Hi, I have a couple of general doubts on kernel timers and the context in which they run. I gather that a kernel timer would run in the bottom_half interrupt context (please correct me if wrong), and as people have pointed out multiple times, one should never write code which may sleep in the bottom_half context? I have an application in which I am required to buffer packets meant to a network card (which I can accomplish with the skb_queue* API). When a particular timer fires, the buffered packets are to be sent to the network card. I am intercepting the packets in the "bonding" drivers send function, here I buffer the packet and set up a couple of timers. So, my question here is that, will sending something of the order of a 500-1000 packets to the network cards queue from within the timeout function a safe operation? Guessing from my little knowledge about the Linux kernel, aren't bottom half contexts meant for bulk processing of kernel jobs, which makes it perfectly alright to deal with such large amounts of code? Also, even though until dev_queue_xmit the send procedure for a packet is executed in the system call (user) context, the actual send may be dealt by a bottom half? That is, once the packet has been added to the network drivers send queue, it is later processed in a bottom_half (most likely a tasklet)? If this is true, would the argument also extend to the timer bottom halves? I would be happy if you can point any mistakes in my general understanding, would help me avoid unnecessary debugging/redesign time. Thanks and regards, - CCK. -- Chandrakanth Chereddi. ECE-UIUC. -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
