feat(disk-cache): use fsync to force flush changes on NFS, and use retry mechanism to archive caches...

rhodecode/apps/repository/views/repo_files.py

@@ -1,1716 +1,1716 b''
 # Copyright (C) 2011-2023 RhodeCode GmbH
 #
 # This program is free software: you can redistribute it and/or modify
 # it under the terms of the GNU Affero General Public License, version 3
 # (only), as published by the Free Software Foundation.
 #
 # This program is distributed in the hope that it will be useful,
 # but WITHOUT ANY WARRANTY; without even the implied warranty of
 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 # GNU General Public License for more details.
 #
 # You should have received a copy of the GNU Affero General Public License
 # along with this program.  If not, see <http://www.gnu.org/licenses/>.
 #
 # This program is dual-licensed. If you wish to learn more about the
 # RhodeCode Enterprise Edition, including its added features, Support services,
 # and proprietary license terms, please see https://rhodecode.com/licenses/
 
 import itertools
 import logging
 import os
 import collections
 import urllib.request
 import urllib.parse
 import urllib.error
 import pathlib
 import time
 import random
 
 from pyramid.httpexceptions import HTTPNotFound, HTTPBadRequest, HTTPFound
 
 from pyramid.renderers import render
 from pyramid.response import Response
 
 import rhodecode
 from rhodecode.apps._base import RepoAppView
 
 
 from rhodecode.lib import diffs, helpers as h, rc_cache
 from rhodecode.lib import audit_logger
 from rhodecode.lib.hash_utils import sha1_safe
 from rhodecode.lib.rc_cache.archive_cache import (
     get_archival_cache_store, get_archival_config, ArchiveCacheGenerationLock, archive_iterator)
 from rhodecode.lib.str_utils import safe_bytes, convert_special_chars
 from rhodecode.lib.view_utils import parse_path_ref
 from rhodecode.lib.exceptions import NonRelativePathError
 from rhodecode.lib.codeblocks import (
     filenode_as_lines_tokens, filenode_as_annotated_lines_tokens)
 from rhodecode.lib.utils2 import convert_line_endings, detect_mode
 from rhodecode.lib.type_utils import str2bool
 from rhodecode.lib.str_utils import safe_str, safe_int
 from rhodecode.lib.auth import (
     LoginRequired, HasRepoPermissionAnyDecorator, CSRFRequired)
 from rhodecode.lib.vcs import path as vcspath
 from rhodecode.lib.vcs.backends.base import EmptyCommit
 from rhodecode.lib.vcs.conf import settings
 from rhodecode.lib.vcs.nodes import FileNode
 from rhodecode.lib.vcs.exceptions import (
     RepositoryError, CommitDoesNotExistError, EmptyRepositoryError,
     ImproperArchiveTypeError, VCSError, NodeAlreadyExistsError,
     NodeDoesNotExistError, CommitError, NodeError)
 
 from rhodecode.model.scm import ScmModel
 from rhodecode.model.db import Repository
 
 log = logging.getLogger(__name__)
 
 
 def get_archive_name(db_repo_id, db_repo_name, commit_sha, ext, subrepos=False, path_sha='', with_hash=True):
     # original backward compat name of archive
     clean_name = safe_str(convert_special_chars(db_repo_name).replace('/', '_'))
 
     # e.g vcsserver-id-abcd-sub-1-abcfdef-archive-all.zip
     # vcsserver-id-abcd-sub-0-abcfdef-COMMIT_SHA-PATH_SHA.zip
     id_sha = sha1_safe(str(db_repo_id))[:4]
     sub_repo = 'sub-1' if subrepos else 'sub-0'
     commit = commit_sha if with_hash else 'archive'
     path_marker = (path_sha if with_hash else '') or 'all'
     archive_name = f'{clean_name}-id-{id_sha}-{sub_repo}-{commit}-{path_marker}{ext}'
 
     return archive_name
 
 
 def get_path_sha(at_path):
     return safe_str(sha1_safe(at_path)[:8])
 
 
 def _get_archive_spec(fname):
     log.debug('Detecting archive spec for: `%s`', fname)
 
     fileformat = None
     ext = None
     content_type = None
     for a_type, content_type, extension in settings.ARCHIVE_SPECS:
 
         if fname.endswith(extension):
             fileformat = a_type
             log.debug('archive is of type: %s', fileformat)
             ext = extension
             break
 
     if not fileformat:
         raise ValueError()
 
     # left over part of whole fname is the commit
     commit_id = fname[:-len(ext)]
 
     return commit_id, ext, fileformat, content_type
 
 
 class RepoFilesView(RepoAppView):
 
     @staticmethod
     def adjust_file_path_for_svn(f_path, repo):
         """
         Computes the relative path of `f_path`.
 
         This is mainly based on prefix matching of the recognized tags and
         branches in the underlying repository.
         """
         tags_and_branches = itertools.chain(
             repo.branches.keys(),
             repo.tags.keys())
         tags_and_branches = sorted(tags_and_branches, key=len, reverse=True)
 
         for name in tags_and_branches:
             if f_path.startswith(f'{name}/'):
                 f_path = vcspath.relpath(f_path, name)
                 break
         return f_path
 
     def load_default_context(self):
         c = self._get_local_tmpl_context(include_app_defaults=True)
         c.rhodecode_repo = self.rhodecode_vcs_repo
         c.enable_downloads = self.db_repo.enable_downloads
         return c
 
     def _ensure_not_locked(self, commit_id='tip'):
         _ = self.request.translate
 
         repo = self.db_repo
         if repo.enable_locking and repo.locked[0]:
             h.flash(_('This repository has been locked by %s on %s')
                     % (h.person_by_id(repo.locked[0]),
                     h.format_date(h.time_to_datetime(repo.locked[1]))),
                     'warning')
             files_url = h.route_path(
                 'repo_files:default_path',
                 repo_name=self.db_repo_name, commit_id=commit_id)
             raise HTTPFound(files_url)
 
     def forbid_non_head(self, is_head, f_path, commit_id='tip', json_mode=False):
         _ = self.request.translate
 
         if not is_head:
             message = _('Cannot modify file. '
                         'Given commit `{}` is not head of a branch.').format(commit_id)
             h.flash(message, category='warning')
 
             if json_mode:
                 return message
 
             files_url = h.route_path(
                 'repo_files', repo_name=self.db_repo_name, commit_id=commit_id,
                 f_path=f_path)
             raise HTTPFound(files_url)
 
     def check_branch_permission(self, branch_name, commit_id='tip', json_mode=False):
         _ = self.request.translate
 
         rule, branch_perm = self._rhodecode_user.get_rule_and_branch_permission(
             self.db_repo_name, branch_name)
         if branch_perm and branch_perm not in ['branch.push', 'branch.push_force']:
             message = _('Branch `{}` changes forbidden by rule {}.').format(
                 h.escape(branch_name), h.escape(rule))
             h.flash(message, 'warning')
 
             if json_mode:
                 return message
 
             files_url = h.route_path(
                 'repo_files:default_path', repo_name=self.db_repo_name, commit_id=commit_id)
 
             raise HTTPFound(files_url)
 
     def _get_commit_and_path(self):
         default_commit_id = self.db_repo.landing_ref_name
         default_f_path = '/'
 
         commit_id = self.request.matchdict.get(
             'commit_id', default_commit_id)
         f_path = self._get_f_path(self.request.matchdict, default_f_path)
         return commit_id, f_path
 
     def _get_default_encoding(self, c):
         enc_list = getattr(c, 'default_encodings', [])
         return enc_list[0] if enc_list else 'UTF-8'
 
     def _get_commit_or_redirect(self, commit_id, redirect_after=True):
         """
         This is a safe way to get commit. If an error occurs it redirects to
         tip with proper message
 
         :param commit_id: id of commit to fetch
         :param redirect_after: toggle redirection
         """
         _ = self.request.translate
 
         try:
             return self.rhodecode_vcs_repo.get_commit(commit_id)
         except EmptyRepositoryError:
             if not redirect_after:
                 return None
 
             add_new = upload_new = ""
             if h.HasRepoPermissionAny(
                     'repository.write', 'repository.admin')(self.db_repo_name):
                 _url = h.route_path(
                     'repo_files_add_file',
                     repo_name=self.db_repo_name, commit_id=0, f_path='')
                 add_new = h.link_to(
                     _('add a new file'), _url, class_="alert-link")
 
                 _url_upld = h.route_path(
                     'repo_files_upload_file',
                     repo_name=self.db_repo_name, commit_id=0, f_path='')
                 upload_new = h.link_to(
                     _('upload a new file'), _url_upld, class_="alert-link")
 
             h.flash(h.literal(
                 _('There are no files yet. Click here to %s or %s.') % (add_new, upload_new)), category='warning')
             raise HTTPFound(
                 h.route_path('repo_summary', repo_name=self.db_repo_name))
 
         except (CommitDoesNotExistError, LookupError) as e:
             msg = _('No such commit exists for this repository. Commit: {}').format(commit_id)
             h.flash(msg, category='error')
             raise HTTPNotFound()
         except RepositoryError as e:
             h.flash(h.escape(safe_str(e)), category='error')
             raise HTTPNotFound()
 
     def _get_filenode_or_redirect(self, commit_obj, path, pre_load=None):
         """
         Returns file_node, if error occurs or given path is directory,
         it'll redirect to top level path
         """
         _ = self.request.translate
 
         try:
             file_node = commit_obj.get_node(path, pre_load=pre_load)
             if file_node.is_dir():
                 raise RepositoryError('The given path is a directory')
         except CommitDoesNotExistError:
             log.exception('No such commit exists for this repository')
             h.flash(_('No such commit exists for this repository'), category='error')
             raise HTTPNotFound()
         except RepositoryError as e:
             log.warning('Repository error while fetching filenode `%s`. Err:%s', path, e)
             h.flash(h.escape(safe_str(e)), category='error')
             raise HTTPNotFound()
 
         return file_node
 
     def _is_valid_head(self, commit_id, repo, landing_ref):
         branch_name = sha_commit_id = ''
         is_head = False
         log.debug('Checking if commit_id `%s` is a head for %s.', commit_id, repo)
 
         for _branch_name, branch_commit_id in repo.branches.items():
             # simple case we pass in branch name, it's a HEAD
             if commit_id == _branch_name:
                 is_head = True
                 branch_name = _branch_name
                 sha_commit_id = branch_commit_id
                 break
             # case when we pass in full sha commit_id, which is a head
             elif commit_id == branch_commit_id:
                 is_head = True
                 branch_name = _branch_name
                 sha_commit_id = branch_commit_id
                 break
 
         if h.is_svn(repo) and not repo.is_empty():
             # Note: Subversion only has one head.
             if commit_id == repo.get_commit(commit_idx=-1).raw_id:
                 is_head = True
                 return branch_name, sha_commit_id, is_head
 
         # checked branches, means we only need to try to get the branch/commit_sha
         if repo.is_empty():
             is_head = True
             branch_name = landing_ref
             sha_commit_id = EmptyCommit().raw_id
         else:
             commit = repo.get_commit(commit_id=commit_id)
             if commit:
                 branch_name = commit.branch
                 sha_commit_id = commit.raw_id
 
         return branch_name, sha_commit_id, is_head
 
     def _get_tree_at_commit(self, c, commit_id, f_path, full_load=False, at_rev=None):
 
         repo_id = self.db_repo.repo_id
         force_recache = self.get_recache_flag()
 
         cache_seconds = safe_int(
             rhodecode.CONFIG.get('rc_cache.cache_repo.expiration_time'))
         cache_on = not force_recache and cache_seconds > 0
         log.debug(
             'Computing FILE TREE for repo_id %s commit_id `%s` and path `%s`'
             'with caching: %s[TTL: %ss]' % (
                 repo_id, commit_id, f_path, cache_on, cache_seconds or 0))
 
         cache_namespace_uid = f'repo.{rc_cache.FILE_TREE_CACHE_VER}.{repo_id}'
         region = rc_cache.get_or_create_region('cache_repo', cache_namespace_uid)
 
         @region.conditional_cache_on_arguments(namespace=cache_namespace_uid, condition=cache_on)
         def compute_file_tree(_name_hash, _repo_id, _commit_id, _f_path, _full_load, _at_rev):
             log.debug('Generating cached file tree at for repo_id: %s, %s, %s',
                       _repo_id, _commit_id, _f_path)
 
             c.full_load = _full_load
             return render(
                 'rhodecode:templates/files/files_browser_tree.mako',
                 self._get_template_context(c), self.request, _at_rev)
 
         return compute_file_tree(
             self.db_repo.repo_name_hash, self.db_repo.repo_id, commit_id, f_path, full_load, at_rev)
 
     def create_pure_path(self, *parts):
         # Split paths and sanitize them, removing any ../ etc
         sanitized_path = [
             x for x in pathlib.PurePath(*parts).parts
             if x not in ['.', '..']]
 
         pure_path = pathlib.PurePath(*sanitized_path)
         return pure_path
 
     def _is_lf_enabled(self, target_repo):
         lf_enabled = False
 
         lf_key_for_vcs_map = {
             'hg': 'extensions_largefiles',
             'git': 'vcs_git_lfs_enabled'
         }
 
         lf_key_for_vcs = lf_key_for_vcs_map.get(target_repo.repo_type)
 
         if lf_key_for_vcs:
             lf_enabled = self._get_repo_setting(target_repo, lf_key_for_vcs)
 
         return lf_enabled
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator(
         'repository.read', 'repository.write', 'repository.admin')
     def repo_archivefile(self):
         # archive cache config
         from rhodecode import CONFIG
         _ = self.request.translate
         self.load_default_context()
         default_at_path = '/'
         fname = self.request.matchdict['fname']
         subrepos = self.request.GET.get('subrepos') == 'true'
         with_hash = str2bool(self.request.GET.get('with_hash', '1'))
         at_path = self.request.GET.get('at_path') or default_at_path
 
         if not self.db_repo.enable_downloads:
             return Response(_('Downloads disabled'))
 
         try:
             commit_id, ext, fileformat, content_type = \
                 _get_archive_spec(fname)
         except ValueError:
             return Response(_('Unknown archive type for: `{}`').format(
                 h.escape(fname)))
 
         try:
             commit = self.rhodecode_vcs_repo.get_commit(commit_id)
         except CommitDoesNotExistError:
             return Response(_('Unknown commit_id {}').format(
                 h.escape(commit_id)))
         except EmptyRepositoryError:
             return Response(_('Empty repository'))
 
         # we used a ref, or a shorter version, lets redirect client ot use explicit hash
         if commit_id != commit.raw_id:
             fname=f'{commit.raw_id}{ext}'
             raise HTTPFound(self.request.current_route_path(fname=fname))
 
         try:
             at_path = commit.get_node(at_path).path or default_at_path
         except Exception:
             return Response(_('No node at path {} for this repository').format(h.escape(at_path)))
 
         path_sha = get_path_sha(at_path)
 
         # used for cache etc, consistent unique archive name
         archive_name_key = get_archive_name(
             self.db_repo.repo_id, self.db_repo_name, commit_sha=commit.short_id, ext=ext, subrepos=subrepos,
             path_sha=path_sha, with_hash=True)
 
         if not with_hash:
             path_sha = ''
 
         # what end client gets served
         response_archive_name = get_archive_name(
             self.db_repo.repo_id, self.db_repo_name, commit_sha=commit.short_id, ext=ext, subrepos=subrepos,
             path_sha=path_sha, with_hash=with_hash)
 
         # remove extension from our archive directory name
         archive_dir_name = response_archive_name[:-len(ext)]
 
         archive_cache_disable = self.request.GET.get('no_cache')
 
         d_cache = get_archival_cache_store(config=CONFIG)
 
         # NOTE: we get the config to pass to a call to lazy-init the SAME type of cache on vcsserver
         d_cache_conf = get_archival_config(config=CONFIG)
 
         # This is also a cache key, and lock key
         reentrant_lock_key = archive_name_key + '.lock'
 
         use_cached_archive = False
         if not archive_cache_disable and archive_name_key in d_cache:
             reader, metadata = d_cache.fetch(archive_name_key)
 
             use_cached_archive = True
             log.debug('Found cached archive as key=%s tag=%s, serving archive from cache reader=%s',
                       archive_name_key, metadata, reader.name)
         else:
             reader = None
             log.debug('Archive with key=%s is not yet cached, creating one now...', archive_name_key)
 
         if not reader:
             # generate new archive, as previous was not found in the cache
             try:
                 with d_cache.get_lock(reentrant_lock_key):
                     try:
                         commit.archive_repo(archive_name_key, archive_dir_name=archive_dir_name,
                                             kind=fileformat, subrepos=subrepos,
                                             archive_at_path=at_path, cache_config=d_cache_conf)
                     except ImproperArchiveTypeError:
                         return _('Unknown archive type')
 
             except ArchiveCacheGenerationLock:
                 retry_after = round(random.uniform(0.3, 3.0), 1)
                 time.sleep(retry_after)
 
                 location = self.request.url
                 response = Response(
                     f"archive {archive_name_key} generation in progress, Retry-After={retry_after}, Location={location}"
                 )
                 response.headers["Retry-After"] = str(retry_after)
                 response.status_code = 307  # temporary redirect
 
                 response.location = location
                 return response
 
-            reader, metadata = d_cache.fetch(archive_name_key)
+            reader, metadata = d_cache.fetch(archive_name_key, retry=True, retry_attempts=30)
 
         response = Response(app_iter=archive_iterator(reader))
         response.content_disposition = f'attachment; filename={response_archive_name}'
         response.content_type = str(content_type)
 
         try:
             return response
         finally:
             # store download action
             audit_logger.store_web(
                 'repo.archive.download', action_data={
                     'user_agent': self.request.user_agent,
                     'archive_name': archive_name_key,
                     'archive_spec': fname,
                     'archive_cached': use_cached_archive},
                 user=self._rhodecode_user,
                 repo=self.db_repo,
                 commit=True
             )
 
     def _get_file_node(self, commit_id, f_path):
         if commit_id not in ['', None, 'None', '0' * 12, '0' * 40]:
             commit = self.rhodecode_vcs_repo.get_commit(commit_id=commit_id)
             try:
                 node = commit.get_node(f_path)
                 if node.is_dir():
                     raise NodeError(f'{node} path is a {type(node)} not a file')
             except NodeDoesNotExistError:
                 commit = EmptyCommit(
                     commit_id=commit_id,
                     idx=commit.idx,
                     repo=commit.repository,
                     alias=commit.repository.alias,
                     message=commit.message,
                     author=commit.author,
                     date=commit.date)
                 node = FileNode(safe_bytes(f_path), b'', commit=commit)
         else:
             commit = EmptyCommit(
                 repo=self.rhodecode_vcs_repo,
                 alias=self.rhodecode_vcs_repo.alias)
             node = FileNode(safe_bytes(f_path), b'', commit=commit)
         return node
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator(
         'repository.read', 'repository.write', 'repository.admin')
     def repo_files_diff(self):
         c = self.load_default_context()
         f_path = self._get_f_path(self.request.matchdict)
         diff1 = self.request.GET.get('diff1', '')
         diff2 = self.request.GET.get('diff2', '')
 
         path1, diff1 = parse_path_ref(diff1, default_path=f_path)
 
         ignore_whitespace = str2bool(self.request.GET.get('ignorews'))
         line_context = self.request.GET.get('context', 3)
 
         if not any((diff1, diff2)):
             h.flash(
                 'Need query parameter "diff1" or "diff2" to generate a diff.',
                 category='error')
             raise HTTPBadRequest()
 
         c.action = self.request.GET.get('diff')
         if c.action not in ['download', 'raw']:
             compare_url = h.route_path(
                 'repo_compare',
                 repo_name=self.db_repo_name,
                 source_ref_type='rev',
                 source_ref=diff1,
                 target_repo=self.db_repo_name,
                 target_ref_type='rev',
                 target_ref=diff2,
                 _query=dict(f_path=f_path))
             # redirect to new view if we render diff
             raise HTTPFound(compare_url)
 
         try:
             node1 = self._get_file_node(diff1, path1)
             node2 = self._get_file_node(diff2, f_path)
         except (RepositoryError, NodeError):
             log.exception("Exception while trying to get node from repository")
             raise HTTPFound(
                 h.route_path('repo_files', repo_name=self.db_repo_name,
                              commit_id='tip', f_path=f_path))
 
         if all(isinstance(node.commit, EmptyCommit)
                for node in (node1, node2)):
             raise HTTPNotFound()
 
         c.commit_1 = node1.commit
         c.commit_2 = node2.commit
 
         if c.action == 'download':
             _diff = diffs.get_gitdiff(node1, node2,
                                       ignore_whitespace=ignore_whitespace,
                                       context=line_context)
             # NOTE: this was using diff_format='gitdiff'
             diff = diffs.DiffProcessor(_diff, diff_format='newdiff')
 
             response = Response(self.path_filter.get_raw_patch(diff))
             response.content_type = 'text/plain'
             response.content_disposition = (
                 f'attachment; filename={f_path}_{diff1}_vs_{diff2}.diff'
             )
             charset = self._get_default_encoding(c)
             if charset:
                 response.charset = charset
             return response
 
         elif c.action == 'raw':
             _diff = diffs.get_gitdiff(node1, node2,
                                       ignore_whitespace=ignore_whitespace,
                                       context=line_context)
             # NOTE: this was using diff_format='gitdiff'
             diff = diffs.DiffProcessor(_diff, diff_format='newdiff')
 
             response = Response(self.path_filter.get_raw_patch(diff))
             response.content_type = 'text/plain'
             charset = self._get_default_encoding(c)
             if charset:
                 response.charset = charset
             return response
 
         # in case we ever end up here
         raise HTTPNotFound()
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator(
         'repository.read', 'repository.write', 'repository.admin')
     def repo_files_diff_2way_redirect(self):
         """
         Kept only to make OLD links work
         """
         f_path = self._get_f_path_unchecked(self.request.matchdict)
         diff1 = self.request.GET.get('diff1', '')
         diff2 = self.request.GET.get('diff2', '')
 
         if not any((diff1, diff2)):
             h.flash(
                 'Need query parameter "diff1" or "diff2" to generate a diff.',
                 category='error')
             raise HTTPBadRequest()
 
         compare_url = h.route_path(
             'repo_compare',
             repo_name=self.db_repo_name,
             source_ref_type='rev',
             source_ref=diff1,
             target_ref_type='rev',
             target_ref=diff2,
             _query=dict(f_path=f_path, diffmode='sideside',
                         target_repo=self.db_repo_name,))
         raise HTTPFound(compare_url)
 
     @LoginRequired()
     def repo_files_default_commit_redirect(self):
         """
         Special page that redirects to the landing page of files based on the default
         commit for repository
         """
         c = self.load_default_context()
         ref_name = c.rhodecode_db_repo.landing_ref_name
         landing_url = h.repo_files_by_ref_url(
             c.rhodecode_db_repo.repo_name,
             c.rhodecode_db_repo.repo_type,
             f_path='',
             ref_name=ref_name,
             commit_id='tip',
             query=dict(at=ref_name)
         )
 
         raise HTTPFound(landing_url)
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator(
         'repository.read', 'repository.write', 'repository.admin')
     def repo_files(self):
         c = self.load_default_context()
 
         view_name = getattr(self.request.matched_route, 'name', None)
 
         c.annotate = view_name == 'repo_files:annotated'
         # default is false, but .rst/.md files later are auto rendered, we can
         # overwrite auto rendering by setting this GET flag
         c.renderer = view_name == 'repo_files:rendered' or not self.request.GET.get('no-render', False)
 
         commit_id, f_path = self._get_commit_and_path()
 
         c.commit = self._get_commit_or_redirect(commit_id)
         c.branch = self.request.GET.get('branch', None)
         c.f_path = f_path
         at_rev = self.request.GET.get('at')
 
         # files or dirs
         try:
             c.file = c.commit.get_node(f_path, pre_load=['is_binary', 'size', 'data'])
 
             c.file_author = True
             c.file_tree = ''
 
             # prev link
             try:
                 prev_commit = c.commit.prev(c.branch)
                 c.prev_commit = prev_commit
                 c.url_prev = h.route_path(
                     'repo_files', repo_name=self.db_repo_name,
                     commit_id=prev_commit.raw_id, f_path=f_path)
                 if c.branch:
                     c.url_prev += '?branch=%s' % c.branch
             except (CommitDoesNotExistError, VCSError):
                 c.url_prev = '#'
                 c.prev_commit = EmptyCommit()
 
             # next link
             try:
                 next_commit = c.commit.next(c.branch)
                 c.next_commit = next_commit
                 c.url_next = h.route_path(
                     'repo_files', repo_name=self.db_repo_name,
                     commit_id=next_commit.raw_id, f_path=f_path)
                 if c.branch:
                     c.url_next += '?branch=%s' % c.branch
             except (CommitDoesNotExistError, VCSError):
                 c.url_next = '#'
                 c.next_commit = EmptyCommit()
 
             # load file content
             if c.file.is_file():
 
                 c.lf_node = {}
 
                 has_lf_enabled = self._is_lf_enabled(self.db_repo)
                 if has_lf_enabled:
                     c.lf_node = c.file.get_largefile_node()
 
                 c.file_source_page = 'true'
                 c.file_last_commit = c.file.last_commit
 
                 c.file_size_too_big = c.file.size > c.visual.cut_off_limit_file
 
                 if not (c.file_size_too_big or c.file.is_binary):
                     if c.annotate:  # annotation has precedence over renderer
                         c.annotated_lines = filenode_as_annotated_lines_tokens(
                             c.file
                         )
                     else:
                         c.renderer = (
                             c.renderer and h.renderer_from_filename(c.file.path)
                         )
                         if not c.renderer:
                             c.lines = filenode_as_lines_tokens(c.file)
 
                 _branch_name, _sha_commit_id, is_head = \
                     self._is_valid_head(commit_id, self.rhodecode_vcs_repo,
                                         landing_ref=self.db_repo.landing_ref_name)
                 c.on_branch_head = is_head
 
                 branch = c.commit.branch if (
                     c.commit.branch and '/' not in c.commit.branch) else None
                 c.branch_or_raw_id = branch or c.commit.raw_id
                 c.branch_name = c.commit.branch or h.short_id(c.commit.raw_id)
 
                 author = c.file_last_commit.author
                 c.authors = [[
                     h.email(author),
                     h.person(author, 'username_or_name_or_email'),
                     1
                 ]]
 
             else:  # load tree content at path
                 c.file_source_page = 'false'
                 c.authors = []
                 # this loads a simple tree without metadata to speed things up
                 # later via ajax we call repo_nodetree_full and fetch whole
                 c.file_tree = self._get_tree_at_commit(c, c.commit.raw_id, f_path, at_rev=at_rev)
 
                 c.readme_data, c.readme_file = \
                     self._get_readme_data(self.db_repo, c.visual.default_renderer,
                                           c.commit.raw_id, f_path)
 
         except RepositoryError as e:
             h.flash(h.escape(safe_str(e)), category='error')
             raise HTTPNotFound()
 
         if self.request.environ.get('HTTP_X_PJAX'):
             html = render('rhodecode:templates/files/files_pjax.mako',
                           self._get_template_context(c), self.request)
         else:
             html = render('rhodecode:templates/files/files.mako',
                           self._get_template_context(c), self.request)
         return Response(html)
 
     @HasRepoPermissionAnyDecorator(
         'repository.read', 'repository.write', 'repository.admin')
     def repo_files_annotated_previous(self):
         self.load_default_context()
 
         commit_id, f_path = self._get_commit_and_path()
         commit = self._get_commit_or_redirect(commit_id)
         prev_commit_id = commit.raw_id
         line_anchor = self.request.GET.get('line_anchor')
         is_file = False
         try:
             _file = commit.get_node(f_path)
             is_file = _file.is_file()
         except (NodeDoesNotExistError, CommitDoesNotExistError, VCSError):
             pass
 
         if is_file:
             history = commit.get_path_history(f_path)
             prev_commit_id = history[1].raw_id \
                 if len(history) > 1 else prev_commit_id
         prev_url = h.route_path(
             'repo_files:annotated', repo_name=self.db_repo_name,
             commit_id=prev_commit_id, f_path=f_path,
             _anchor=f'L{line_anchor}')
 
         raise HTTPFound(prev_url)
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator(
         'repository.read', 'repository.write', 'repository.admin')
     def repo_nodetree_full(self):
         """
         Returns rendered html of file tree that contains commit date,
         author, commit_id for the specified combination of
         repo, commit_id and file path
         """
         c = self.load_default_context()
 
         commit_id, f_path = self._get_commit_and_path()
         commit = self._get_commit_or_redirect(commit_id)
         try:
             dir_node = commit.get_node(f_path)
         except RepositoryError as e:
             return Response(f'error: {h.escape(safe_str(e))}')
 
         if dir_node.is_file():
             return Response('')
 
         c.file = dir_node
         c.commit = commit
         at_rev = self.request.GET.get('at')
 
         html = self._get_tree_at_commit(
             c, commit.raw_id, dir_node.path, full_load=True, at_rev=at_rev)
 
         return Response(html)
 
     def _get_attachement_headers(self, f_path):
         f_name = safe_str(f_path.split(Repository.NAME_SEP)[-1])
         safe_path = f_name.replace('"', '\\"')
         encoded_path = urllib.parse.quote(f_name)
 
         headers = "attachment; " \
                   "filename=\"{}\"; " \
                   "filename*=UTF-8\'\'{}".format(safe_path, encoded_path)
 
         return safe_bytes(headers).decode('latin-1', errors='replace')
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator(
         'repository.read', 'repository.write', 'repository.admin')
     def repo_file_raw(self):
         """
         Action for show as raw, some mimetypes are "rendered",
         those include images, icons.
         """
         c = self.load_default_context()
 
         commit_id, f_path = self._get_commit_and_path()
         commit = self._get_commit_or_redirect(commit_id)
         file_node = self._get_filenode_or_redirect(commit, f_path)
 
         raw_mimetype_mapping = {
             # map original mimetype to a mimetype used for "show as raw"
             # you can also provide a content-disposition to override the
             # default "attachment" disposition.
             # orig_type: (new_type, new_dispo)
 
             # show images inline:
             # Do not re-add SVG: it is unsafe and permits XSS attacks. One can
             # for example render an SVG with javascript inside or even render
             # HTML.
             'image/x-icon': ('image/x-icon', 'inline'),
             'image/png': ('image/png', 'inline'),
             'image/gif': ('image/gif', 'inline'),
             'image/jpeg': ('image/jpeg', 'inline'),
             'application/pdf': ('application/pdf', 'inline'),
         }
 
         mimetype = file_node.mimetype
         try:
             mimetype, disposition = raw_mimetype_mapping[mimetype]
         except KeyError:
             # we don't know anything special about this, handle it safely
             if file_node.is_binary:
                 # do same as download raw for binary files
                 mimetype, disposition = 'application/octet-stream', 'attachment'
             else:
                 # do not just use the original mimetype, but force text/plain,
                 # otherwise it would serve text/html and that might be unsafe.
                 # Note: underlying vcs library fakes text/plain mimetype if the
                 # mimetype can not be determined and it thinks it is not
                 # binary.This might lead to erroneous text display in some
                 # cases, but helps in other cases, like with text files
                 # without extension.
                 mimetype, disposition = 'text/plain', 'inline'
 
         if disposition == 'attachment':
             disposition = self._get_attachement_headers(f_path)
 
         stream_content = file_node.stream_bytes()
 
         response = Response(app_iter=stream_content)
         response.content_disposition = disposition
         response.content_type = mimetype
 
         charset = self._get_default_encoding(c)
         if charset:
             response.charset = charset
 
         return response
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator(
         'repository.read', 'repository.write', 'repository.admin')
     def repo_file_download(self):
         c = self.load_default_context()
 
         commit_id, f_path = self._get_commit_and_path()
         commit = self._get_commit_or_redirect(commit_id)
         file_node = self._get_filenode_or_redirect(commit, f_path)
 
         if self.request.GET.get('lf'):
             # only if lf get flag is passed, we download this file
             # as LFS/Largefile
             lf_node = file_node.get_largefile_node()
             if lf_node:
                 # overwrite our pointer with the REAL large-file
                 file_node = lf_node
 
         disposition = self._get_attachement_headers(f_path)
 
         stream_content = file_node.stream_bytes()
 
         response = Response(app_iter=stream_content)
         response.content_disposition = disposition
         response.content_type = file_node.mimetype
 
         charset = self._get_default_encoding(c)
         if charset:
             response.charset = charset
 
         return response
 
     def _get_nodelist_at_commit(self, repo_name, repo_id, commit_id, f_path):
 
         cache_seconds = safe_int(
             rhodecode.CONFIG.get('rc_cache.cache_repo.expiration_time'))
         cache_on = cache_seconds > 0
         log.debug(
             'Computing FILE SEARCH for repo_id %s commit_id `%s` and path `%s`'
             'with caching: %s[TTL: %ss]' % (
                 repo_id, commit_id, f_path, cache_on, cache_seconds or 0))
 
         cache_namespace_uid = f'repo.{repo_id}'
         region = rc_cache.get_or_create_region('cache_repo', cache_namespace_uid)
 
         @region.conditional_cache_on_arguments(namespace=cache_namespace_uid, condition=cache_on)
         def compute_file_search(_name_hash, _repo_id, _commit_id, _f_path):
             log.debug('Generating cached nodelist for repo_id:%s, %s, %s',
                       _repo_id, commit_id, f_path)
             try:
                 _d, _f = ScmModel().get_quick_filter_nodes(repo_name, _commit_id, _f_path)
             except (RepositoryError, CommitDoesNotExistError, Exception) as e:
                 log.exception(safe_str(e))
                 h.flash(h.escape(safe_str(e)), category='error')
                 raise HTTPFound(h.route_path(
                     'repo_files', repo_name=self.db_repo_name,
                     commit_id='tip', f_path='/'))
 
             return _d + _f
 
         result = compute_file_search(self.db_repo.repo_name_hash, self.db_repo.repo_id,
                                      commit_id, f_path)
         return filter(lambda n: self.path_filter.path_access_allowed(n['name']), result)
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator(
         'repository.read', 'repository.write', 'repository.admin')
     def repo_nodelist(self):
         self.load_default_context()
 
         commit_id, f_path = self._get_commit_and_path()
         commit = self._get_commit_or_redirect(commit_id)
 
         metadata = self._get_nodelist_at_commit(
             self.db_repo_name, self.db_repo.repo_id, commit.raw_id, f_path)
         return {'nodes': [x for x in metadata]}
 
     def _create_references(self, branches_or_tags, symbolic_reference, f_path, ref_type):
         items = []
         for name, commit_id in branches_or_tags.items():
             sym_ref = symbolic_reference(commit_id, name, f_path, ref_type)
             items.append((sym_ref, name, ref_type))
         return items
 
     def _symbolic_reference(self, commit_id, name, f_path, ref_type):
         return commit_id
 
     def _symbolic_reference_svn(self, commit_id, name, f_path, ref_type):
         return commit_id
 
         # NOTE(dan): old code we used in "diff" mode compare
         new_f_path = vcspath.join(name, f_path)
         return f'{new_f_path}@{commit_id}'
 
     def _get_node_history(self, commit_obj, f_path, commits=None):
         """
         get commit history for given node
 
         :param commit_obj: commit to calculate history
         :param f_path: path for node to calculate history for
         :param commits: if passed don't calculate history and take
             commits defined in this list
         """
         _ = self.request.translate
 
         # calculate history based on tip
         tip = self.rhodecode_vcs_repo.get_commit()
         if commits is None:
             pre_load = ["author", "branch"]
             try:
                 commits = tip.get_path_history(f_path, pre_load=pre_load)
             except (NodeDoesNotExistError, CommitError):
                 # this node is not present at tip!
                 commits = commit_obj.get_path_history(f_path, pre_load=pre_load)
 
         history = []
         commits_group = ([], _("Changesets"))
         for commit in commits:
             branch = ' (%s)' % commit.branch if commit.branch else ''
             n_desc = f'r{commit.idx}:{commit.short_id}{branch}'
             commits_group[0].append((commit.raw_id, n_desc, 'sha'))
         history.append(commits_group)
 
         symbolic_reference = self._symbolic_reference
 
         if self.rhodecode_vcs_repo.alias == 'svn':
             adjusted_f_path = RepoFilesView.adjust_file_path_for_svn(
                 f_path, self.rhodecode_vcs_repo)
             if adjusted_f_path != f_path:
                 log.debug(
                     'Recognized svn tag or branch in file "%s", using svn '
                     'specific symbolic references', f_path)
                 f_path = adjusted_f_path
                 symbolic_reference = self._symbolic_reference_svn
 
         branches = self._create_references(
             self.rhodecode_vcs_repo.branches, symbolic_reference, f_path, 'branch')
         branches_group = (branches, _("Branches"))
 
         tags = self._create_references(
             self.rhodecode_vcs_repo.tags, symbolic_reference, f_path, 'tag')
         tags_group = (tags, _("Tags"))
 
         history.append(branches_group)
         history.append(tags_group)
 
         return history, commits
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator(
         'repository.read', 'repository.write', 'repository.admin')
     def repo_file_history(self):
         self.load_default_context()
 
         commit_id, f_path = self._get_commit_and_path()
         commit = self._get_commit_or_redirect(commit_id)
         file_node = self._get_filenode_or_redirect(commit, f_path)
 
         if file_node.is_file():
             file_history, _hist = self._get_node_history(commit, f_path)
 
             res = []
             for section_items, section in file_history:
                 items = []
                 for obj_id, obj_text, obj_type in section_items:
                     at_rev = ''
                     if obj_type in ['branch', 'bookmark', 'tag']:
                         at_rev = obj_text
                     entry = {
                         'id': obj_id,
                         'text': obj_text,
                         'type': obj_type,
                         'at_rev': at_rev
                     }
 
                     items.append(entry)
 
                 res.append({
                     'text': section,
                     'children': items
                 })
 
             data = {
                 'more': False,
                 'results': res
             }
             return data
 
         log.warning('Cannot fetch history for directory')
         raise HTTPBadRequest()
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator(
         'repository.read', 'repository.write', 'repository.admin')
     def repo_file_authors(self):
         c = self.load_default_context()
 
         commit_id, f_path = self._get_commit_and_path()
         commit = self._get_commit_or_redirect(commit_id)
         file_node = self._get_filenode_or_redirect(commit, f_path)
 
         if not file_node.is_file():
             raise HTTPBadRequest()
 
         c.file_last_commit = file_node.last_commit
         if self.request.GET.get('annotate') == '1':
             # use _hist from annotation if annotation mode is on
             commit_ids = {x[1] for x in file_node.annotate}
             _hist = (
                 self.rhodecode_vcs_repo.get_commit(commit_id)
                 for commit_id in commit_ids)
         else:
             _f_history, _hist = self._get_node_history(commit, f_path)
         c.file_author = False
 
         unique = collections.OrderedDict()
         for commit in _hist:
             author = commit.author
             if author not in unique:
                 unique[commit.author] = [
                     h.email(author),
                     h.person(author, 'username_or_name_or_email'),
                     1  # counter
                 ]
 
             else:
                 # increase counter
                 unique[commit.author][2] += 1
 
         c.authors = [val for val in unique.values()]
 
         return self._get_template_context(c)
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator('repository.write', 'repository.admin')
     def repo_files_check_head(self):
         self.load_default_context()
 
         commit_id, f_path = self._get_commit_and_path()
         _branch_name, _sha_commit_id, is_head = \
             self._is_valid_head(commit_id, self.rhodecode_vcs_repo,
                                 landing_ref=self.db_repo.landing_ref_name)
 
         new_path = self.request.POST.get('path')
         operation = self.request.POST.get('operation')
         path_exist = ''
 
         if new_path and operation in ['create', 'upload']:
             new_f_path = os.path.join(f_path.lstrip('/'), new_path)
             try:
                 commit_obj = self.rhodecode_vcs_repo.get_commit(commit_id)
                 # NOTE(dan): construct whole path without leading /
                 file_node = commit_obj.get_node(new_f_path)
                 if file_node is not None:
                     path_exist = new_f_path
             except EmptyRepositoryError:
                 pass
             except Exception:
                 pass
 
         return {
             'branch': _branch_name,
             'sha': _sha_commit_id,
             'is_head': is_head,
             'path_exists': path_exist
         }
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator('repository.write', 'repository.admin')
     def repo_files_remove_file(self):
         _ = self.request.translate
         c = self.load_default_context()
         commit_id, f_path = self._get_commit_and_path()
 
         self._ensure_not_locked()
         _branch_name, _sha_commit_id, is_head = \
             self._is_valid_head(commit_id, self.rhodecode_vcs_repo,
                                 landing_ref=self.db_repo.landing_ref_name)
 
         self.forbid_non_head(is_head, f_path)
         self.check_branch_permission(_branch_name)
 
         c.commit = self._get_commit_or_redirect(commit_id)
         c.file = self._get_filenode_or_redirect(c.commit, f_path)
 
         c.default_message = _(
             'Deleted file {} via RhodeCode Enterprise').format(f_path)
         c.f_path = f_path
 
         return self._get_template_context(c)
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator('repository.write', 'repository.admin')
     @CSRFRequired()
     def repo_files_delete_file(self):
         _ = self.request.translate
 
         c = self.load_default_context()
         commit_id, f_path = self._get_commit_and_path()
 
         self._ensure_not_locked()
         _branch_name, _sha_commit_id, is_head = \
             self._is_valid_head(commit_id, self.rhodecode_vcs_repo,
                                 landing_ref=self.db_repo.landing_ref_name)
 
         self.forbid_non_head(is_head, f_path)
         self.check_branch_permission(_branch_name)
 
         c.commit = self._get_commit_or_redirect(commit_id)
         c.file = self._get_filenode_or_redirect(c.commit, f_path)
 
         c.default_message = _(
             'Deleted file {} via RhodeCode Enterprise').format(f_path)
         c.f_path = f_path
         node_path = f_path
         author = self._rhodecode_db_user.full_contact
         message = self.request.POST.get('message') or c.default_message
         try:
             nodes = {
                 safe_bytes(node_path): {
                     'content': b''
                 }
             }
             ScmModel().delete_nodes(
                 user=self._rhodecode_db_user.user_id, repo=self.db_repo,
                 message=message,
                 nodes=nodes,
                 parent_commit=c.commit,
                 author=author,
             )
 
             h.flash(
                 _('Successfully deleted file `{}`').format(
                     h.escape(f_path)), category='success')
         except Exception:
             log.exception('Error during commit operation')
             h.flash(_('Error occurred during commit'), category='error')
         raise HTTPFound(
             h.route_path('repo_commit', repo_name=self.db_repo_name,
                          commit_id='tip'))
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator('repository.write', 'repository.admin')
     def repo_files_edit_file(self):
         _ = self.request.translate
         c = self.load_default_context()
         commit_id, f_path = self._get_commit_and_path()
 
         self._ensure_not_locked()
         _branch_name, _sha_commit_id, is_head = \
             self._is_valid_head(commit_id, self.rhodecode_vcs_repo,
                                 landing_ref=self.db_repo.landing_ref_name)
 
         self.forbid_non_head(is_head, f_path, commit_id=commit_id)
         self.check_branch_permission(_branch_name, commit_id=commit_id)
 
         c.commit = self._get_commit_or_redirect(commit_id)
         c.file = self._get_filenode_or_redirect(c.commit, f_path)
 
         if c.file.is_binary:
             files_url = h.route_path(
                 'repo_files',
                 repo_name=self.db_repo_name,
                 commit_id=c.commit.raw_id, f_path=f_path)
             raise HTTPFound(files_url)
 
         c.default_message = _('Edited file {} via RhodeCode Enterprise').format(f_path)
         c.f_path = f_path
 
         return self._get_template_context(c)
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator('repository.write', 'repository.admin')
     @CSRFRequired()
     def repo_files_update_file(self):
         _ = self.request.translate
         c = self.load_default_context()
         commit_id, f_path = self._get_commit_and_path()
 
         self._ensure_not_locked()
 
         c.commit = self._get_commit_or_redirect(commit_id)
         c.file = self._get_filenode_or_redirect(c.commit, f_path)
 
         if c.file.is_binary:
             raise HTTPFound(h.route_path('repo_files', repo_name=self.db_repo_name,
                                          commit_id=c.commit.raw_id, f_path=f_path))
 
         _branch_name, _sha_commit_id, is_head = \
             self._is_valid_head(commit_id, self.rhodecode_vcs_repo,
                                 landing_ref=self.db_repo.landing_ref_name)
 
         self.forbid_non_head(is_head, f_path, commit_id=commit_id)
         self.check_branch_permission(_branch_name, commit_id=commit_id)
 
         c.default_message = _('Edited file {} via RhodeCode Enterprise').format(f_path)
         c.f_path = f_path
 
         old_content = c.file.str_content
         sl = old_content.splitlines(1)
         first_line = sl[0] if sl else ''
 
         r_post = self.request.POST
         # line endings:  0 - Unix, 1 - Mac, 2 - DOS
         line_ending_mode = detect_mode(first_line, 0)
         content = convert_line_endings(r_post.get('content', ''), line_ending_mode)
 
         message = r_post.get('message') or c.default_message
 
         org_node_path = c.file.str_path
         filename = r_post['filename']
 
         root_path = c.file.dir_path
         pure_path = self.create_pure_path(root_path, filename)
         node_path = pure_path.as_posix()
 
         default_redirect_url = h.route_path('repo_commit', repo_name=self.db_repo_name,
                                             commit_id=commit_id)
         if content == old_content and node_path == org_node_path:
             h.flash(_('No changes detected on {}').format(h.escape(org_node_path)),
                     category='warning')
             raise HTTPFound(default_redirect_url)
 
         try:
             mapping = {
                 c.file.bytes_path: {
                     'org_filename': org_node_path,
                     'filename': safe_bytes(node_path),
                     'content': safe_bytes(content),
                     'lexer': '',
                     'op': 'mod',
                     'mode': c.file.mode
                 }
             }
 
             commit = ScmModel().update_nodes(
                 user=self._rhodecode_db_user.user_id,
                 repo=self.db_repo,
                 message=message,
                 nodes=mapping,
                 parent_commit=c.commit,
             )
 
             h.flash(_('Successfully committed changes to file `{}`').format(
                     h.escape(f_path)), category='success')
             default_redirect_url = h.route_path(
                 'repo_commit', repo_name=self.db_repo_name, commit_id=commit.raw_id)
 
         except Exception:
             log.exception('Error occurred during commit')
             h.flash(_('Error occurred during commit'), category='error')
 
         raise HTTPFound(default_redirect_url)
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator('repository.write', 'repository.admin')
     def repo_files_add_file(self):
         _ = self.request.translate
         c = self.load_default_context()
         commit_id, f_path = self._get_commit_and_path()
 
         self._ensure_not_locked()
 
         # Check if we need to use this page to upload binary
         upload_binary = str2bool(self.request.params.get('upload_binary', False))
 
         c.commit = self._get_commit_or_redirect(commit_id, redirect_after=False)
         if c.commit is None:
             c.commit = EmptyCommit(alias=self.rhodecode_vcs_repo.alias)
 
         if self.rhodecode_vcs_repo.is_empty():
             # for empty repository we cannot check for current branch, we rely on
             # c.commit.branch instead
             _branch_name, _sha_commit_id, is_head = c.commit.branch, '', True
         else:
             _branch_name, _sha_commit_id, is_head = \
                 self._is_valid_head(commit_id, self.rhodecode_vcs_repo,
                                     landing_ref=self.db_repo.landing_ref_name)
 
         self.forbid_non_head(is_head, f_path, commit_id=commit_id)
         self.check_branch_permission(_branch_name, commit_id=commit_id)
 
         c.default_message = (_('Added file via RhodeCode Enterprise')) \
             if not upload_binary else (_('Edited file {} via RhodeCode Enterprise').format(f_path))
         c.f_path = f_path.lstrip('/')  # ensure not relative path
         c.replace_binary = upload_binary
 
         return self._get_template_context(c)
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator('repository.write', 'repository.admin')
     @CSRFRequired()
     def repo_files_create_file(self):
         _ = self.request.translate
         c = self.load_default_context()
         commit_id, f_path = self._get_commit_and_path()
 
         self._ensure_not_locked()
 
         c.commit = self._get_commit_or_redirect(commit_id, redirect_after=False)
         if c.commit is None:
             c.commit = EmptyCommit(alias=self.rhodecode_vcs_repo.alias)
 
         # calculate redirect URL
         if self.rhodecode_vcs_repo.is_empty():
             default_redirect_url = h.route_path(
                 'repo_summary', repo_name=self.db_repo_name)
         else:
             default_redirect_url = h.route_path(
                 'repo_commit', repo_name=self.db_repo_name, commit_id='tip')
 
         if self.rhodecode_vcs_repo.is_empty():
             # for empty repository we cannot check for current branch, we rely on
             # c.commit.branch instead
             _branch_name, _sha_commit_id, is_head = c.commit.branch, '', True
         else:
             _branch_name, _sha_commit_id, is_head = \
                 self._is_valid_head(commit_id, self.rhodecode_vcs_repo,
                                     landing_ref=self.db_repo.landing_ref_name)
 
         self.forbid_non_head(is_head, f_path, commit_id=commit_id)
         self.check_branch_permission(_branch_name, commit_id=commit_id)
 
         c.default_message = (_('Added file via RhodeCode Enterprise'))
         c.f_path = f_path
 
         r_post = self.request.POST
         message = r_post.get('message') or c.default_message
         filename = r_post.get('filename')
         unix_mode = 0
 
         if not filename:
             # If there's no commit, redirect to repo summary
             if type(c.commit) is EmptyCommit:
                 redirect_url = h.route_path(
                     'repo_summary', repo_name=self.db_repo_name)
             else:
                 redirect_url = default_redirect_url
             h.flash(_('No filename specified'), category='warning')
             raise HTTPFound(redirect_url)
 
         root_path = f_path
         pure_path = self.create_pure_path(root_path, filename)
         node_path = pure_path.as_posix().lstrip('/')
 
         author = self._rhodecode_db_user.full_contact
         content = convert_line_endings(r_post.get('content', ''), unix_mode)
         nodes = {
             safe_bytes(node_path): {
                 'content': safe_bytes(content)
             }
         }
 
         try:
 
             commit = ScmModel().create_nodes(
                 user=self._rhodecode_db_user.user_id,
                 repo=self.db_repo,
                 message=message,
                 nodes=nodes,
                 parent_commit=c.commit,
                 author=author,
             )
 
             h.flash(_('Successfully committed new file `{}`').format(
                     h.escape(node_path)), category='success')
 
             default_redirect_url = h.route_path(
                 'repo_commit', repo_name=self.db_repo_name, commit_id=commit.raw_id)
 
         except NonRelativePathError:
             log.exception('Non Relative path found')
             h.flash(_('The location specified must be a relative path and must not '
                       'contain .. in the path'), category='warning')
             raise HTTPFound(default_redirect_url)
         except (NodeError, NodeAlreadyExistsError) as e:
             h.flash(h.escape(safe_str(e)), category='error')
         except Exception:
             log.exception('Error occurred during commit')
             h.flash(_('Error occurred during commit'), category='error')
 
         raise HTTPFound(default_redirect_url)
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator('repository.write', 'repository.admin')
     @CSRFRequired()
     def repo_files_upload_file(self):
         _ = self.request.translate
         c = self.load_default_context()
         commit_id, f_path = self._get_commit_and_path()
 
         self._ensure_not_locked()
 
         c.commit = self._get_commit_or_redirect(commit_id, redirect_after=False)
         if c.commit is None:
             c.commit = EmptyCommit(alias=self.rhodecode_vcs_repo.alias)
 
         # calculate redirect URL
         if self.rhodecode_vcs_repo.is_empty():
             default_redirect_url = h.route_path(
                 'repo_summary', repo_name=self.db_repo_name)
         else:
             default_redirect_url = h.route_path(
                 'repo_commit', repo_name=self.db_repo_name, commit_id='tip')
 
         if self.rhodecode_vcs_repo.is_empty():
             # for empty repository we cannot check for current branch, we rely on
             # c.commit.branch instead
             _branch_name, _sha_commit_id, is_head = c.commit.branch, '', True
         else:
             _branch_name, _sha_commit_id, is_head = \
                 self._is_valid_head(commit_id, self.rhodecode_vcs_repo,
                                     landing_ref=self.db_repo.landing_ref_name)
 
         error = self.forbid_non_head(is_head, f_path, json_mode=True)
         if error:
             return {
                 'error': error,
                 'redirect_url': default_redirect_url
             }
         error = self.check_branch_permission(_branch_name, json_mode=True)
         if error:
             return {
                 'error': error,
                 'redirect_url': default_redirect_url
             }
 
         c.default_message = (_('Added file via RhodeCode Enterprise'))
         c.f_path = f_path
 
         r_post = self.request.POST
 
         message = c.default_message
         user_message = r_post.getall('message')
         if isinstance(user_message, list) and user_message:
             # we take the first from duplicated results if it's not empty
             message = user_message[0] if user_message[0] else message
 
         nodes = {}
 
         for file_obj in r_post.getall('files_upload') or []:
             content = file_obj.file
             filename = file_obj.filename
 
             root_path = f_path
             pure_path = self.create_pure_path(root_path, filename)
             node_path = pure_path.as_posix().lstrip('/')
 
             nodes[safe_bytes(node_path)] = {
                 'content': content
             }
 
         if not nodes:
             error = 'missing files'
             return {
                 'error': error,
                 'redirect_url': default_redirect_url
             }
 
         author = self._rhodecode_db_user.full_contact
 
         try:
             commit = ScmModel().create_nodes(
                 user=self._rhodecode_db_user.user_id,
                 repo=self.db_repo,
                 message=message,
                 nodes=nodes,
                 parent_commit=c.commit,
                 author=author,
             )
             if len(nodes) == 1:
                 flash_message = _('Successfully committed {} new files').format(len(nodes))
             else:
                 flash_message = _('Successfully committed 1 new file')
 
             h.flash(flash_message, category='success')
 
             default_redirect_url = h.route_path(
                 'repo_commit', repo_name=self.db_repo_name, commit_id=commit.raw_id)
 
         except NonRelativePathError:
             log.exception('Non Relative path found')
             error = _('The location specified must be a relative path and must not '
                       'contain .. in the path')
             h.flash(error, category='warning')
 
             return {
                 'error': error,
                 'redirect_url': default_redirect_url
             }
         except (NodeError, NodeAlreadyExistsError) as e:
             error = h.escape(e)
             h.flash(error, category='error')
 
             return {
                 'error': error,
                 'redirect_url': default_redirect_url
             }
         except Exception:
             log.exception('Error occurred during commit')
             error = _('Error occurred during commit')
             h.flash(error, category='error')
             return {
                 'error': error,
                 'redirect_url': default_redirect_url
             }
 
         return {
             'error': None,
             'redirect_url': default_redirect_url
         }
 
     @LoginRequired()
     @HasRepoPermissionAnyDecorator('repository.write', 'repository.admin')
     @CSRFRequired()
     def repo_files_replace_file(self):
         _ = self.request.translate
         c = self.load_default_context()
         commit_id, f_path = self._get_commit_and_path()
 
         self._ensure_not_locked()
 
         c.commit = self._get_commit_or_redirect(commit_id, redirect_after=False)
         if c.commit is None:
             c.commit = EmptyCommit(alias=self.rhodecode_vcs_repo.alias)
 
         if self.rhodecode_vcs_repo.is_empty():
             default_redirect_url = h.route_path(
                 'repo_summary', repo_name=self.db_repo_name)
         else:
             default_redirect_url = h.route_path(
                 'repo_commit', repo_name=self.db_repo_name, commit_id='tip')
 
         if self.rhodecode_vcs_repo.is_empty():
             # for empty repository we cannot check for current branch, we rely on
             # c.commit.branch instead
             _branch_name, _sha_commit_id, is_head = c.commit.branch, '', True
         else:
             _branch_name, _sha_commit_id, is_head = \
                 self._is_valid_head(commit_id, self.rhodecode_vcs_repo,
                                     landing_ref=self.db_repo.landing_ref_name)
 
         error = self.forbid_non_head(is_head, f_path, json_mode=True)
         if error:
             return {
                 'error': error,
                 'redirect_url': default_redirect_url
             }
         error = self.check_branch_permission(_branch_name, json_mode=True)
         if error:
             return {
                 'error': error,
                 'redirect_url': default_redirect_url
             }
 
         c.default_message = (_('Edited file {} via RhodeCode Enterprise').format(f_path))
         c.f_path = f_path
 
         r_post = self.request.POST
 
         message = c.default_message
         user_message = r_post.getall('message')
         if isinstance(user_message, list) and user_message:
             # we take the first from duplicated results if it's not empty
             message = user_message[0] if user_message[0] else message
 
         data_for_replacement = r_post.getall('files_upload') or []
         if (objects_count := len(data_for_replacement)) > 1:
             return {
                 'error': 'too many files for replacement',
                 'redirect_url': default_redirect_url
             }
         elif not objects_count:
             return {
                 'error': 'missing files',
                 'redirect_url': default_redirect_url
             }
 
         content = data_for_replacement[0].file
         retrieved_filename = data_for_replacement[0].filename
 
         if retrieved_filename.split('.')[-1] != f_path.split('.')[-1]:
             return {
                 'error': 'file extension of uploaded file doesn\'t match an original file\'s extension',
                 'redirect_url': default_redirect_url
             }
 
         author = self._rhodecode_db_user.full_contact
 
         try:
             commit = ScmModel().update_binary_node(
                 user=self._rhodecode_db_user.user_id,
                 repo=self.db_repo,
                 message=message,
                 node={
                     'content': content,
                     'file_path': f_path.encode(),
                 },
                 parent_commit=c.commit,
                 author=author,
             )
 
             h.flash(_('Successfully committed 1 new file'), category='success')
 
             default_redirect_url = h.route_path(
                 'repo_commit', repo_name=self.db_repo_name, commit_id=commit.raw_id)
 
         except (NodeError, NodeAlreadyExistsError) as e:
             error = h.escape(e)
             h.flash(error, category='error')
 
             return {
                 'error': error,
                 'redirect_url': default_redirect_url
             }
         except Exception:
             log.exception('Error occurred during commit')
             error = _('Error occurred during commit')
             h.flash(error, category='error')
             return {
                 'error': error,
                 'redirect_url': default_redirect_url
             }
 
         return {
             'error': None,
             'redirect_url': default_redirect_url
         }

rhodecode/lib/rc_cache/archive_cache/fanout_cache.py

@@ -1,449 +1,456 b''
 # Copyright (C) 2015-2024 RhodeCode GmbH
 #
 # This program is free software: you can redistribute it and/or modify
 # it under the terms of the GNU Affero General Public License, version 3
 # (only), as published by the Free Software Foundation.
 #
 # This program is distributed in the hope that it will be useful,
 # but WITHOUT ANY WARRANTY; without even the implied warranty of
 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 # GNU General Public License for more details.
 #
 # You should have received a copy of the GNU Affero General Public License
 # along with this program.  If not, see <http://www.gnu.org/licenses/>.
 #
 # This program is dual-licensed. If you wish to learn more about the
 # RhodeCode Enterprise Edition, including its added features, Support services,
 # and proprietary license terms, please see https://rhodecode.com/licenses/
 
 import codecs
 import contextlib
 import functools
 import os
 import logging
 import time
 import typing
 import zlib
 import sqlite3
 
 from ...ext_json import json
 from .lock import GenerationLock
 from .utils import format_size
 
 log = logging.getLogger(__name__)
 
 cache_meta = None
 
 UNKNOWN = -241
 NO_VAL = -917
 
 MODE_BINARY = 'BINARY'
 
 
 EVICTION_POLICY = {
     'none': {
         'evict': None,
     },
     'least-recently-stored': {
         'evict': 'SELECT {fields} FROM archive_cache ORDER BY store_time',
     },
     'least-recently-used': {
         'evict': 'SELECT {fields} FROM archive_cache ORDER BY access_time',
     },
     'least-frequently-used': {
         'evict': 'SELECT {fields} FROM archive_cache ORDER BY access_count',
     },
 }
 
 
 class DB:
 
     def __init__(self):
         self.connection = sqlite3.connect(':memory:')
         self._init_db()
 
     def _init_db(self):
         qry = '''
             CREATE TABLE IF NOT EXISTS archive_cache (
              rowid INTEGER PRIMARY KEY,
              key_file TEXT,
              key_file_path TEXT,
              filename TEXT,
              full_path TEXT,
              store_time REAL,
              access_time REAL,
              access_count INTEGER DEFAULT 0,
              size INTEGER DEFAULT 0
              )
         '''
 
         self.sql(qry)
         self.connection.commit()
 
     @property
     def sql(self):
         return self.connection.execute
 
     def bulk_insert(self, rows):
         qry = '''
             INSERT INTO archive_cache (
              rowid,
              key_file,
              key_file_path,
              filename,
              full_path,
              store_time,
              access_time,
              access_count,
              size
             )
             VALUES (
             ?, ?, ?, ?, ?, ?, ?, ?, ?
             )
         '''
         cursor = self.connection.cursor()
         cursor.executemany(qry, rows)
         self.connection.commit()
 
 
 class FileSystemCache:
 
     def __init__(self, index, directory, **settings):
         self._index = index
         self._directory = directory
 
     @property
     def directory(self):
         """Cache directory."""
         return self._directory
 
     def _write_file(self, full_path, iterator, mode, encoding=None):
         full_dir, _ = os.path.split(full_path)
 
         for count in range(1, 11):
             with contextlib.suppress(OSError):
                 os.makedirs(full_dir)
 
             try:
                 # Another cache may have deleted the directory before
                 # the file could be opened.
                 writer = open(full_path, mode, encoding=encoding)
             except OSError:
                 if count == 10:
                     # Give up after 10 tries to open the file.
                     raise
                 continue
 
             with writer:
                 size = 0
                 for chunk in iterator:
                     size += len(chunk)
                     writer.write(chunk)
+                writer.flush()
+                # Get the file descriptor
+                fd = writer.fileno()
+
+                # Sync the file descriptor to disk, helps with NFS cases...
+                os.fsync(fd)
+                log.debug('written new archive cache under %s', full_path)
                 return size
 
     def _get_keyfile(self, key):
         return os.path.join(self._directory, f'{key}.key')
 
     def store(self, key, value_reader, metadata):
         filename, full_path = self.random_filename()
         key_file = self._get_keyfile(key)
 
         # STORE METADATA
         _metadata = {
             "version": "v1",
             "filename": filename,
             "full_path": full_path,
             "key_file": key_file,
             "store_time": time.time(),
             "access_count": 1,
             "access_time": 0,
             "size": 0
         }
         if metadata:
             _metadata.update(metadata)
 
         reader = functools.partial(value_reader.read, 2**22)
 
         iterator = iter(reader, b'')
         size = self._write_file(full_path, iterator, 'xb')
         metadata['size'] = size
 
         # after archive is finished, we create a key to save the presence of the binary file
         with open(key_file, 'wb') as f:
             f.write(json.dumps(_metadata))
 
         return key, size, MODE_BINARY, filename, _metadata
 
     def fetch(self, key, retry=False, retry_attempts=10) -> tuple[typing.BinaryIO, dict]:
 
         if retry:
             for attempt in range(retry_attempts):
                 if key in self:
                     break
                 # we dind't find the key, wait 1s, and re-check
                 time.sleep(1)
 
         if key not in self:
             log.exception('requested {key} not found in {self}', key, self)
             raise KeyError(key)
 
         key_file = self._get_keyfile(key)
         with open(key_file, 'rb') as f:
             metadata = json.loads(f.read())
 
         filename = metadata['filename']
 
         try:
             return open(os.path.join(self.directory, filename), 'rb'), metadata
         finally:
             # update usage stats, count and accessed
             metadata["access_count"] = metadata.get("access_count", 0) + 1
             metadata["access_time"] = time.time()
 
             with open(key_file, 'wb') as f:
                 f.write(json.dumps(metadata))
 
     def random_filename(self):
         """Return filename and full-path tuple for file storage.
 
         Filename will be a randomly generated 28 character hexadecimal string
         with ".archive_cache" suffixed. Two levels of sub-directories will be used to
         reduce the size of directories. On older filesystems, lookups in
         directories with many files may be slow.
         """
 
         hex_name = codecs.encode(os.urandom(16), 'hex').decode('utf-8')
         sub_dir = os.path.join(hex_name[:2], hex_name[2:4])
         name = hex_name[4:] + '.archive_cache'
         filename = os.path.join(sub_dir, name)
         full_path = os.path.join(self.directory, filename)
         return filename, full_path
 
     def hash(self, key):
         """Compute portable hash for `key`.
 
         :param key: key to hash
         :return: hash value
 
         """
         mask = 0xFFFFFFFF
         return zlib.adler32(key.encode('utf-8')) & mask  # noqa
 
     def __contains__(self, key):
         """Return `True` if `key` matching item is found in cache.
 
         :param key: key matching item
         :return: True if key matching item
 
         """
         key_file = self._get_keyfile(key)
         return os.path.exists(key_file)
 
     def __repr__(self):
         return f'FileSystemCache(index={self._index}, dir={self.directory})'
 
 
 class FanoutCache:
     """Cache that shards keys and values."""
 
     def __init__(
         self, directory=None, **settings
     ):
         """Initialize cache instance.
 
         :param str directory: cache directory
         :param settings: settings dict
 
         """
         if directory is None:
             raise ValueError('directory cannot be None')
 
         directory = str(directory)
         directory = os.path.expanduser(directory)
         directory = os.path.expandvars(directory)
         self._directory = directory
 
         self._count = settings.pop('cache_shards')
         self._locking_url = settings.pop('locking_url')
 
         self._eviction_policy = settings['cache_eviction_policy']
         self._cache_size_limit = settings['cache_size_limit']
 
         self._shards = tuple(
             FileSystemCache(
                 index=num,
                 directory=os.path.join(directory, 'shard_%03d' % num),
                 **settings,
             )
             for num in range(self._count)
         )
         self._hash = self._shards[0].hash
 
     @property
     def directory(self):
         """Cache directory."""
         return self._directory
 
     def get_lock(self, lock_key):
         return GenerationLock(lock_key, self._locking_url)
 
     def _get_shard(self, key) -> FileSystemCache:
         index = self._hash(key) % self._count
         shard = self._shards[index]
         return shard
 
     def store(self, key, value_reader, metadata=None):
         shard = self._get_shard(key)
         return shard.store(key, value_reader, metadata)
 
     def fetch(self, key, retry=False, retry_attempts=10):
         """Return file handle corresponding to `key` from cache.
         """
         shard = self._get_shard(key)
         return shard.fetch(key, retry=retry, retry_attempts=retry_attempts)
 
     def has_key(self, key):
         """Return `True` if `key` matching item is found in cache.
 
         :param key: key for item
         :return: True if key is found
 
         """
         shard = self._get_shard(key)
         return key in shard
 
     def __contains__(self, item):
         return self.has_key(item)
 
     def evict(self, policy=None, size_limit=None):
         """
         Remove old items based on the conditions
 
 
         explanation of this algo:
         iterate over each shard, then for each shard iterate over the .key files
         read the key files metadata stored. This gives us a full list of keys, cached_archived, their size and
         access data, time creation, and access counts.
 
         Store that into a memory DB so we can run different sorting strategies easily.
         Summing the size is a sum sql query.
 
         Then we run a sorting strategy based on eviction policy.
         We iterate over sorted keys, and remove each checking if we hit the overall limit.
         """
 
         policy = policy or self._eviction_policy
         size_limit = size_limit or self._cache_size_limit
 
         select_policy = EVICTION_POLICY[policy]['evict']
 
         log.debug('Running eviction policy \'%s\', and checking for size limit: %s',
                   policy, format_size(size_limit))
 
         if select_policy is None:
             return 0
 
         db = DB()
 
         data = []
         cnt = 1
         for shard in self._shards:
             for key_file in os.listdir(shard.directory):
                 if key_file.endswith('.key'):
                     key_file_path = os.path.join(shard.directory, key_file)
                     with open(key_file_path, 'rb') as f:
                         metadata = json.loads(f.read())
 
                         size = metadata.get('size')
                         filename = metadata.get('filename')
                         full_path = metadata.get('full_path')
 
                         if not size:
                             # in case we don't have size re-calc it...
                             size = os.stat(full_path).st_size
 
                         data.append([
                           cnt,
                           key_file,
                           key_file_path,
                           filename,
                           full_path,
                           metadata.get('store_time', 0),
                           metadata.get('access_time', 0),
                           metadata.get('access_count', 0),
                           size,
                         ])
                         cnt += 1
 
         # Insert bulk data using executemany
         db.bulk_insert(data)
 
         ((total_size,),) = db.sql('SELECT COALESCE(SUM(size), 0) FROM archive_cache').fetchall()
         log.debug('Analyzed %s keys, occupied: %s', len(data), format_size(total_size))
         select_policy_qry = select_policy.format(fields='key_file_path, full_path, size')
         sorted_keys = db.sql(select_policy_qry).fetchall()
 
         removed_items = 0
         removed_size = 0
         for key, cached_file, size in sorted_keys:
             # simulate removal impact BEFORE removal
             total_size -= size
 
             if total_size <= size_limit:
                 # we obtained what we wanted...
                 break
 
             os.remove(cached_file)
             os.remove(key)
             removed_items += 1
             removed_size += size
 
         log.debug('Removed %s cache archives, and reduced size: %s', removed_items, format_size(removed_size))
         return removed_items
 
 
 def get_archival_config(config):
 
     final_config = {
 
     }
 
     for k, v in config.items():
         if k.startswith('archive_cache'):
             final_config[k] = v
 
     return final_config
 
 
 def get_archival_cache_store(config):
 
     global cache_meta
     if cache_meta is not None:
         return cache_meta
 
     config = get_archival_config(config)
     backend = config['archive_cache.backend.type']
     if backend != 'filesystem':
         raise ValueError('archive_cache.backend.type only supports "filesystem"')
 
     archive_cache_locking_url = config['archive_cache.locking.url']
     archive_cache_dir = config['archive_cache.filesystem.store_dir']
     archive_cache_size_gb = config['archive_cache.filesystem.cache_size_gb']
     archive_cache_shards = config['archive_cache.filesystem.cache_shards']
     archive_cache_eviction_policy = config['archive_cache.filesystem.eviction_policy']
 
     log.debug('Initializing archival cache instance under %s', archive_cache_dir)
 
     # check if it's ok to write, and re-create the archive cache
     if not os.path.isdir(archive_cache_dir):
         os.makedirs(archive_cache_dir, exist_ok=True)
 
     d_cache = FanoutCache(
         archive_cache_dir,
         locking_url=archive_cache_locking_url,
         cache_shards=archive_cache_shards,
         cache_size_limit=archive_cache_size_gb * 1024 * 1024 * 1024,
         cache_eviction_policy=archive_cache_eviction_policy
     )
     cache_meta = d_cache
     return cache_meta