Hi Naman,
Thanks a lot for your help!
I will fix for your comment, and send as v4 patch next week :)
Please help me understand if this is correct. The actual problem you are
highlighting lies here when sizeof(wchat_t) is more than 16, i.e. 32?
Yes, so we should not cast array of __u16 to wchar_t*.
I confirmed it on Linux as following:
```
[yasuenag@fc42 sizeof]$ cat test.c
#include <stdio.h>
#include <stddef.h>
int main(){
printf("%ld\n", sizeof(wchar_t));
return 0;
}
[yasuenag@fc42 sizeof]$ rpm -q gcc
gcc-15.1.1-2.fc42.x86_64
[yasuenag@fc42 sizeof]$ gcc test.c
[yasuenag@fc42 sizeof]$ ./a.out
4
```
In Windows, it returns 2:
```
PS C:\test\sizeof> cl.exe .\test.c
Microsoft(R) C/C++ Optimizing Compiler Version 19.44.35208 for x86
Copyright (C) Microsoft Corporation. All rights reserved.
test.c
Microsoft (R) Incremental Linker Version 14.44.35208.0
Copyright (C) Microsoft Corporation. All rights reserved.
/out:test.exe
test.obj
PS C:\test\sizeof> .\test.exe
2
```
Thanks,
Yasumasa
On 2025/06/04 21:36, Naman Jain wrote:
On 6/4/2025 2:40 PM, Yasumasa Suenaga wrote:
Hi Naman,
Sorry for bothering you, but I have some questions:
Hi,
Happy to help. I have tested your patch on my setup, and it did not break
anything for me. Please find response and additional questions inline.
1. Create a new patch file for every new version and then send it.
Currently it seems you are manually editing the same patch file in the
subject and sending it, so each patch version is showing up in the same
thread.
I've committed changes with "git amend" and created formatted patch with "git
format-patch",
then I sent all of *.patch files via "git send-email".
Do you mean I should commit to fix commits reviewers and send cover letter and
incremental
patches only? I couldn't find about it from the guide.
I checked linux-hyperv list, all of patches have been sent in each version
AFAICS.
* make changes
* git add, git commit, git format-patch, git send-email <path to v1 patch)
* make changes for v2
* git add, git commit --amend, git format-patch -v2, git send-email <path to v2
patch *only*>
and so on.
I follow above practice, and the patches go in new threads. I am not sure why
they are landing in same thread for you. Anyhow, its not a big thing.
3. Keep a minimum of 1-2 weeks gap between successive patch versions to
give time to people to review your changes.
Does it include changes of commit message too?
Other reviewers may be reviewing older versions and it becomes confusing if new
versions come too soon. To give everyone enough time to review your changes and
to not have unnecessary patch versions, its better to wait for a few days,
collect the feedback and address it together. We also need to mention what we
changed since last time, to help reviewers know if their comments got addressed.
For single patch, we usually don't add a cover letter. You can refer
https://lore.kernel.org/all/20250102145243.2088-1-namj...@linux.microsoft.com/
The comments between the two "---" does not make it to git log.
4. In the commit msg of v3, it is still not very clear what problem, you
are trying to fix here. Do you mean to say that fcopy does not work on
Linux? Or you are assuming it won't work and fixing some generic
problem?
Fcopy is supposed to work fine on Linux VM on HyperV with windows host. If
there are some errors, please share in cover letter/comments
in the patch along with steps of execution.
I have a problem on my PC:
Host: Windows 11 Pro (24H2 build 26100.4202)
Guest: Fedora 42
- kernel-6.14.4-300.fc42.x86_64
- hypervfcopyd-6.10-1.fc42.x86_64
How to reproduce: run Copy-VMFile commandlet on Host:
Following log is in Japanese because my PC is set to Japanese, sorry.
But it says fcopy could not transfer file (test.ps1) to /tmp/ on Linux guest
because it already exists.
I confirmed /tmp/test.ps1 does not exist of course.
```
Copy-VMFile
cmdlet Copy-VMFile at command pipeline position 1
Supply values for the following parameters:
Name[0]: Fedora42
Name[1]:
SourcePath: test.ps1
DestinationPath: /tmp/
FileSource: Host
Copy-VMFile: ゲストへのファイルのコピーを開始できませんでした。
ソース ファイル 'C:\test\test.ps1' をゲストの宛先 '/tmp/' にコピーできま せんでした。
'Fedora42' はゲスト: ファイルがあります。 (0x80070050) へのファイルのコ ピーを開始できませんでした。(仮想マシン ID
9BFDF23D-CCAA-4748- A770-6D654E09A133)
'Fedora42' は、コピー元ファイル 'C:\test\test.ps1' をゲスト上のコピー先 '/tmp/' にコピーできませんでした:
ファイルがあります。 (0x80070050)。(仮 想マシン ID 9BFDF23D-CCAA-4748-A770-6D654E09A133)
```
I got following fcopy log from journald - it is strange because "/tmp/
test.ps1" should be shown here.
```
6月 04 17:48:24 fc42 HV_UIO_FCOPY[1080]: File: / exists
```
As I wrote in commit message, wchar_t is 32 bit in Linux. I confirmed it with
"sizeof(wchar_t)".
However fcopyd handles it as 16 bit value (__u16), thus I think this is a bug
in fcopy, and
I think it would also not work on other environments.
Actually it works fine with my patch to handle values as 16 bit.
Thanks,
Yasumasa
On 2025/06/04 15:19, Naman Jain wrote:
On 6/4/2025 5:13 AM, yasue...@gmail.com wrote:
From: Yasumasa Suenaga <yasue...@gmail.com>
Handle file copy request from the host (e.g. Copy-VMFile commandlet)
correctly.
Store file path and name as __u16 arrays in struct hv_start_fcopy.
Convert directly to UTF-8 string without casting to wchar_t* in fcopyd.
Fix string conversion failure caused by wchar_t size difference between
Linux (32bit) and Windows (16bit). Convert each character to char
if the value is less than 0x80 instead of using wcstombs() call.
Add new check to snprintf() call for target path creation to handle
length differences between PATH_MAX (Linux) and W_MAX_PATH (Windows).
Signed-off-by: Yasumasa Suenaga <yasue...@gmail.com>
---
tools/hv/hv_fcopy_uio_daemon.c | 37 ++++++++++++++--------------------
1 file changed, 15 insertions(+), 22 deletions(-)
diff --git a/tools/hv/hv_fcopy_uio_daemon.c b/tools/hv/ hv_fcopy_uio_daemon.c
index 0198321d1..86702f39e 100644
--- a/tools/hv/hv_fcopy_uio_daemon.c
+++ b/tools/hv/hv_fcopy_uio_daemon.c
@@ -62,8 +62,11 @@ static int hv_fcopy_create_file(char *file_name, char
*path_name, __u32 flags)
filesize = 0;
p = path_name;
- snprintf(target_fname, sizeof(target_fname), "%s/%s",
- path_name, file_name);
+ if (snprintf(target_fname, sizeof(target_fname), "%s/%s",
+ path_name, file_name) >= sizeof(target_fname)) {
+ /* target file name is too long */
Please add a syslog for this. It will help in debugging issues.
+ goto done;
+ }
/*
* Check to see if the path is already in place; if not,
@@ -273,6 +276,8 @@ static void wcstoutf8(char *dest, const __u16 *src, size_t
dest_size)
while (len < dest_size) {
if (src[len] < 0x80)
dest[len++] = (char)(*src++);
+ else if (src[len] == '0')
+ break;
While this also works, I think you can add it to the while loop itself
while (len < dest_size && src[len])
Is this related to the FCopy issue that you see or this is just a fix for
correct destination file name?
else
dest[len++] = 'X';
}
@@ -282,27 +287,15 @@ static void wcstoutf8(char *dest, const __u16 *src,
size_t dest_size)
static int hv_fcopy_start(struct hv_start_fcopy *smsg_in)
{
- setlocale(LC_ALL, "en_US.utf8");
- size_t file_size, path_size;
- char *file_name, *path_name;
- char *in_file_name = (char *)smsg_in->file_name;
- char *in_path_name = (char *)smsg_in->path_name;
-
- file_size = wcstombs(NULL, (const wchar_t *restrict)in_file_name, 0) + 1;
- path_size = wcstombs(NULL, (const wchar_t *restrict)in_path_name, 0) + 1;
-
Please help me understand if this is correct. The actual problem you are
highlighting lies here when sizeof(wchat_t) is more than 16, i.e. 32?
- file_name = (char *)malloc(file_size * sizeof(char));
- path_name = (char *)malloc(path_size * sizeof(char));
-
- if (!file_name || !path_name) {
- free(file_name);
- free(path_name);
- syslog(LOG_ERR, "Can't allocate memory for file name and/or path
name");
- return HV_E_FAIL;
- }
+ /*
+ * file_name and path_name should have same length with appropriate
+ * member of hv_start_fcopy.
+ */
+ char file_name[W_MAX_PATH], path_name[W_MAX_PATH];
- wcstoutf8(file_name, (__u16 *)in_file_name, file_size);
- wcstoutf8(path_name, (__u16 *)in_path_name, path_size);
+ setlocale(LC_ALL, "en_US.utf8");
+ wcstoutf8(file_name, smsg_in->file_name, W_MAX_PATH - 1);
+ wcstoutf8(path_name, smsg_in->path_name, W_MAX_PATH - 1);
return hv_fcopy_create_file(file_name, path_name, smsg_in- >copy_flags);
}
Hi,
I understand this is your first patch for upstreaming. Here are a few
things you should consider:
1. Create a new patch file for every new version and then send it.
Currently it seems you are manually editing the same patch file in the
subject and sending it, so each patch version is showing up in the same
thread.
2. Read, re-read, absorb the information in the link that Easwar also
mentioned:
https://www.kernel.org/doc/html/latest/process/submitting-patches.html
3. Keep a minimum of 1-2 weeks gap between successive patch versions to
give time to people to review your changes.
4. In the commit msg of v3, it is still not very clear what problem, you
are trying to fix here. Do you mean to say that fcopy does not work on
Linux? Or you are assuming it won't work and fixing some generic
problem?
Fcopy is supposed to work fine on Linux VM on HyperV with windows host. If
there are some errors, please share in cover letter/comments
in the patch along with steps of execution.
5. If its a fix, we should have a proper Fixes tag with the commit you
are fixing.
6. Have a look at existing conversations at lore to get to know common
practices with single patch, multi patch, cover letters etc.
https://lore.kernel.org/linux-hyperv/?q=linux-hyperv
Regards,
Naman
Regards,
Naman
