This is the mail archive of the
ecos-patches@sourceware.org
mailing list for the eCos project.
Patch_or_Contribution granted: [Bug 1001633] DHCP Client may hang
- From: bugzilla-daemon at bugs dot ecos dot sourceware dot org
- To: ecos-patches at sourceware dot org
- Date: Thu, 26 Jul 2012 10:21:52 +0100
- Subject: Patch_or_Contribution granted: [Bug 1001633] DHCP Client may hang
- Auto-submitted: auto-generated
- References: <bug-1001633-0@http.bugs.ecos.sourceware.org/>
Bernd Edlinger <bernd.edlinger@hotmail.de> has granted Patch_or_Contribution:
Bug 1001633: DHCP Client may hang
http://bugs.ecos.sourceware.org/show_bug.cgi?id=1001633
------- Additional Comments from Bernd Edlinger <bernd.edlinger@hotmail.de>
when a DHCP Response is received it should contain a time slice value.
however if the time slice value is zero, the device enters an endless loop.
Hence, this configuration error in the DHCP server crashes the devices.
Reproduce:
install tftpd32 on an own subnet, and configure the dhcp server
to use zero-length time slices.
The time slice value is not checked by the dhcp server, and just used as is.
When the eCos DHCP Client is booted, it gets an address assigned, and freezes
immediately, for instance no ping or arp responses are sent.
Additionally, the code that sends new dhcp requests
uses a wrong formula for bp_secs:
bp_secs = cyg_current_time() / 100;
But that is only correct for 100 Ticks/second.
bp_secs = cyg_current_time() /
((1000000000LL*CYGNUM_HAL_RTC_DENOMINATOR)/CYGNUM_HAL_RTC_NUMERATOR);
Howto Fix:
Attached you will find a patch for the file
"ecos/packages/net/common/current/src/dhcp_prot.c"
which checks for zero timeout and silently changes this to 1 second timeout.