2 files changed, 54 insertions(+), 50 deletions(-) diff --git a/.dir-locals.el b/.dir-locals.el index d3ec2dd00d..419911972d 100644 =2D-- a/.dir-locals.el +++ b/.dir-locals.el @@ -37,7 +37,7 @@ (eval . (put 'with-file-lock 'scheme-indent-function 1)) (eval . (put 'with-file-lock/no-wait 'scheme-indent-function 1)) (eval . (put 'with-profile-lock 'scheme-indent-function 1)) =2D (eval . (put 'with-writable-file 'scheme-indent-function 1)) + (eval . (put 'with-writable-file 'scheme-indent-function 2)) =20 (eval . (put 'package 'scheme-indent-function 0)) (eval . (put 'origin 'scheme-indent-function 0)) diff --git a/guix/store/deduplication.scm b/guix/store/deduplication.scm index df959bdd06..0655ceb890 100644 =2D-- a/guix/store/deduplication.scm +++ b/guix/store/deduplication.scm @@ -94,8 +94,8 @@ LINK-PREFIX." (try (tempname-in link-prefix)) (apply throw args)))))) =20 =2D(define (call-with-writable-file file thunk) =2D (if (string=3D? file (%store-directory)) +(define (call-with-writable-file file store thunk) + (if (string=3D? file store) (thunk) ;don't meddle with the store's permiss= ions (let ((stat (lstat file))) (dynamic-wind @@ -106,17 +106,18 @@ LINK-PREFIX." (set-file-time file stat) (chmod file (stat:mode stat))))))) =20 =2D(define-syntax-rule (with-writable-file file exp ...) +(define-syntax-rule (with-writable-file file store exp ...) "Make FILE writable for the dynamic extent of EXP..., except if FILE is = the store." =2D (call-with-writable-file file (lambda () exp ...))) + (call-with-writable-file file store (lambda () exp ...))) =20 ;; There are 3 main kinds of errors we can get from hardlinking: "Too many ;; things link to this" (EMLINK), "this link already exists" (EEXIST), and ;; "can't fit more stuff in this directory" (ENOSPC). =20 (define* (replace-with-link target to-replace =2D #:key (swap-directory (dirname target))) + #:key (swap-directory (dirname target)) + (store (%store-directory))) "Atomically replace the file TO-REPLACE with a link to TARGET. Use SWAP-DIRECTORY as the directory to store temporary hard links. Upon ENOSPC and EMLINK, TO-REPLACE is left unchanged. @@ -137,7 +138,7 @@ Note: TARGET, TO-REPLACE, and SWAP-DIRECTORY must be on= the same file system." ;; If we couldn't create TEMP-LINK, that's OK: just don't do the ;; replacement, which means TO-REPLACE won't be deduplicated. (when temp-link =2D (with-writable-file (dirname to-replace) + (with-writable-file (dirname to-replace) store (catch 'system-error (lambda () (rename-file temp-link to-replace)) @@ -154,46 +155,49 @@ under STORE." (define links-directory (string-append store "/.links")) =20 =2D (mkdir-p links-directory) =2D (let loop ((path path) =2D (type (stat:type (lstat path))) =2D (hash hash)) =2D (if (eq? 'directory type) =2D ;; Can't hardlink directories, so hardlink their atoms. =2D (for-each (match-lambda =2D ((file . properties) =2D (unless (member file '("." "..")) =2D (let* ((file (string-append path "/" file)) =2D (type (match (assoc-ref properties 'type) =2D ((or 'unknown #f) =2D (stat:type (lstat file))) =2D (type type)))) =2D (loop file type =2D (and (not (eq? 'directory type)) =2D (nar-sha256 file))))))) =2D (scandir* path)) =2D (let ((link-file (string-append links-directory "/" =2D (bytevector->nix-base32-string h= ash)))) =2D (if (file-exists? link-file) =2D (replace-with-link link-file path =2D #:swap-directory links-directory) =2D (catch 'system-error =2D (lambda () =2D (link path link-file)) =2D (lambda args =2D (let ((errno (system-error-errno args))) =2D (cond ((=3D errno EEXIST) =2D ;; Someone else put an entry for PATH in =2D ;; LINKS-DIRECTORY before we could. Let's us= e it. =2D (replace-with-link path link-file =2D #:swap-directory links-dir= ectory)) =2D ((=3D errno ENOSPC) =2D ;; There's not enough room in the directory i= ndex for =2D ;; more entries in .links, but that's fine: w= e can =2D ;; just stop. =2D #f) =2D ((=3D errno EMLINK) =2D ;; PATH has reached the maximum number of lin= ks, but =2D ;; that's OK: we just can't deduplicate it mo= re. =2D #f) =2D (else (apply throw args))))))))))) + (mkdir-p links-directory) + (let loop ((path path) + (type (stat:type (lstat path))) + (hash hash)) + (if (eq? 'directory type) + ;; Can't hardlink directories, so hardlink their atoms. + (for-each (match-lambda + ((file . properties) + (unless (member file '("." "..")) + (let* ((file (string-append path "/" file)) + (type (match (assoc-ref properties 'type) + ((or 'unknown #f) + (stat:type (lstat file))) + (type type)))) + (loop file type + (and (not (eq? 'directory type)) + (nar-sha256 file))))))) + (scandir* path)) + (let ((link-file (string-append links-directory "/" + (bytevector->nix-base32-string h= ash)))) + (if (file-exists? link-file) + (replace-with-link link-file path + #:swap-directory links-directory + #:store store) + (catch 'system-error + (lambda () + (link path link-file)) + (lambda args + (let ((errno (system-error-errno args))) + (cond ((=3D errno EEXIST) + ;; Someone else put an entry for PATH in + ;; LINKS-DIRECTORY before we could. Let's us= e it. + (replace-with-link path link-file + #:swap-directory + links-directory + #:store store)) + ((=3D errno ENOSPC) + ;; There's not enough room in the directory i= ndex for + ;; more entries in .links, but that's fine: w= e can + ;; just stop. + #f) + ((=3D errno EMLINK) + ;; PATH has reached the maximum number of lin= ks, but + ;; that's OK: we just can't deduplicate it mo= re. + #f) + (else (apply throw args))))))))))) =2D-=20 2.28.0 --=-=-= Content-Type: text/x-patch Content-Disposition: attachment; filename=0003-deduplication-retry-on-ENOENT.patch Content-Transfer-Encoding: quoted-printable From=20b992a3aaac7e3b30222e0bf1df09093f18e25e6a Mon Sep 17 00:00:00 2001 From: Caleb Ristvedt Date: Sat, 8 Aug 2020 11:25:57 -0500 Subject: [PATCH 3/6] deduplication: retry on ENOENT. It's possible for the garbage collector to remove the "canonical" link after it's been detected as existing by 'deduplicate'. This would cause an ENOENT error when replace-with-link attempts to create the temporary link. This changes it so that it will properly handle that by retrying. * guix/store/deduplication.scm (replace-with-link): renamed to canonicalize-with-link, now also handles the case where the target link doesn't exist yet, and retries on ENOENT. (deduplicate): modified to use canonicalize-with-link. =2D-- guix/store/deduplication.scm | 103 ++++++++++++++++++----------------- 1 file changed, 53 insertions(+), 50 deletions(-) diff --git a/guix/store/deduplication.scm b/guix/store/deduplication.scm index 0655ceb890..035a4218fb 100644 =2D-- a/guix/store/deduplication.scm +++ b/guix/store/deduplication.scm @@ -115,9 +115,9 @@ store." ;; things link to this" (EMLINK), "this link already exists" (EEXIST), and ;; "can't fit more stuff in this directory" (ENOSPC). =20 =2D(define* (replace-with-link target to-replace =2D #:key (swap-directory (dirname target)) =2D (store (%store-directory))) +(define* (canonicalize-with-link target to-replace + #:key (swap-directory (dirname target)) + (store (%store-directory))) "Atomically replace the file TO-REPLACE with a link to TARGET. Use SWAP-DIRECTORY as the directory to store temporary hard links. Upon ENOSPC and EMLINK, TO-REPLACE is left unchanged. @@ -126,7 +126,32 @@ Note: TARGET, TO-REPLACE, and SWAP-DIRECTORY must be o= n the same file system." (define temp-link (catch 'system-error (lambda () =2D (get-temp-link target swap-directory)) + (let retry () + (if (file-exists? target) + (catch 'system-error + (lambda () + (get-temp-link target swap-directory)) + (lambda args + (let ((errno (system-error-errno args))) + (if (=3D errno ENOENT) + ;; either SWAP-DIRECTORY has missing directory + ;; components or TARGET was deleted - this is a + ;; fundamental ambiguity to the errno produced by + ;; link() + (if (file-exists? swap-directory) + ;; we must assume link failed because target d= oesn't + ;; exist, so create it. + (retry) + (apply throw args)) + (apply throw args))))) + (catch 'system-error + (lambda () + (link to-replace target) + #f) + (lambda args + (if (=3D (system-error-errno args) EEXIST) + (retry) + (apply throw args))))))) (lambda args ;; We get ENOSPC when we can't fit an additional entry in ;; SWAP-DIRECTORY. If it's EMLINK, then TARGET has reached its @@ -155,49 +180,27 @@ under STORE." (define links-directory (string-append store "/.links")) =20 =2D (mkdir-p links-directory) =2D (let loop ((path path) =2D (type (stat:type (lstat path))) =2D (hash hash)) =2D (if (eq? 'directory type) =2D ;; Can't hardlink directories, so hardlink their atoms. =2D (for-each (match-lambda =2D ((file . properties) =2D (unless (member file '("." "..")) =2D (let* ((file (string-append path "/" file)) =2D (type (match (assoc-ref properties 'type) =2D ((or 'unknown #f) =2D (stat:type (lstat file))) =2D (type type)))) =2D (loop file type =2D (and (not (eq? 'directory type)) =2D (nar-sha256 file))))))) =2D (scandir* path)) =2D (let ((link-file (string-append links-directory "/" =2D (bytevector->nix-base32-string= hash)))) =2D (if (file-exists? link-file) =2D (replace-with-link link-file path =2D #:swap-directory links-directory =2D #:store store) =2D (catch 'system-error =2D (lambda () =2D (link path link-file)) =2D (lambda args =2D (let ((errno (system-error-errno args))) =2D (cond ((=3D errno EEXIST) =2D ;; Someone else put an entry for PATH in =2D ;; LINKS-DIRECTORY before we could. Let's = use it. =2D (replace-with-link path link-file =2D #:swap-directory =2D links-directory =2D #:store store)) =2D ((=3D errno ENOSPC) =2D ;; There's not enough room in the directory= index for =2D ;; more entries in .links, but that's fine:= we can =2D ;; just stop. =2D #f) =2D ((=3D errno EMLINK) =2D ;; PATH has reached the maximum number of l= inks, but =2D ;; that's OK: we just can't deduplicate it = more. =2D #f) =2D (else (apply throw args))))))))))) + (mkdir-p links-directory) + (let loop ((path path) + (type (stat:type (lstat path))) + (hash hash)) + (if (eq? 'directory type) + ;; Can't hardlink directories, so hardlink their atoms. + (for-each (match-lambda + ((file . properties) + (unless (member file '("." "..")) + (let* ((file (string-append path "/" file)) + (type (match (assoc-ref properties 'type) + ((or 'unknown #f) + (stat:type (lstat file))) + (type type)))) + (loop file type + (and (not (eq? 'directory type)) + (nar-sha256 file))))))) + (scandir* path)) + (let ((link-file (string-append links-directory "/" + (bytevector->nix-base32-string + hash)))) + (canonicalize-with-link link-file path + #:swap-directory links-directory + #:store store))))) =2D-=20 2.28.0 --=-=-= Content-Type: text/x-patch Content-Disposition: attachment; filename=0004-nar-ensure-finalization-target-is-a-temp-root-during.patch Content-Transfer-Encoding: quoted-printable From=206b7d011680c77642f24396be0eb0015c20413d1a Mon Sep 17 00:00:00 2001 From: Caleb Ristvedt Date: Sat, 8 Aug 2020 08:31:38 -0500 Subject: [PATCH 4/6] nar: ensure finalization target is a temp root during registration. Note that this is currently unnecessary, as finalize-store-file is only used from the offload hook, and the daemon process that spawned the offload hook will have already registered the derivation outputs as temp roots prior to attempting to offload (see haveDerivation() in nix/libstore/build.cc). But it's necessary to ensure that the register-items invocation works properly when finalize-store-file is used in other contexts. * guix/nar.scm (finalize-store-file): make target a temp root during extent= of register-items invocation. =2D-- guix/nar.scm | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-) diff --git a/guix/nar.scm b/guix/nar.scm index a2aacf585c..ffb487e185 100644 =2D-- a/guix/nar.scm +++ b/guix/nar.scm @@ -111,13 +111,24 @@ held." (when (file-exists? target) (delete-file-recursively target)) =20 =2D ;; Install the new TARGET. =2D (rename-file source target) + ;; Register TARGET as a temp root. Note that this may not alway= s be + ;; necessary, but adding a temp root multiple times doesn't chan= ge + ;; anything (except taking a little more space in the temproots + ;; file). Note that currently this will only ensure that TARGET + ;; doesn't get deleted between now and when registration finishe= s; + ;; it may well be deleted by the time this procedure returns. =20 =2D ;; Register TARGET. As a side effect, it resets the timestamp= s of all =2D ;; its files, recursively, and runs a deduplication pass. =2D (register-items db =2D (list (store-info target deriver references)))) + ;; TODO: don't use an RPC for this, add it to *this process's* t= emp + ;; roots file. + (with-store store + (add-temp-root store target) + ;; Install the new TARGET. + (rename-file source target) + + ;; Register TARGET. As a side effect, it resets the timestamp= s of + ;; all its files, recursively, and runs a deduplication pass. + (register-items db + (list (store-info target deriver references)))= )) =20 (when lock? (delete-file (string-append target ".lock")) =2D-=20 2.28.0 --=-=-= Content-Type: text/x-patch Content-Disposition: attachment; filename=0005-database-document-extra-registration-requirements.patch Content-Transfer-Encoding: quoted-printable From=2055dd48e88d641bbc17b4d9484d6ee84acfb29766 Mon Sep 17 00:00:00 2001 From: Caleb Ristvedt Date: Wed, 8 Jul 2020 11:33:23 -0500 Subject: [PATCH 5/6] database: document extra registration requirements. It's necessary that store items be locked and protected from garbage collection while they are being registered. This documents that. * guix/store/database.scm (register-path, register-items): document GC protection and locking requirements. =2D-- guix/store/database.scm | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/guix/store/database.scm b/guix/store/database.scm index 50b66ce282..e39a1603af 100644 =2D-- a/guix/store/database.scm +++ b/guix/store/database.scm @@ -397,7 +397,10 @@ absolute file name to the state directory of the store= being initialized. Return #t on success. =20 Use with care as it directly modifies the store! This is primarily meant = to =2Dbe used internally by the daemon's build hook." +be used internally by the daemon's build hook. + +PATH must be protected from GC and locked during execution of this, typica= lly +by adding it as a temp-root." (define db-file (store-database-file #:prefix prefix #:state-directory state-directory)) @@ -423,7 +426,9 @@ be used internally by the daemon's build hook." "Register all of ITEMS, a list of records as returned by 'read-reference-graph', in DB. ITEMS must be in topological order (with leaves first.) REGISTRATION-TIME must be the registration time to be reco= rded =2Din the database; #f means \"now\". Write a progress report to LOG-PORT." +in the database; #f means \"now\". Write a progress report to LOG-PORT. = All +of ITEMS must be protected from GC and locked during execution of this, +typically by adding them as temp-roots." (define store-dir (if prefix (string-append prefix %storedir) =2D-=20 2.28.0 --=-=-= Content-Type: text/x-patch Content-Disposition: attachment; filename=0006-database-register-items-reduce-transaction-scope.patch Content-Transfer-Encoding: quoted-printable From=2030afb453ce4eb161bb87645a0e6314e6af82a61f Mon Sep 17 00:00:00 2001 From: Christopher Baines Date: Tue, 23 Jun 2020 17:36:49 +0100 Subject: [PATCH 6/6] database: register-items: reduce transaction scope. It was made transactional in a4678c6ba18d8dbd79d931f80426eebf61be7ebe, with the reasoning to prevent broken intermediate states from being visible. I think this means something like an entry being in ValidPaths, but the Refs = not being inserted. Using a transaction for this makes sense, but I think using one single transaction for the whole register-items call is unnecessary to avoid broken states from being visible, and could block other writes to the store databa= se while register-items is running. Because the deduplication and resetting timestamps happens within the transaction as well, even though these things don't involve the database, writes to the database will still be blocked wh= ile this is happening. To reduce the potential for register-items to block other writers to the database for extended periods, this commit moves the transaction to just wr= ap the call to sqlite-register. This is the one place where writes occur, so t= hat should prevent the broken intermediate states issue above. The one differen= ce this will make is some of the registered items will be visible to other connections while others may be still being added. I think this is OK, as i= t's equivalent to just registering different items. * guix/store/database.scm (register-items): Reduce transaction scope. =2D-- guix/store/database.scm | 37 +++++++++++++++++++------------------ 1 file changed, 19 insertions(+), 18 deletions(-) diff --git a/guix/store/database.scm b/guix/store/database.scm index e39a1603af..2ea63b17aa 100644 =2D-- a/guix/store/database.scm +++ b/guix/store/database.scm @@ -457,24 +457,25 @@ typically by adding them as temp-roots." (when reset-timestamps? (reset-timestamps real-file-name)) (let-values (((hash nar-size) (nar-sha256 real-file-name))) =2D (sqlite-register db #:path to-register =2D #:references (store-info-references item) =2D #:deriver (store-info-deriver item) =2D #:hash (string-append "sha256:" =2D (bytevector->base16-strin= g hash)) =2D #:nar-size nar-size =2D #:time registration-time) + (call-with-retrying-transaction db + (lambda () + (sqlite-register db #:path to-register + #:references (store-info-references item) + #:deriver (store-info-deriver item) + #:hash (string-append + "sha256:" + (bytevector->base16-string hash)) + #:nar-size nar-size + #:time registration-time))) (when deduplicate? (deduplicate real-file-name hash #:store store-dir))))) =20 =2D (call-with-retrying-transaction db =2D (lambda () =2D (let* ((prefix (format #f "registering ~a items" (length items= ))) =2D (progress (progress-reporter/bar (length items) =2D prefix log-port))) =2D (call-with-progress-reporter progress =2D (lambda (report) =2D (for-each (lambda (item) =2D (register db item) =2D (report)) =2D items))))))) + (let* ((prefix (format #f "registering ~a items" (length items))) + (progress (progress-reporter/bar (length items) + prefix log-port))) + (call-with-progress-reporter progress + (lambda (report) + (for-each (lambda (item) + (register db item) + (report)) + items))))) =2D-=20 2.28.0 --=-=-=-- --==-=-= Content-Type: application/pgp-signature; name="signature.asc" -----BEGIN PGP SIGNATURE----- iQEzBAEBCAAdFiEEdNapMPRLm4SepVYGwWaqSV9/GJwFAl8vd/AACgkQwWaqSV9/ GJzdygf/fnQKTxIQzOI8DAvh8Avz3GXx7pESnxYhXCSa1vHN2LD0gQwytS4bHN1B 1vEWqaG+i4aIBF97oIEWMQXm/YrlhEburbQ+/aoKMJzYui4p3vQZedCEt0hrlEsD 5qf7AdIQDwPAScJeeeADvkRLy30+H9ui72iRKqpqPgm6kCHd+1NZS8KGqFYhTe2Y VPo+12xW8dkJguT8FMVYJYj7V/JRtSrGs+0sItBcwpd78HCSV914l0FUdEbPzcqk QbtzyY0a6r+Ul8FpPUYWLvqUNUCNK+kMWE57MpHEzKvosBB/9IerU00WBEC48LIP sZusS4tkpSmbUB3N19lSAV9XizrPfw== =CeVX -----END PGP SIGNATURE----- --==-=-=--