Fatal error on thread creation in low memory condition: local storage

[vstinner]

BPO	8197
Nosy	@vstinner

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2010-09-10.22:29:29.420>
created_at = <Date 2010-03-22.01:08:22.931>
labels = ['type-crash']
title = 'Fatal error on thread creation in low memory condition: local storage'
updated_at = <Date 2010-09-10.22:29:29.419>
user = 'https://github.com/vstinner'

bugs.python.org fields:

activity = <Date 2010-09-10.22:29:29.419>
actor = 'vstinner'
assignee = 'none'
closed = True
closed_date = <Date 2010-09-10.22:29:29.420>
closer = 'vstinner'
components = []
creation = <Date 2010-03-22.01:08:22.931>
creator = 'vstinner'
dependencies = []
files = []
hgrepos = []
issue_num = 8197
keywords = []
message_count = 2.0
messages = ['101474', '116052']
nosy_count = 1.0
nosy_names = ['vstinner']
pr_nums = []
priority = 'low'
resolution = 'wont fix'
stage = None
status = 'closed'
superseder = None
type = 'crash'
url = 'https://bugs.python.org/issue8197'
versions = ['Python 2.7', 'Python 3.2']

[vstinner]

I wrote a patch to preallocate Python thread state before creating the thread to avoid a fatal error: bpo-7544 (it's now closed). This patch is not enough to avoid fatal errors in low memory condition.

Just after the creation of the thread, _PyGILState_NoteThreadState() is called. This function can fail with a fatal error (Couldn't create autoTLSkey mapping) in low memory condition if a memory allocation fails in find_key(), called by PyThread_set_key_value().

PyThread_set_key_value() fills a global single linked list of type 'struct key', the head is 'keyhead'. This list contains values and uses an index composed of (long thread id, int key). Only one key is used: autoTLSkey (0). The list is used to get the thread state (eg. in PyGILState_Ensure).

Note: This issue is very unlikely, but it does exist :-)

[vstinner]

I don't know how to fix this issue, and I don't know if it can be fixed. As the issue is very unlikely, I prefer to close it.