Systemd/Microsoft Effort For A Global Counter On Block/Disk Changes Coming To Linux 5.15

[skeevy420]

Same man, same. I'm 35, gonna be 36 in September. I haven't watched really in over a decade. But the 90s era with Ultimate Warrior, Macho Man, Undertaker, and those guys and then the Attitude Era in the '2000s with Stone Cold, The Rock, Mankind, etc. Best times. Some of these dudes are still around (Undertaker). incredible. Got lots of respect for wrestlers and wrestling.

Ditto, change the month to November.

[Filiprino]

you cannot tell whether it's because the right uevent has not arrived yet, or it was already sent and you missed it. So you cannot tell whether you should wait for it or not.

You can time-out and retry. It's a distributed system, it's a network. TCP, UDP... right?
Imagine you have a non well-behaving device and you don't time out... you'll be stuck forever waiting for it if you assume an event will always happen.

[teknoraver]

I wonder if this developer had any experience with the Windows kernel code.
I would love to read an article by someone with experience with both the Linux kernel and the Windows kernel, and to hear opinions, impressions, preferences, pros/cons, and such.

I have zero experience with Windows code.
I don't know people with experience with both kernels.
Maybe the Red Hat developers which maintain the VirtIO Windows drivers can give you some insight.

[teknoraver]

You can time-out and retry. It's a distributed system, it's a network. TCP, UDP... right?
Imagine you have a non well-behaving device and you don't time out... you'll be stuck forever waiting for it if you assume an event will always happen.

Completely wrong. This is not TCP or UDP. This is not even IP networking, events are passed by the kernel to an userspace handler.

[thunderbird32]

Same man, same. I'm 35, gonna be 36 in September. I haven't watched really in over a decade. But the 90s era with Ultimate Warrior, Macho Man, Undertaker, and those guys and then the Attitude Era in the '2000s with Stone Cold, The Rock, Mankind, etc. Best times. Some of these dudes are still around (Undertaker). incredible. Got lots of respect for wrestlers and wrestling.

Undertaker is fully retired now, just FYI. As of last November's Survivor Series. From that era, Edge and Christian are both still wrestling (though Christian is in AEW now since, reportedly, WWE didn't want to give him a full time contract).

EDIT: Also, Matt and Jeff Hardy are both still wrestling as well. Same situation as Edge and Christian though, with one in AEW and one in WWE.

[perpetually high]

Undertaker is fully retired now, just FYI. As of last November's Survivor Series. From that era, Edge and Christian are both still wrestling (though Christian is in AEW now since, reportedly, WWE didn't want to give him a full time contract).

Ah, ok. What a career for Taker. Easily possibly going down at the GOAT. And I just saw that promo for Edge coming back the other day. Looked awesome

[cl333r]

Somewhat off topic: anyone knows why udev_monitor notifies of partition added/removed events, but mount/unmount never or almost never?

struct udev_monitor *monitor = udev_monitor_new_from_netlink(udev, "udev");
const char *subsys = "block";
udev_monitor_filter_add_match_subsystem_devtype(monitor, subsys, "disk");
udev_monitor_filter_add_match_subsystem_devtype(monitor, subsys, "partition");
udev_monitor_enable_receiving(monitor);
int fd = udev_monitor_get_fd(monitor);
struct timeval tv;

while (1) {
// reading events using select()
 // ONLY added/removed events are received regularly.
// mount/unmount only sometimes which defeats the purpose.
usleep(500 * 1000); /// 500 ms
} 

PS: The code insertion is glitchy AF, randomly removes whitespaces and inserts in other places, inserts smilies, Michael wth?

[gens]

Somewhat off topic: anyone knows why udev_monitor notifies of partition added/removed events, but mount/unmount never or almost never?

struct udev_monitor *monitor = udev_monitor_new_from_netlink(udev, "udev");
const char *subsys = "block";
udev_monitor_filter_add_match_subsystem_devtype(monitor, subsys, "disk");
udev_monitor_filter_add_match_subsystem_devtype(monitor, subsys, "partition");
udev_monitor_enable_receiving(monitor);
int fd = udev_monitor_get_fd(monitor);
struct timeval tv;

while (1) {
// reading events using select()
// ONLY added/removed events are received regularly.
// mount/unmount only sometimes which defeats the purpose.
usleep(500 * 1000); /// 500 ms
} 

PS: The code insertion is glitchy AF, randomly removes whitespaces and inserts in other places, inserts smilies, Michael wth?

Just a guess..

It is because mounting does not have anything directly to do with a device. `mount` command, for example, takes a "device file" as a source (example /dev/sda1). That device file is.., lets say, made by udev.

[sindr]

I have zero experience with Windows code.
I don't know people with experience with both kernels.
Maybe the Red Hat developers which maintain the VirtIO Windows drivers can give you some insight.

There's some comments on HN comparing Linux to the Windows NT kernel. The brainchild of ex-DEC engineers, it could be considered the spiritual successor to VMS . I'm told at the API level, the NT kernel compares favourably to Linux.

[MadeUpName]

Is there a use case where this would be preferable to looking at the SMART data which is actually stored on the disk?