From: Akihiko Odaki <akihiko.odaki@daynix.com>
To: "Daniel P. Berrangé" <berrange@redhat.com>
Cc: qemu-devel@nongnu.org, qemu-block@nongnu.org,
Yuval Shaia <yuval.shaia.ml@gmail.com>,
Marcel Apfelbaum <marcel.apfelbaum@gmail.com>,
Konstantin Kostiuk <kkostiuk@redhat.com>,
Michael Roth <michael.roth@amd.com>,
Paolo Bonzini <pbonzini@redhat.com>, Fam Zheng <fam@euphon.net>,
"Dr . David Alan Gilbert" <dgilbert@redhat.com>,
Stefan Hajnoczi <stefanha@redhat.com>,
Gerd Hoffmann <kraxel@redhat.com>, Stefan Weil <sw@weilnetz.de>,
Yan Vugenfirer <yan@daynix.com>
Subject: Re: [PATCH v4 1/7] util: Introduce qemu_get_runtime_dir()
Date: Tue, 16 Jul 2024 20:27:39 +0900 [thread overview]
Message-ID: <7bee9c26-d75e-4d83-bc2a-c97b06416080@daynix.com> (raw)
In-Reply-To: <ZpZRYyW74wfqR6Zr@redhat.com>
On 2024/07/16 19:54, Daniel P. Berrangé wrote:
> On Tue, Jul 16, 2024 at 07:52:00PM +0900, Akihiko Odaki wrote:
>> On 2024/07/16 18:53, Daniel P. Berrangé wrote:
>>> On Tue, Jul 16, 2024 at 04:27:31PM +0900, Akihiko Odaki wrote:
>>>> qemu_get_runtime_dir() returns a dynamically allocated directory path
>>>> that is appropriate for storing runtime files. It corresponds to "run"
>>>> directory in Unix.
>>>>
>>>> With a tree-wide search, it was found that there are several cases
>>>> where such a functionality is implemented so let's have one as a common
>>>> utlity function.
>>>>
>>>> A notable feature of qemu_get_runtime_dir() is that it uses
>>>> $XDG_RUNTIME_DIR if available. While the function is often called by
>>>> executables which requires root privileges, it is still possible that
>>>> they are called from a user without privilege to write the system
>>>> runtime directory. In fact, I decided to write this patch when I ran
>>>> virtiofsd in a Linux namespace created by a normal user and realized
>>>> it tries to write the system runtime directory, not writable in this
>>>> case. $XDG_RUNTIME_DIR should provide a writable directory in such
>>>> cases.
>>>>
>>>> This function does not use qemu_get_local_state_dir() or its logic
>>>> for Windows. Actually the implementation of qemu_get_local_state_dir()
>>>> for Windows seems not right as it calls g_get_system_data_dirs(),
>>>> which refers to $XDG_DATA_DIRS. In Unix terminology, it is basically
>>>> "/usr/share", not "/var", which qemu_get_local_state_dir() is intended
>>>> to provide. Instead, this function try to use the following in order:
>>>> - $XDG_RUNTIME_DIR
>>>> - LocalAppData folder
>>>> - get_relocated_path(CONFIG_QEMU_LOCALSTATEDIR "/run")
>>>>
>>>> This function does not use g_get_user_runtime_dir() either as it
>>>> falls back to g_get_user_cache_dir() when $XDG_DATA_DIRS is not
>>>> available. In the case, we rather use:
>>>> get_relocated_path(CONFIG_QEMU_LOCALSTATEDIR "/run")
>>>>
>>>> Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
>>>> Message-Id: <20230921075425.16738-2-akihiko.odaki@daynix.com>
>>>> ---
>>>> include/qemu/osdep.h | 12 ++++++++++++
>>>> util/oslib-posix.c | 11 +++++++++++
>>>> util/oslib-win32.c | 26 ++++++++++++++++++++++++++
>>>> 3 files changed, 49 insertions(+)
>>>>
>>>> diff --git a/include/qemu/osdep.h b/include/qemu/osdep.h
>>>> index 191916f38e6d..fe8609fc1375 100644
>>>> --- a/include/qemu/osdep.h
>>>> +++ b/include/qemu/osdep.h
>>>> @@ -670,6 +670,18 @@ void qemu_set_cloexec(int fd);
>>>> */
>>>> char *qemu_get_local_state_dir(void);
>>>> +/**
>>>> + * qemu_get_runtime_dir:
>>>> + *
>>>> + * Return a dynamically allocated directory path that is appropriate for storing
>>>> + * runtime files. It corresponds to "run" directory in Unix, and uses
>>>> + * $XDG_RUNTIME_DIR if available.
>>>> + *
>>>> + * The caller is responsible for releasing the value returned with g_free()
>>>> + * after use.
>>>> + */
>>>> +char *qemu_get_runtime_dir(void);
>>>> +
>>>> /**
>>>> * qemu_getauxval:
>>>> * @type: the auxiliary vector key to lookup
>>>> diff --git a/util/oslib-posix.c b/util/oslib-posix.c
>>>> index e76441695bdc..9599509a9aa7 100644
>>>> --- a/util/oslib-posix.c
>>>> +++ b/util/oslib-posix.c
>>>> @@ -278,6 +278,17 @@ qemu_get_local_state_dir(void)
>>>> return get_relocated_path(CONFIG_QEMU_LOCALSTATEDIR);
>>>> }
>>>> +char *
>>>> +qemu_get_runtime_dir(void)
>>>> +{
>>>> + char *env = getenv("XDG_RUNTIME_DIR");
>>>> + if (env) {
>>>> + return g_strdup(env);
>>>> + }
>>>> +
>>>> + return get_relocated_path(CONFIG_QEMU_LOCALSTATEDIR "/run");
>>>> +}
>>>
>>> I'm not convinced this is the correct logic to be following.
>>>
>>> In the cover letter you mention not using g_get_user_runtime_dir()
>>> because it falls back to XDG_CACHE_HOME, and we need to fallback
>>> to LOCALSTATEDIR/run. This is not right for normal users though,
>>> where falling back to LOCALSTATEDIR/run is always wrong, as it
>>> won't be writable - the g_get_user_runtime_dir() fallback is
>>> desirable for non-root users.
>>
>> It also checks LocalAppData, which should be usually available.
>>
>> g_get_user_runtime_dir() is not a proper fallback in case neither of
>> XDG_RUNTIME_DIR and LocalAppData are available. g_get_user_cache_dir(),
>> which gets called by g_get_user_runtime_dir(), internally uses:
>> - XDG_CACHE_HOME or
>> - FOLDERID_InternetCache
>>
>> g_get_user_cache_dir() just returns NULL if neither of them is available.
>>
>> We can't expect XDG_CACHE_HOME is present when XDG_RUNTIME_DIR is missing.
>> FOLDERID_InternetCache points to %LOCALAPPDATA%\Microsoft\Windows\Temporary
>> Internet Files, according to:
>> https://learn.microsoft.com/en-us/windows/win32/shell/knownfolderid
>>
>> So we can't expect FOLDERID_InternetCache is available when LocalAppData is
>> missing.
>
> XDG_CACHE_HOME isn't required to be present. Glib will use a fallback
> location if XDG_CACHE_HOME isn't set, and it will mkdir() the location
> if it doesn't exist.
The fallback location is FOLDERID_InternetCache, which is not better
than looking at LocalAppData.
Regards,
Akihiko Odaki
next prev parent reply other threads:[~2024-07-16 11:30 UTC|newest]
Thread overview: 22+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-07-16 7:27 [PATCH v4 0/7] util: Introduce qemu_get_runtime_dir() Akihiko Odaki
2024-07-16 7:27 ` [PATCH v4 1/7] " Akihiko Odaki
2024-07-16 9:53 ` Daniel P. Berrangé
2024-07-16 10:52 ` Akihiko Odaki
2024-07-16 10:54 ` Daniel P. Berrangé
2024-07-16 11:27 ` Akihiko Odaki [this message]
2024-07-16 7:27 ` [PATCH v4 2/7] ivshmem-server: Use qemu_get_runtime_dir() Akihiko Odaki
2024-07-16 7:27 ` [PATCH v4 3/7] qga: " Akihiko Odaki
2024-07-16 7:27 ` [PATCH v4 4/7] scsi: " Akihiko Odaki
2024-07-16 7:27 ` [PATCH v4 5/7] module: " Akihiko Odaki
2024-07-16 7:27 ` [PATCH v4 6/7] util: Remove qemu_get_local_state_dir() Akihiko Odaki
2024-07-16 7:27 ` [PATCH v4 7/7] spice-app: Use qemu_get_runtime_dir() Akihiko Odaki
2024-07-16 8:06 ` [PATCH v4 0/7] util: Introduce qemu_get_runtime_dir() Michael Tokarev
2024-07-16 9:32 ` Akihiko Odaki
2024-07-16 9:41 ` Michael Tokarev
2024-07-16 9:56 ` Daniel P. Berrangé
2024-07-16 10:43 ` Paolo Bonzini
2024-07-16 12:45 ` Akihiko Odaki
2024-07-16 13:29 ` Paolo Bonzini
2024-07-16 13:35 ` Akihiko Odaki
2024-07-16 8:45 ` Paolo Bonzini
2025-01-11 5:15 ` Akihiko Odaki
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=7bee9c26-d75e-4d83-bc2a-c97b06416080@daynix.com \
--to=akihiko.odaki@daynix.com \
--cc=berrange@redhat.com \
--cc=dgilbert@redhat.com \
--cc=fam@euphon.net \
--cc=kkostiuk@redhat.com \
--cc=kraxel@redhat.com \
--cc=marcel.apfelbaum@gmail.com \
--cc=michael.roth@amd.com \
--cc=pbonzini@redhat.com \
--cc=qemu-block@nongnu.org \
--cc=qemu-devel@nongnu.org \
--cc=stefanha@redhat.com \
--cc=sw@weilnetz.de \
--cc=yan@daynix.com \
--cc=yuval.shaia.ml@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).