From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Turner Subject: [PATCH v6 21/32] refs: resolve symbolic refs first Date: Wed, 24 Feb 2016 17:58:53 -0500 Message-ID: <1456354744-8022-22-git-send-email-dturner@twopensource.com> References: <1456354744-8022-1-git-send-email-dturner@twopensource.com> Cc: David Turner To: git@vger.kernel.org, mhagger@alum.mit.edu, pclouds@gmail.com X-From: git-owner@vger.kernel.org Thu Feb 25 00:00:56 2016 Return-path: Envelope-to: gcvg-git-2@plane.gmane.org Received: from vger.kernel.org ([209.132.180.67]) by plane.gmane.org with esmtp (Exim 4.69) (envelope-from ) id 1aYiQQ-0004YF-98 for gcvg-git-2@plane.gmane.org; Thu, 25 Feb 2016 00:00:54 +0100 Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1759268AbcBXXAw (ORCPT ); Wed, 24 Feb 2016 18:00:52 -0500 Received: from mail-qg0-f41.google.com ([209.85.192.41]:34661 "EHLO mail-qg0-f41.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1759171AbcBXW7i (ORCPT ); Wed, 24 Feb 2016 17:59:38 -0500 Received: by mail-qg0-f41.google.com with SMTP id b67so27014181qgb.1 for ; Wed, 24 Feb 2016 14:59:38 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=twopensource-com.20150623.gappssmtp.com; s=20150623; h=from:to:cc:subject:date:message-id:in-reply-to:references; bh=cB546VAIBP0MnFvkVntT8IreLsHWcJ1Z3BIJTcSV4GE=; b=jsvv4qjoxhYjQtDSyzlyTPK98CdkzYAYyV/MIdP+PCQRl1Q6yPYhzTGOSxxqnLqgc6 oMWpvBrlEuQxFh15nQpZ0YMJvzmI6eGInPg6wGwftNnyIsIahx2LS7flVUniZSz4sxFw UwnIAWY++RozVn67YzAsmepsKs6JgPhZEaheT16jkWvkx+ER9OCPqPqsem9oRXY9xJpE bLXK6xAndUR460wla2iuKoklxIgbcLXyqh4ntiCj3USvgAAkwrD4r4m/Nb+2nz0urnTn Z3YhKevxhmIoB+TQDztNH0nliW1pkhh5ilbyHP6ZpPUViy8LbMUzfkmVKcbSz4EDINpU wgyA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20130820; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references; bh=cB546VAIBP0MnFvkVntT8IreLsHWcJ1Z3BIJTcSV4GE=; b=Y+oncwAitrNwBvavRjSe2dh6/uMY1cjRC9a/Kl2/YB4/hiUgY2xcCi7ar7y2E56cCC lShRjH3kVKC+VhleODFKJ3hlRtdIu6IsaR3n1mu+jdIhF0bKokJjdQk+9HmvzGDe8rVn sdatNOpqvM3uHQznbkQ/IGPm6gQL+zOvNsdWIAcBO3ECtLZzZVj1lBHHGdGJN9n5u/rO jbOk6vM6l2dmPfzZW12S8lic1Y5IuGmlPQqIXdnR6tlqjCIz4iBJJq/R9DqreVhGJmhJ hKcNMfGxV1CO3M8YEOKhP+2eQqZS4HqWtwwUTVkpFqimrUxNZ/eEi978bPAyY/09lz5w VH2g== X-Gm-Message-State: AG10YOTunZWCNS1g0DVbjeQO2anqrGJg/Bc8a5n4oLD+M9BX9zXOJ2NTV5EPHR2r4BJmGw== X-Received: by 10.140.96.84 with SMTP id j78mr50688442qge.93.1456354777696; Wed, 24 Feb 2016 14:59:37 -0800 (PST) Received: from ubuntu.twitter.biz ([192.133.79.145]) by smtp.gmail.com with ESMTPSA id 66sm2154254qhp.4.2016.02.24.14.59.36 (version=TLSv1/SSLv3 cipher=OTHER); Wed, 24 Feb 2016 14:59:36 -0800 (PST) X-Mailer: git-send-email 2.4.2.767.g62658d5-twtrsrc In-Reply-To: <1456354744-8022-1-git-send-email-dturner@twopensource.com> Sender: git-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: git@vger.kernel.org Archived-At: Before committing ref updates, split symbolic ref updates into two parts: an update to the underlying ref, and a log-only update to the symbolic ref. This ensures that both references are locked correctly while their reflogs are updated. It is still possible to confuse git by concurrent updates, since the splitting of symbolic refs does not happen under lock. So a symbolic ref could be replaced by a plain ref in the middle of this operation, which would lead to reflog discontinuities and missed old-ref checks. All callers to lock_ref_sha1_basic outside of ref_transaction_commit now set REF_NODEREF. And ref_transaction_commit only happens on already-derefernced refs. So we can assume REF_NODEREF when resolving inside this function. Signed-off-by: David Turner --- refs.c | 69 ++++++++++++++++++++++++++ refs/files-backend.c | 137 ++++++++++++++++++++++++++------------------------- refs/refs-internal.h | 8 +++ 3 files changed, 148 insertions(+), 66 deletions(-) diff --git a/refs.c b/refs.c index daf92ce..bdde808 100644 --- a/refs.c +++ b/refs.c @@ -1351,6 +1351,71 @@ int refs_init_db(int shared, struct strbuf *err) return the_refs_backend->init_db(shared, err); } +/* + * Special case for symbolic refs when REF_NODEREF is not turned on. + * Dereference them here, mark them REF_LOG_ONLY, and add an update + * for the underlying ref. + */ +static int dereference_symrefs(struct ref_transaction *transaction, + struct strbuf *err) +{ + int i; + int nr = transaction->nr; + + for (i = 0; i < nr; i++) { + struct ref_update *update = transaction->updates[i]; + const char *resolved; + unsigned char sha1[20]; + int resolve_flags = 0; + int mustexist = update->flags & REF_HAVE_OLD && + !is_null_sha1(update->old_sha1); + int deleting = (update->flags & REF_HAVE_NEW) && + is_null_sha1(update->new_sha1); + + if (mustexist) + resolve_flags |= RESOLVE_REF_READING; + if (deleting) + resolve_flags |= RESOLVE_REF_ALLOW_BAD_NAME | + RESOLVE_REF_NO_RECURSE; + + if (strcmp(update->refname, "HEAD")) + update->flags |= REF_IS_NOT_HEAD; + + resolved = resolve_ref_unsafe(update->refname, resolve_flags, + sha1, &update->type); + if (!resolved) { + /* + * We may notice this breakage later and die + * with a sensible error message + */ + update->type |= REF_ISBROKEN; + continue; + } + + hashcpy(update->read_sha1, sha1); + + if (update->flags & REF_NODEREF || + !(update->type & REF_ISSYMREF)) + continue; + + /* Create a new transaction for the underlying ref */ + if (ref_transaction_update(transaction, + resolved, + update->new_sha1, + (update->flags & REF_HAVE_OLD) ? + update->old_sha1 : NULL, + update->flags & ~REF_IS_NOT_HEAD, + update->msg, err)) + return -1; + + /* Make the symbolic ref update non-recursive */ + update->flags |= REF_LOG_ONLY | REF_NODEREF; + update->flags &= ~REF_HAVE_OLD; + } + + return 0; +} + int ref_transaction_commit(struct ref_transaction *transaction, struct strbuf *err) { @@ -1367,6 +1432,10 @@ int ref_transaction_commit(struct ref_transaction *transaction, return 0; } + ret = dereference_symrefs(transaction, err); + if (ret) + goto done; + if (get_affected_refnames(transaction, &affected_refnames, err)) { ret = TRANSACTION_GENERIC_ERROR; goto done; diff --git a/refs/files-backend.c b/refs/files-backend.c index f1045b4..f378f06 100644 --- a/refs/files-backend.c +++ b/refs/files-backend.c @@ -7,7 +7,6 @@ struct ref_lock { char *ref_name; - char *orig_ref_name; struct lock_file *lk; struct object_id old_oid; }; @@ -1723,7 +1722,6 @@ static void unlock_ref(struct ref_lock *lock) if (lock->lk) rollback_lock_file(lock->lk); free(lock->ref_name); - free(lock->orig_ref_name); free(lock); } @@ -1779,6 +1777,7 @@ static int remove_empty_directories(struct strbuf *path) */ static struct ref_lock *lock_ref_sha1_basic(const char *refname, const unsigned char *old_sha1, + const unsigned char *read_sha1, const struct string_list *extras, const struct string_list *skip, unsigned int flags, int *type_p, @@ -1786,14 +1785,14 @@ static struct ref_lock *lock_ref_sha1_basic(const char *refname, { struct strbuf ref_file = STRBUF_INIT; struct strbuf orig_ref_file = STRBUF_INIT; - const char *orig_refname = refname; struct ref_lock *lock; int last_errno = 0; int type; int lflags = 0; int mustexist = (old_sha1 && !is_null_sha1(old_sha1)); - int resolve_flags = 0; + int resolve_flags = RESOLVE_REF_NO_RECURSE; int attempts_remaining = 3; + int resolved; assert(err); @@ -1803,65 +1802,65 @@ static struct ref_lock *lock_ref_sha1_basic(const char *refname, resolve_flags |= RESOLVE_REF_READING; if (flags & REF_DELETING) resolve_flags |= RESOLVE_REF_ALLOW_BAD_NAME; - if (flags & REF_NODEREF) { - resolve_flags |= RESOLVE_REF_NO_RECURSE; + if (flags & REF_NODEREF) lflags |= LOCK_NO_DEREF; - } - refname = resolve_ref_unsafe(refname, resolve_flags, - lock->old_oid.hash, &type); - if (!refname && errno == EISDIR) { - /* - * we are trying to lock foo but we used to - * have foo/bar which now does not exist; - * it is normal for the empty directory 'foo' - * to remain. - */ - strbuf_git_path(&orig_ref_file, "%s", orig_refname); - if (remove_empty_directories(&orig_ref_file)) { + if (type_p && *type_p & REF_ISSYMREF) { + hashcpy(lock->old_oid.hash, read_sha1); + } else { + resolved = !!resolve_ref_unsafe(refname, resolve_flags, + lock->old_oid.hash, &type); + if (!resolved && errno == EISDIR) { + /* + * we are trying to lock foo but we used to + * have foo/bar which now does not exist; + * it is normal for the empty directory 'foo' + * to remain. + */ + strbuf_git_path(&orig_ref_file, "%s", refname); + if (remove_empty_directories(&orig_ref_file)) { + struct ref_dir *loose_refs; + loose_refs = get_loose_refs(&ref_cache); + last_errno = errno; + if (!verify_refname_available_dir(refname, extras, skip, + loose_refs, err)) + strbuf_addf(err, "there are still refs under '%s'", + refname); + goto error_return; + } + resolved = !!resolve_ref_unsafe(refname, resolve_flags, + lock->old_oid.hash, &type); + } + + if (type_p) + *type_p = type; + if (!resolved) { last_errno = errno; - if (!verify_refname_available_dir(orig_refname, extras, skip, + if (last_errno != ENOTDIR || + !verify_refname_available_dir(refname, extras, skip, get_loose_refs(&ref_cache), err)) - strbuf_addf(err, "there are still refs under '%s'", - orig_refname); + strbuf_addf(err, + "unable to resolve reference %s: %s", + refname, strerror(last_errno)); + + goto error_return; + } + /* + * If the ref did not exist and we are creating it, make sure + * there is no existing packed ref whose name begins with our + * refname, nor a packed ref whose name is a proper prefix of + * our refname. + */ + if (is_null_oid(&lock->old_oid) && + verify_refname_available_dir(refname, extras, skip, + get_packed_refs(&ref_cache), err)) { + last_errno = ENOTDIR; goto error_return; } - refname = resolve_ref_unsafe(orig_refname, resolve_flags, - lock->old_oid.hash, &type); - } - if (type_p) - *type_p = type; - if (!refname) { - last_errno = errno; - if (last_errno != ENOTDIR || - !verify_refname_available_dir(orig_refname, extras, skip, - get_loose_refs(&ref_cache), err)) - strbuf_addf(err, "unable to resolve reference %s: %s", - orig_refname, strerror(last_errno)); - - goto error_return; - } - - if (flags & REF_NODEREF) - refname = orig_refname; - - /* - * If the ref did not exist and we are creating it, make sure - * there is no existing packed ref whose name begins with our - * refname, nor a packed ref whose name is a proper prefix of - * our refname. - */ - if (is_null_oid(&lock->old_oid) && - verify_refname_available_dir(refname, extras, skip, - get_packed_refs(&ref_cache), err)) { - last_errno = ENOTDIR; - goto error_return; } - lock->lk = xcalloc(1, sizeof(struct lock_file)); lock->ref_name = xstrdup(refname); - lock->orig_ref_name = xstrdup(orig_refname); strbuf_git_path(&ref_file, "%s", refname); retry: @@ -1893,7 +1892,13 @@ static struct ref_lock *lock_ref_sha1_basic(const char *refname, goto error_return; } } - if (verify_lock(lock, old_sha1, mustexist, err)) { + + if (type_p && *type_p & REF_ISSYMREF && !(*type_p & REF_ISBROKEN)) { + /* + * Old hash verification for symrefs happens on their + * base ref. + */ + } else if (verify_lock(lock, old_sha1, mustexist, err)) { last_errno = errno; goto error_return; } @@ -2428,8 +2433,8 @@ static int files_rename_ref(const char *oldrefname, const char *newrefname, logmoved = log; - lock = lock_ref_sha1_basic(newrefname, NULL, NULL, NULL, REF_NODEREF, - NULL, &err); + lock = lock_ref_sha1_basic(newrefname, NULL, NULL, NULL, NULL, + REF_NODEREF, NULL, &err); if (!lock) { error("unable to rename '%s' to '%s': %s", oldrefname, newrefname, err.buf); strbuf_release(&err); @@ -2447,8 +2452,9 @@ static int files_rename_ref(const char *oldrefname, const char *newrefname, return 0; rollback: - lock = lock_ref_sha1_basic(oldrefname, NULL, NULL, NULL, REF_NODEREF, - NULL, &err); + lock = lock_ref_sha1_basic(oldrefname, NULL, NULL, NULL, NULL, + REF_NODEREF, NULL, &err); + if (!lock) { error("unable to lock %s for rollback: %s", oldrefname, err.buf); strbuf_release(&err); @@ -2676,9 +2682,7 @@ static int commit_ref_update(struct ref_lock *lock, int flags, struct strbuf *err) { clear_loose_ref_cache(&ref_cache); - if (log_ref_write(lock->ref_name, lock->old_oid.hash, sha1, logmsg, flags, err) < 0 || - (strcmp(lock->ref_name, lock->orig_ref_name) && - log_ref_write(lock->orig_ref_name, lock->old_oid.hash, sha1, logmsg, flags, err) < 0)) { + if (log_ref_write(lock->ref_name, lock->old_oid.hash, sha1, logmsg, flags, err) < 0) { char *old_msg = strbuf_detach(err, NULL); strbuf_addf(err, "Cannot update the ref '%s': %s", lock->ref_name, old_msg); @@ -2686,7 +2690,7 @@ static int commit_ref_update(struct ref_lock *lock, unlock_ref(lock); return -1; } - if (strcmp(lock->orig_ref_name, "HEAD") != 0) { + if (flags & REF_IS_NOT_HEAD) { /* * Special hack: If a branch is updated directly and HEAD * points to it (may happen on the remote side of a push @@ -2781,8 +2785,8 @@ static int files_create_symref(const char *refname, struct ref_lock *lock; int ret; - lock = lock_ref_sha1_basic(refname, NULL, NULL, NULL, REF_NODEREF, NULL, - &err); + lock = lock_ref_sha1_basic(refname, NULL, NULL, NULL, NULL, REF_NODEREF, + NULL, &err); if (!lock) { error("%s", err.buf); strbuf_release(&err); @@ -3050,6 +3054,7 @@ static int files_transaction_commit(struct ref_transaction *transaction, update->refname, ((update->flags & REF_HAVE_OLD) ? update->old_sha1 : NULL), + update->read_sha1, affected_refnames, NULL, update->flags, &update->type, @@ -3296,7 +3301,7 @@ static int files_reflog_expire(const char *refname, const unsigned char *sha1, struct ref_lock *lock; char *log_file; int status = 0; - int type; + int type = 0; struct strbuf err = STRBUF_INIT; memset(&cb, 0, sizeof(cb)); @@ -3309,7 +3314,7 @@ static int files_reflog_expire(const char *refname, const unsigned char *sha1, * reference itself, plus we might need to update the * reference if --updateref was specified: */ - lock = lock_ref_sha1_basic(refname, sha1, NULL, NULL, REF_NODEREF, + lock = lock_ref_sha1_basic(refname, sha1, NULL, NULL, NULL, REF_NODEREF, &type, &err); if (!lock) { error("cannot lock ref '%s': %s", refname, err.buf); diff --git a/refs/refs-internal.h b/refs/refs-internal.h index dd76246..0d3f9e7 100644 --- a/refs/refs-internal.h +++ b/refs/refs-internal.h @@ -49,6 +49,8 @@ */ #define REF_LOG_ONLY 0x80 +#define REF_IS_NOT_HEAD 0x100 + /* * Return true iff refname is minimally safe. "Safe" here means that * deleting a loose reference by this name will not do any damage, for @@ -147,6 +149,12 @@ struct ref_update { */ unsigned char old_sha1[20]; /* + * During the symbolic ref split stage, we resolve refs. + * We'll re-resolve non-symbolic refs once they are locked, + * but we store this to avoid re-resolving symbolic refs. + */ + unsigned char read_sha1[20]; + /* * One or more of REF_HAVE_NEW, REF_HAVE_OLD, REF_NODEREF, * REF_DELETING, and REF_ISPRUNING: */ -- 2.4.2.767.g62658d5-twtrsrc