Fix infinite loop when using characters from unicode supplementary pl…

[Ratio2]

…anes ('🔈' for example)

Fixes # .

Checklist

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the existing documentation
• My changes generate no new warnings
• I have updated the change log (Add/Change/Fix)
• I have cleaned up the commit history (use rebase and squash)

Changes proposed in this pull request:

[Liryna]

Hi @Ratio2 ,

Thank you for the PR. There are a couple of changes in your PR probably not all related to the issue.
Can you clean it ?
I imagine you are talking about put_utf8 for the infinite loop ?

[Ratio2]

@Liryna Make cleanup as you requested.

[Liryna]

Hi @Ratio2 ,

I have tried to reproduce your issue with a small sample outside dokan_fuse and could not reproduce the infinite loop.

 unsigned char src[] = "🔈";
 unsigned char dst[16];
 convert_char(get_utf16, put_utf8, src, 2, dst);

Am I doing something different ?

[Ratio2]

@Liryna I reproduce bug in encfs bundle: https://github.com/jetwhiz/encfs4win/releases/download/v1.10.1/encfs-installer.exe, but bug in dokan itself.

[Liryna]

Hi @jetwhiz,

Do you have any knowledge about such issue ?

[Ratio2]

@Liryna Bug occured when we convert utf-16 to utf-8. We got 4 bytes, but with invalid value. When we decode this utf-8, only 2 bytes consumed instead 4, second 2 bytes return error with negative result and str len increased (negative cast to big number => undefined behavior) instead reducing.

[Liryna]

Do you have sample data (src buffer) that I could use to reproduce ?

[Ratio2]

@Liryna I don't report this issue to encfs team. Just find problem and fix it in dokan ;)

[Ratio2]

@Liryna https://pastebin.com/YzwHM948

[Ratio2]

@Liryna Sorry, my test was bad, i fixed it:

int main()
{
	char utf8[5] = "\xf0\x9f\x94\x88";
	wchar_t utf8_to_utf16[3] = {};
	auto res = convert_char(get_utf8, put_utf16, utf8, 4, utf8_to_utf16);
	assert(res == 4);
	char utf16_to_utf8[5] = {};
	res = convert_char(get_utf16, put_utf8, utf8_to_utf16, 4, utf16_to_utf8);
	assert(res == 4);
	assert(!strcmp(utf8, utf16_to_utf8));
}

[Ratio2]

@Liryna One more simple test:

int main()
{
	wchar_t utf16[3] = L"\xd83d\xdd08";
	char utf8[5] = {};
	auto res = convert_char(get_utf16, put_utf8, utf16, 4, utf8);
	assert(res == 4);
	assert(utf8[0] != 0); // wtf?
}

[Liryna]

Thank you very much @Ratio2 ! 🏆

I am not the author of the dokan_fuse. I was wondering if you understood why the author used the condition if (c < 0x40) ?

[Ratio2]

@Liryna
We have unicode character with code 0x1f508: https://www.unicode.org/charts/PDF/U1F300.pdf
In bits ‭11111010100001000‬ - 17 bits.
According to utf-8 we can encode 16 bits to 3 bytes and 21 bits to 4 bytes: https://en.wikipedia.org/wiki/UTF-8#Description
Split to 4 byte utf-8: 000 011111 010100 001000. First byte is zero, but it must be exists, author of original code make mistake assuming that first byte can not be zero and remain garbage in first byte ;)

[Ratio2]

@Liryna Condition if (c < 0x40) check if remain 6 bits or less (max bits per utf-8 byte if encoded more then one).

[Liryna]

Thank you @Ratio2 this explanation is very helpful 👍
More special thank for your PR, please do not hesitate if you find anything else 🚀

[jetwhiz]

Thanks @Ratio2 and @Liryna !

I wonder if this is related to the issue regarding Thai characters eating up 100% CPU resources? jetwhiz/encfs4win#94 Do Thai characters fall into this category?

[Liryna]

Hi @jetwhiz

http://www.ltg.ed.ac.uk/~richard/utf-8.cgi?input=%E0%B8%8B%E0%B9%88&mode=char
utf8 code also starts with a zero so therefore the same case as @Ratio2 reported.
I just made and test and I can confirm. Previous code corrupted the character but with @Ratio2 fix, we get the correct value 👍