On a failed bind cmdPASV jumped straight back to .next_port, and
nextpasvport wraps within [start, end], so a range with no usable port
spun forever without reading the control connection again: the client
hangs.
Count the ports in the range (end - start + 1), decrement after a
failed bind and give up at zero, then close the listener, reset the
state and answer 425 Can't open data connection. The counter lives in a
new thread_data field rather than a register, because mcall bind
clobbers eax/ebx/ecx/edx/esi/edi and esi already carries the sockaddr
length for bind. Decrementing after the failed bind and not before the
next attempt is what keeps a single-port range (start = end) working:
it still gets its one attempt.
Also validate the pair read from the ini. With start > end the range
degenerates and no port is bindable, so fall back to 2000/5000.
---------
Co-authored-by: Igor Shutrov <kolibridev@inbox.ru>
Reviewed-on: KolibriOS/kolibrios#710
Reviewed-by: Kiril Lipatov <lipatov.kiril@gmail.com>
Reviewed-by: Burer <burer@kolibrios.org>
Co-authored-by: Igor shutrov <kolibridev@mail.ru>
mcall close takes the socket number in ecx, but the operand was
commented out in the two places that close the PASV listener, so close
ran with whatever ecx happened to hold. Both call sites had just put the
listener in ecx, so the socket was closed by accident: the code worked
only as long as that accident held. The cmdPASV cleanup was commented
out completely, so a second PASV before the client connected left the
previous listener open.
Restore the operand in threadloop (ftpd.asm) and in open_datasock, and
re-enable the cmdPASV prologue, which also resets passivesocknum to -1.
Closing the data socket needed the same care: datasocketnum only holds
a live socket in MODE_PASSIVE_OK and MODE_ACTIVE, so the new
close_data_sock helper closes it in those two states only. The
.cannot_open label is also reachable before open_datasock runs (the
length checks at the top of cmdSTOR), where datasocketnum is either
stale or never initialised, and closing it unconditionally there could
have closed an unrelated socket that had reused the number. The field
is now initialised to -1 at thread start and reset to -1 after every
close.
---------
Co-authored-by: leency <lipatov.kiril@gmail.com>
Co-authored-by: Igor Shutrov <kolibridev@inbox.ru>
Reviewed-on: KolibriOS/kolibrios#707
Reviewed-by: Kiril Lipatov <lipatov.kiril@gmail.com>
Reviewed-by: Burer <burer@kolibrios.org>
Co-authored-by: igorsh <kolibridev@mail.ru>
cmdSTOR opens the destination with O_CREATE + O_WRITE, and libio only
truncates on write when the position is already at EOF (libio.asm:314),
which never happens here. Replacing a longer file with a shorter one
therefore kept the old tail: 4096-byte A overwritten by 1024-byte B
produced a 4096-byte file holding B followed by A[1024:4096], and
overwriting with zero bytes left the file completely untouched.
Call file.truncate (SF 70.4, size = Position = 0) right after a
successful open so that STOR replaces the file, as RFC 959 requires.
The descriptor survives the call in ebx because sendFTP clobbers the
registers. A failed truncate closes the descriptor and answers 550
through a new .truncate_error path.
The libio import list gains file_truncate, exported since 2009.
-------------------------
STOR: усекать файл при перезаписи, чтобы не оставался старый хвост
Проблема
cmdSTOR открывает целевой файл с O_CREATE + O_WRITE, а libio усекает файл при записи только если позиция уже находится в EOF (libio.asm:314). В этом сценарии условие никогда не выполняется. Поэтому при замене длинного файла коротким сохранялся старый хвост: 4096-байтный A, перезаписанный 1024-байтным B, давал 4096-байтный файл, содержащий B, а затем A[1024:4096]. Перезапись нулевым количеством байт вообще не изменяла файл.
Решение
Сразу после успешного открытия вызывается file.truncate (SF 70.4, size = Position = 0), чтобы STOR заменял файл, как требует RFC 959. Дескриптор сохраняется в ebx после вызова, так как sendFTP затирает регистры. Если truncate завершается ошибкой, дескриптор закрывается, а ответ 550 отправляется через новый обработчик .truncate_error.
Дополнительно
В список импортов libio добавлен file_truncate (экспортируется с 2009 года).
Co-authored-by: leency <lipatov.kiril@gmail.com>
Co-authored-by: Igor Shutrov <kolibridev@inbox.ru>
Reviewed-on: KolibriOS/kolibrios#706
Reviewed-by: Kiril Lipatov <lipatov.kiril@gmail.com>
Reviewed-by: Burer <burer@kolibrios.org>
Co-authored-by: igorsh <kolibridev@mail.ru>
Sysfn 76 returns -1 for a missing device. The ARP conflict counter
check in .maintain_link treats any change of the value as a new
conflict, so when a device disappeared the counter jumped 0 -> -1 -> 0
across the reconnect and a bogus "IP address conflict" notification
popped up on every replug of a USB network device.
Skip the check when the query fails.
Assisted-by: Claude Opus 5 <noreply@anthropic.com>
Reviewed-on: KolibriOS/kolibrios#677
Reviewed-by: hidnplayr <hidnplayr@gmail.com>
Reviewed-by: Burer <burer@kolibrios.org>
Co-authored-by: leency <lipatov.kiril@gmail.com>
Sysfn 74 returns -1 for a missing device, not 0, so the device-presence
check in .maintain_link never took the .link_down path and the interface
stayed in connected state forever. As a result DHCP was never restarted
when a device reappeared on the same slot (e.g. an Android phone
re-enumerating in RNDIS tethering mode with a new random MAC), leaving
the interface with a stale IP and a dead gateway.
Compare against 1 (Ethernet), the same way the .link_up? path already
does, so a vanished device properly resets the interface state and DHCP
runs again on reconnect.
Assisted-by: Claude Opus 5 <noreply@anthropic.com>
Reviewed-on: KolibriOS/kolibrios#676
Reviewed-by: hidnplayr <hidnplayr@gmail.com>
Reviewed-by: Burer <burer@kolibrios.org>
Co-authored-by: leency <lipatov.kiril@gmail.com>
- Move source code from `trunk` into root directory.
- Update build files and ASM include paths. Add SPDX license header where appropriate.
- These changes also update the CI build files in `.gitea/workflows/`.
- Note: Line endings standardised from `CRLF` > `LF`, so best to view diffs with whitespace changes hidden.
(Work towards [#75](KolibriOS/kolibrios#75), point 3)
---------
Co-authored-by: Burer <burer@kolibrios.org>
Reviewed-on: KolibriOS/kolibrios#425
Reviewed-by: Mikhail Frolov <mixa.frolov2003@gmail.com>
Reviewed-by: Burer <burer@kolibrios.org>
Co-authored-by: Andrew <dent.ace@gmail.com>
Co-committed-by: Andrew <dent.ace@gmail.com>
- fix for bugs with buffers 'remote_list_buf', 'buf_buffer2';
- fix stack leak when navigating files;
- change sorting of folders and files, clean code;
- added option to show|hide log;
- icons are taken from @reshare;
- added icons to the 'disconnect' and 'search' buttons.
Reviewed-on: KolibriOS/kolibrios#185
Reviewed-by: Max Logaev <maxlogaev@proton.me>
Co-authored-by: IgorA <aie85playm@gmail.com>
Co-committed-by: IgorA <aie85playm@gmail.com>
Lossless optimization of png image files, using `pngslim` and other tools.
Program graphics (compiled or included with software):
- Slimmed 102 B (~102 B kpacked), average ~11% saving for 2 files.
git-svn-id: svn://kolibrios.org@10024 a494cfbc-eb01-0410-851d-a64ba20cac60
-Use HMAC and CTR/CBC from libcrash instead of our own implementations
-Fixed stack allocation for keystroke handler thread
git-svn-id: svn://kolibrios.org@9987 a494cfbc-eb01-0410-851d-a64ba20cac60
igorshandIgor Shutrov
Doczom
Egor00fandBurer
ace-dent
b00bl1k
IgorA
ace_dent
hidnplayr
Doczom
turbocat
vitalkrilov
sdongles
IgorA
dunkaist