rongqing.li at windriver.com
2013-Aug-22 01:00 UTC
[syslinux] [PATCH] tftp-hpa: add error check for disk filled up
From: "Roy.Li" <rongqing.li at windriver.com> Add error check when the write-buffer is finally flushed to the file, the caller can detect if the disk filled up (or had an i/o error) and return a NOSAPCE nak to the other side. Signed-off-by: Ming Liu <ming.liu at windriver.com> Signed-off-by: Roy.Li <rongqing.li at windriver.com> --- common/tftpsubs.c | 8 +++++--- tftpd/tftpd.c | 12 ++++++++++-- 2 files changed, 15 insertions(+), 5 deletions(-) diff --git a/common/tftpsubs.c b/common/tftpsubs.c index 8c999f6..b4d4ffe 100644 --- a/common/tftpsubs.c +++ b/common/tftpsubs.c @@ -206,20 +206,22 @@ int write_behind(FILE * file, int convert) p = buf; ct = count; + count = 0; while (ct--) { /* loop over the buffer */ c = *p++; /* pick up a character */ if (prevchar == '\r') { /* if prev char was cr */ if (c == '\n') /* if have cr,lf then just */ - fseek(file, -1, 1); /* smash lf on top of the cr */ + count = count - 1; else if (c == '\0') /* if have cr,nul then */ goto skipit; /* just skip over the putc */ /* else just fall through and allow it */ } - putc(c, file); + buf[count] = c; + count ++; skipit: prevchar = c; } - return count; + return write(fileno(file), buf, count); } /* When an error has occurred, it is possible that the two sides diff --git a/tftpd/tftpd.c b/tftpd/tftpd.c index 1873e70..c2adbda 100644 --- a/tftpd/tftpd.c +++ b/tftpd/tftpd.c @@ -1681,7 +1681,11 @@ static void tftp_recvfile(const struct formats *pf, struct tftphdr *oap, int oac syslog(LOG_WARNING, "tftpd: write(ack): %m"); goto abort; } - write_behind(file, pf->f_convert); + if(write_behind(file, pf->f_convert) < 0) { + nak(ENOSPACE, NULL); + (void)fclose(file); + goto abort; + } for (;;) { n = recv_time(peer, dp, PKTSIZE, 0, &r_timeout); if (n < 0) { /* really? */ @@ -1712,7 +1716,11 @@ static void tftp_recvfile(const struct formats *pf, struct tftphdr *oap, int oac goto abort; } } while (size == segsize); - write_behind(file, pf->f_convert); + if(write_behind(file, pf->f_convert) < 0) { + nak(ENOSPACE, NULL); + (void)fclose(file); + goto abort; + } (void)fclose(file); /* close data file */ ap->th_opcode = htons((u_short) ACK); /* send the "final" ack */ -- 1.7.10.4
H. Peter Anvin
2013-Aug-28 19:13 UTC
[syslinux] [PATCH] tftp-hpa: add error check for disk filled up
On 08/21/2013 06:00 PM, rongqing.li at windriver.com wrote:> From: "Roy.Li" <rongqing.li at windriver.com> > > Add error check when the write-buffer is finally flushed to the file, > the caller can detect if the disk filled up (or had an i/o error) and > return a NOSAPCE nak to the other side. > > Signed-off-by: Ming Liu <ming.liu at windriver.com> > Signed-off-by: Roy.Li <rongqing.li at windriver.com>Don't mix stdio and raw I/O this way. The right thing to do is either to use fflush() and ferror() to detect an error condition, or consistently use raw I/O. I would prefer the former. -hpa
Rongqing Li
2013-Sep-03 07:02 UTC
[syslinux] [PATCH] tftp-hpa: add error check for disk filled up
On 08/29/2013 03:13 AM, H. Peter Anvin wrote:> On 08/21/2013 06:00 PM, rongqing.li at windriver.com wrote: >> From: "Roy.Li" <rongqing.li at windriver.com> >> >> Add error check when the write-buffer is finally flushed to the file, >> the caller can detect if the disk filled up (or had an i/o error) and >> return a NOSAPCE nak to the other side. >> >> Signed-off-by: Ming Liu <ming.liu at windriver.com> >> Signed-off-by: Roy.Li <rongqing.li at windriver.com> > > Don't mix stdio and raw I/O this way. > > The right thing to do is either to use fflush() and ferror() to detect > an error condition, or consistently use raw I/O. I would prefer the former. > > -hpa >Could we check putc return value directly? -Roy> > >-- Best Reagrds, Roy | RongQing Li
Possibly Parallel Threads
- [PATCH v2] tftp-hpa: add error check for disk filled up
- [PATCH] tftp-hpa: add error check for disk filled up
- Bug#771441: [PATCH tftpd-hpa] tftpd: don't use AI_CANONNAME and AI_ADDRCONFIG to resolve addresses for bind
- [PATCH tftpd-hpa] tftpd: don't use AI_CANONNAME and AI_ADDRCONFIG to resolve addresses for bind
- Bug#771441: [PATCH tftpd-hpa] tftpd: don't use AI_CANONNAME and AI_ADDRCONFIG to resolve addresses for bind