sapling/eden/fs/fuse/RequestData.cpp

166 lines
4.6 KiB
C++
Raw Normal View History

/*
* Copyright (c) Facebook, Inc. and its affiliates.
*
* This software may be used and distributed according to the terms of the
* GNU General Public License version 2.
*/
#include "eden/fs/fuse/RequestData.h"
#include <folly/Subprocess.h>
#include <folly/logging/xlog.h>
#include "eden/fs/fuse/Dispatcher.h"
#include "eden/fs/notifications/Notifications.h"
#include "eden/fs/telemetry/RequestMetricsScope.h"
#include "eden/fs/utils/SystemError.h"
using namespace folly;
using namespace std::chrono;
namespace facebook {
namespace eden {
const std::string RequestData::kKey("fuse");
remove dep on libfuse Summary: This serves a few purposes: 1. We can avoid some conditional code inside eden if we know that we have a specific fuse_kernel.h header implementation. 2. We don't have to figure out a way to propagate the kernel capabilities through the graceful restart process. 3. libfuse3 removed the channel/session hooks that we've been using thus far to interject ourselves for mounting and graceful restarting, so we were already effectively the walking dead here. 4. We're now able to take advtange of the latest aspects of the fuse kernel interface without being tied to the implementation of libfuse2 or libfuse3. We're interested in the readdirplus functionality and will look at enabling that in a future diff. This may make some things slightly harder for the more immediate macOS port but I belive that we're in a much better place overall. This diff is relatively mechanical and sadly is (unavoidably) large. The main aspects of this diff are: 1. The `fuse_ino_t` type was provided by libfuse so we needed to replace it with our own definition. This has decent penetration throughout the codebase. 2. The confusing `fuse_file_info` type that was multi-purpose and had fields that were sometimes *in* parameters and sometimes *out* parameters has been removed and replaced with a simpler *flags* parameter that corresponds to the `open(2)` flags parameter. The *out* portions are subsumed by existing file handle metadata methods. 3. The fuse parameters returned from variations of the `LOOKUP` opcode now return the fuse kernel type for this directly. I suspect that we may need to introduce a compatibility type when we revisit the macOS port, but this at least makes this diff slightly simpler. You'll notice that some field and symbol name prefixes vary as a result of this. 4. Similarly for `setattr`, libfuse separated the kernel data into two parameters that were a little awkward to use; we're now just passing the kernel data through and this, IMO, makes the interface slightly more understandable. 5. The bulk of the code from `Dispatcher.cpp` that shimmed the libfuse callbacks into the C++ virtual methods has been removed and replaced by a `switch` statement based dispatcher in `FuseChannel`. I'm not married to this being `switch` based and may revise this to be driven by an `unordered_map` of opcode -> dispatcher method defined in `FuseChannel`. Regardless, `Dispatcher.cpp` is now much slimmer and should be easier to replace by rolling it together into `EdenDispatcher.cpp` in the future should we desire to do so. 6. This diff disables dispatching `poll` and `ioctl` calls. We didn't make use of them and their interfaces are a bit fiddly. 7. `INTERRUPT` is also disabled here. I will re-enable it in a follow-up diff where I can also revise how we track outstanding requests for graceful shutdown. 8. I've imported `fuse_kernel.h` from libfuse. This is included under the permissive 2-clause BSD license that it allows for exactly this integration purpose. Reviewed By: simpkins Differential Revision: D6576472 fbshipit-source-id: 7cb088af5e06fe27bf22a1bed295c18c17d8006c
2018-01-03 03:25:03 +03:00
RequestData::RequestData(
FuseChannel* channel,
const fuse_in_header& fuseHeader,
Dispatcher* dispatcher)
: channel_(channel), fuseHeader_(fuseHeader), dispatcher_(dispatcher) {}
bool RequestData::isFuseRequest() {
return folly::RequestContext::get()->getContextData(kKey) != nullptr;
}
RequestData& RequestData::get() {
const auto data = folly::RequestContext::get()->getContextData(kKey);
if (UNLIKELY(!data)) {
XLOG(FATAL) << "boom for missing RequestData";
throw std::runtime_error("no fuse request data set in this context!");
}
return *dynamic_cast<RequestData*>(data);
}
remove dep on libfuse Summary: This serves a few purposes: 1. We can avoid some conditional code inside eden if we know that we have a specific fuse_kernel.h header implementation. 2. We don't have to figure out a way to propagate the kernel capabilities through the graceful restart process. 3. libfuse3 removed the channel/session hooks that we've been using thus far to interject ourselves for mounting and graceful restarting, so we were already effectively the walking dead here. 4. We're now able to take advtange of the latest aspects of the fuse kernel interface without being tied to the implementation of libfuse2 or libfuse3. We're interested in the readdirplus functionality and will look at enabling that in a future diff. This may make some things slightly harder for the more immediate macOS port but I belive that we're in a much better place overall. This diff is relatively mechanical and sadly is (unavoidably) large. The main aspects of this diff are: 1. The `fuse_ino_t` type was provided by libfuse so we needed to replace it with our own definition. This has decent penetration throughout the codebase. 2. The confusing `fuse_file_info` type that was multi-purpose and had fields that were sometimes *in* parameters and sometimes *out* parameters has been removed and replaced with a simpler *flags* parameter that corresponds to the `open(2)` flags parameter. The *out* portions are subsumed by existing file handle metadata methods. 3. The fuse parameters returned from variations of the `LOOKUP` opcode now return the fuse kernel type for this directly. I suspect that we may need to introduce a compatibility type when we revisit the macOS port, but this at least makes this diff slightly simpler. You'll notice that some field and symbol name prefixes vary as a result of this. 4. Similarly for `setattr`, libfuse separated the kernel data into two parameters that were a little awkward to use; we're now just passing the kernel data through and this, IMO, makes the interface slightly more understandable. 5. The bulk of the code from `Dispatcher.cpp` that shimmed the libfuse callbacks into the C++ virtual methods has been removed and replaced by a `switch` statement based dispatcher in `FuseChannel`. I'm not married to this being `switch` based and may revise this to be driven by an `unordered_map` of opcode -> dispatcher method defined in `FuseChannel`. Regardless, `Dispatcher.cpp` is now much slimmer and should be easier to replace by rolling it together into `EdenDispatcher.cpp` in the future should we desire to do so. 6. This diff disables dispatching `poll` and `ioctl` calls. We didn't make use of them and their interfaces are a bit fiddly. 7. `INTERRUPT` is also disabled here. I will re-enable it in a follow-up diff where I can also revise how we track outstanding requests for graceful shutdown. 8. I've imported `fuse_kernel.h` from libfuse. This is included under the permissive 2-clause BSD license that it allows for exactly this integration purpose. Reviewed By: simpkins Differential Revision: D6576472 fbshipit-source-id: 7cb088af5e06fe27bf22a1bed295c18c17d8006c
2018-01-03 03:25:03 +03:00
RequestData& RequestData::create(
FuseChannel* channel,
const fuse_in_header& fuseHeader,
Dispatcher* dispatcher) {
folly::RequestContext::get()->setContextData(
remove dep on libfuse Summary: This serves a few purposes: 1. We can avoid some conditional code inside eden if we know that we have a specific fuse_kernel.h header implementation. 2. We don't have to figure out a way to propagate the kernel capabilities through the graceful restart process. 3. libfuse3 removed the channel/session hooks that we've been using thus far to interject ourselves for mounting and graceful restarting, so we were already effectively the walking dead here. 4. We're now able to take advtange of the latest aspects of the fuse kernel interface without being tied to the implementation of libfuse2 or libfuse3. We're interested in the readdirplus functionality and will look at enabling that in a future diff. This may make some things slightly harder for the more immediate macOS port but I belive that we're in a much better place overall. This diff is relatively mechanical and sadly is (unavoidably) large. The main aspects of this diff are: 1. The `fuse_ino_t` type was provided by libfuse so we needed to replace it with our own definition. This has decent penetration throughout the codebase. 2. The confusing `fuse_file_info` type that was multi-purpose and had fields that were sometimes *in* parameters and sometimes *out* parameters has been removed and replaced with a simpler *flags* parameter that corresponds to the `open(2)` flags parameter. The *out* portions are subsumed by existing file handle metadata methods. 3. The fuse parameters returned from variations of the `LOOKUP` opcode now return the fuse kernel type for this directly. I suspect that we may need to introduce a compatibility type when we revisit the macOS port, but this at least makes this diff slightly simpler. You'll notice that some field and symbol name prefixes vary as a result of this. 4. Similarly for `setattr`, libfuse separated the kernel data into two parameters that were a little awkward to use; we're now just passing the kernel data through and this, IMO, makes the interface slightly more understandable. 5. The bulk of the code from `Dispatcher.cpp` that shimmed the libfuse callbacks into the C++ virtual methods has been removed and replaced by a `switch` statement based dispatcher in `FuseChannel`. I'm not married to this being `switch` based and may revise this to be driven by an `unordered_map` of opcode -> dispatcher method defined in `FuseChannel`. Regardless, `Dispatcher.cpp` is now much slimmer and should be easier to replace by rolling it together into `EdenDispatcher.cpp` in the future should we desire to do so. 6. This diff disables dispatching `poll` and `ioctl` calls. We didn't make use of them and their interfaces are a bit fiddly. 7. `INTERRUPT` is also disabled here. I will re-enable it in a follow-up diff where I can also revise how we track outstanding requests for graceful shutdown. 8. I've imported `fuse_kernel.h` from libfuse. This is included under the permissive 2-clause BSD license that it allows for exactly this integration purpose. Reviewed By: simpkins Differential Revision: D6576472 fbshipit-source-id: 7cb088af5e06fe27bf22a1bed295c18c17d8006c
2018-01-03 03:25:03 +03:00
RequestData::kKey,
std::make_unique<RequestData>(channel, fuseHeader, dispatcher));
return get();
}
void RequestData::startRequest(
EdenStats* stats,
FuseThreadStats::HistogramPtr histogram,
std::shared_ptr<RequestMetricsScope::LockedRequestWatchList>&
requestWatches) {
startTime_ = steady_clock::now();
DCHECK(latencyHistogram_ == nullptr);
latencyHistogram_ = histogram;
stats_ = stats;
channelThreadLocalStats_ = requestWatches;
requestMetricsScope_ = RequestMetricsScope(channelThreadLocalStats_.get());
}
void RequestData::finishRequest() {
const auto now = steady_clock::now();
const auto now_since_epoch = duration_cast<seconds>(now.time_since_epoch());
const auto diff = now - startTime_;
const auto diff_us = duration_cast<microseconds>(diff);
const auto diff_ns = duration_cast<nanoseconds>(diff);
stats_->getFuseStatsForCurrentThread().recordLatency(
latencyHistogram_, diff_us, now_since_epoch);
latencyHistogram_ = nullptr;
stats_ = nullptr;
{ auto temp = std::move(requestMetricsScope_); }
channelThreadLocalStats_.reset();
auto& pal = channel_->getProcessAccessLog();
if (getEdenTopStats().didImportFromBackingStore()) {
auto type = ProcessAccessLog::AccessType::FuseBackingStoreImport;
pal.recordAccess(examineReq().pid, type);
}
pal.recordDuration(examineReq().pid, diff_ns);
}
remove dep on libfuse Summary: This serves a few purposes: 1. We can avoid some conditional code inside eden if we know that we have a specific fuse_kernel.h header implementation. 2. We don't have to figure out a way to propagate the kernel capabilities through the graceful restart process. 3. libfuse3 removed the channel/session hooks that we've been using thus far to interject ourselves for mounting and graceful restarting, so we were already effectively the walking dead here. 4. We're now able to take advtange of the latest aspects of the fuse kernel interface without being tied to the implementation of libfuse2 or libfuse3. We're interested in the readdirplus functionality and will look at enabling that in a future diff. This may make some things slightly harder for the more immediate macOS port but I belive that we're in a much better place overall. This diff is relatively mechanical and sadly is (unavoidably) large. The main aspects of this diff are: 1. The `fuse_ino_t` type was provided by libfuse so we needed to replace it with our own definition. This has decent penetration throughout the codebase. 2. The confusing `fuse_file_info` type that was multi-purpose and had fields that were sometimes *in* parameters and sometimes *out* parameters has been removed and replaced with a simpler *flags* parameter that corresponds to the `open(2)` flags parameter. The *out* portions are subsumed by existing file handle metadata methods. 3. The fuse parameters returned from variations of the `LOOKUP` opcode now return the fuse kernel type for this directly. I suspect that we may need to introduce a compatibility type when we revisit the macOS port, but this at least makes this diff slightly simpler. You'll notice that some field and symbol name prefixes vary as a result of this. 4. Similarly for `setattr`, libfuse separated the kernel data into two parameters that were a little awkward to use; we're now just passing the kernel data through and this, IMO, makes the interface slightly more understandable. 5. The bulk of the code from `Dispatcher.cpp` that shimmed the libfuse callbacks into the C++ virtual methods has been removed and replaced by a `switch` statement based dispatcher in `FuseChannel`. I'm not married to this being `switch` based and may revise this to be driven by an `unordered_map` of opcode -> dispatcher method defined in `FuseChannel`. Regardless, `Dispatcher.cpp` is now much slimmer and should be easier to replace by rolling it together into `EdenDispatcher.cpp` in the future should we desire to do so. 6. This diff disables dispatching `poll` and `ioctl` calls. We didn't make use of them and their interfaces are a bit fiddly. 7. `INTERRUPT` is also disabled here. I will re-enable it in a follow-up diff where I can also revise how we track outstanding requests for graceful shutdown. 8. I've imported `fuse_kernel.h` from libfuse. This is included under the permissive 2-clause BSD license that it allows for exactly this integration purpose. Reviewed By: simpkins Differential Revision: D6576472 fbshipit-source-id: 7cb088af5e06fe27bf22a1bed295c18c17d8006c
2018-01-03 03:25:03 +03:00
fuse_in_header RequestData::stealReq() {
if (fuseHeader_.opcode == 0) {
throw std::runtime_error("req_ has been released");
}
remove dep on libfuse Summary: This serves a few purposes: 1. We can avoid some conditional code inside eden if we know that we have a specific fuse_kernel.h header implementation. 2. We don't have to figure out a way to propagate the kernel capabilities through the graceful restart process. 3. libfuse3 removed the channel/session hooks that we've been using thus far to interject ourselves for mounting and graceful restarting, so we were already effectively the walking dead here. 4. We're now able to take advtange of the latest aspects of the fuse kernel interface without being tied to the implementation of libfuse2 or libfuse3. We're interested in the readdirplus functionality and will look at enabling that in a future diff. This may make some things slightly harder for the more immediate macOS port but I belive that we're in a much better place overall. This diff is relatively mechanical and sadly is (unavoidably) large. The main aspects of this diff are: 1. The `fuse_ino_t` type was provided by libfuse so we needed to replace it with our own definition. This has decent penetration throughout the codebase. 2. The confusing `fuse_file_info` type that was multi-purpose and had fields that were sometimes *in* parameters and sometimes *out* parameters has been removed and replaced with a simpler *flags* parameter that corresponds to the `open(2)` flags parameter. The *out* portions are subsumed by existing file handle metadata methods. 3. The fuse parameters returned from variations of the `LOOKUP` opcode now return the fuse kernel type for this directly. I suspect that we may need to introduce a compatibility type when we revisit the macOS port, but this at least makes this diff slightly simpler. You'll notice that some field and symbol name prefixes vary as a result of this. 4. Similarly for `setattr`, libfuse separated the kernel data into two parameters that were a little awkward to use; we're now just passing the kernel data through and this, IMO, makes the interface slightly more understandable. 5. The bulk of the code from `Dispatcher.cpp` that shimmed the libfuse callbacks into the C++ virtual methods has been removed and replaced by a `switch` statement based dispatcher in `FuseChannel`. I'm not married to this being `switch` based and may revise this to be driven by an `unordered_map` of opcode -> dispatcher method defined in `FuseChannel`. Regardless, `Dispatcher.cpp` is now much slimmer and should be easier to replace by rolling it together into `EdenDispatcher.cpp` in the future should we desire to do so. 6. This diff disables dispatching `poll` and `ioctl` calls. We didn't make use of them and their interfaces are a bit fiddly. 7. `INTERRUPT` is also disabled here. I will re-enable it in a follow-up diff where I can also revise how we track outstanding requests for graceful shutdown. 8. I've imported `fuse_kernel.h` from libfuse. This is included under the permissive 2-clause BSD license that it allows for exactly this integration purpose. Reviewed By: simpkins Differential Revision: D6576472 fbshipit-source-id: 7cb088af5e06fe27bf22a1bed295c18c17d8006c
2018-01-03 03:25:03 +03:00
fuse_in_header res = fuseHeader_;
fuseHeader_.opcode = 0;
return res;
}
remove dep on libfuse Summary: This serves a few purposes: 1. We can avoid some conditional code inside eden if we know that we have a specific fuse_kernel.h header implementation. 2. We don't have to figure out a way to propagate the kernel capabilities through the graceful restart process. 3. libfuse3 removed the channel/session hooks that we've been using thus far to interject ourselves for mounting and graceful restarting, so we were already effectively the walking dead here. 4. We're now able to take advtange of the latest aspects of the fuse kernel interface without being tied to the implementation of libfuse2 or libfuse3. We're interested in the readdirplus functionality and will look at enabling that in a future diff. This may make some things slightly harder for the more immediate macOS port but I belive that we're in a much better place overall. This diff is relatively mechanical and sadly is (unavoidably) large. The main aspects of this diff are: 1. The `fuse_ino_t` type was provided by libfuse so we needed to replace it with our own definition. This has decent penetration throughout the codebase. 2. The confusing `fuse_file_info` type that was multi-purpose and had fields that were sometimes *in* parameters and sometimes *out* parameters has been removed and replaced with a simpler *flags* parameter that corresponds to the `open(2)` flags parameter. The *out* portions are subsumed by existing file handle metadata methods. 3. The fuse parameters returned from variations of the `LOOKUP` opcode now return the fuse kernel type for this directly. I suspect that we may need to introduce a compatibility type when we revisit the macOS port, but this at least makes this diff slightly simpler. You'll notice that some field and symbol name prefixes vary as a result of this. 4. Similarly for `setattr`, libfuse separated the kernel data into two parameters that were a little awkward to use; we're now just passing the kernel data through and this, IMO, makes the interface slightly more understandable. 5. The bulk of the code from `Dispatcher.cpp` that shimmed the libfuse callbacks into the C++ virtual methods has been removed and replaced by a `switch` statement based dispatcher in `FuseChannel`. I'm not married to this being `switch` based and may revise this to be driven by an `unordered_map` of opcode -> dispatcher method defined in `FuseChannel`. Regardless, `Dispatcher.cpp` is now much slimmer and should be easier to replace by rolling it together into `EdenDispatcher.cpp` in the future should we desire to do so. 6. This diff disables dispatching `poll` and `ioctl` calls. We didn't make use of them and their interfaces are a bit fiddly. 7. `INTERRUPT` is also disabled here. I will re-enable it in a follow-up diff where I can also revise how we track outstanding requests for graceful shutdown. 8. I've imported `fuse_kernel.h` from libfuse. This is included under the permissive 2-clause BSD license that it allows for exactly this integration purpose. Reviewed By: simpkins Differential Revision: D6576472 fbshipit-source-id: 7cb088af5e06fe27bf22a1bed295c18c17d8006c
2018-01-03 03:25:03 +03:00
const fuse_in_header& RequestData::getReq() const {
if (fuseHeader_.opcode == 0) {
throw std::runtime_error("req_ has been released");
}
remove dep on libfuse Summary: This serves a few purposes: 1. We can avoid some conditional code inside eden if we know that we have a specific fuse_kernel.h header implementation. 2. We don't have to figure out a way to propagate the kernel capabilities through the graceful restart process. 3. libfuse3 removed the channel/session hooks that we've been using thus far to interject ourselves for mounting and graceful restarting, so we were already effectively the walking dead here. 4. We're now able to take advtange of the latest aspects of the fuse kernel interface without being tied to the implementation of libfuse2 or libfuse3. We're interested in the readdirplus functionality and will look at enabling that in a future diff. This may make some things slightly harder for the more immediate macOS port but I belive that we're in a much better place overall. This diff is relatively mechanical and sadly is (unavoidably) large. The main aspects of this diff are: 1. The `fuse_ino_t` type was provided by libfuse so we needed to replace it with our own definition. This has decent penetration throughout the codebase. 2. The confusing `fuse_file_info` type that was multi-purpose and had fields that were sometimes *in* parameters and sometimes *out* parameters has been removed and replaced with a simpler *flags* parameter that corresponds to the `open(2)` flags parameter. The *out* portions are subsumed by existing file handle metadata methods. 3. The fuse parameters returned from variations of the `LOOKUP` opcode now return the fuse kernel type for this directly. I suspect that we may need to introduce a compatibility type when we revisit the macOS port, but this at least makes this diff slightly simpler. You'll notice that some field and symbol name prefixes vary as a result of this. 4. Similarly for `setattr`, libfuse separated the kernel data into two parameters that were a little awkward to use; we're now just passing the kernel data through and this, IMO, makes the interface slightly more understandable. 5. The bulk of the code from `Dispatcher.cpp` that shimmed the libfuse callbacks into the C++ virtual methods has been removed and replaced by a `switch` statement based dispatcher in `FuseChannel`. I'm not married to this being `switch` based and may revise this to be driven by an `unordered_map` of opcode -> dispatcher method defined in `FuseChannel`. Regardless, `Dispatcher.cpp` is now much slimmer and should be easier to replace by rolling it together into `EdenDispatcher.cpp` in the future should we desire to do so. 6. This diff disables dispatching `poll` and `ioctl` calls. We didn't make use of them and their interfaces are a bit fiddly. 7. `INTERRUPT` is also disabled here. I will re-enable it in a follow-up diff where I can also revise how we track outstanding requests for graceful shutdown. 8. I've imported `fuse_kernel.h` from libfuse. This is included under the permissive 2-clause BSD license that it allows for exactly this integration purpose. Reviewed By: simpkins Differential Revision: D6576472 fbshipit-source-id: 7cb088af5e06fe27bf22a1bed295c18c17d8006c
2018-01-03 03:25:03 +03:00
return fuseHeader_;
}
const fuse_in_header& RequestData::examineReq() const {
// Will just return the fuseHeader_ and not throw(unlike getReq)
// The caller is responsible to check the opcode and ignore if zero
return fuseHeader_;
}
Dispatcher* RequestData::getDispatcher() const {
return dispatcher_;
}
RequestData::EdenTopStats& RequestData::getEdenTopStats() {
return edenTopStats_;
}
void RequestData::replyError(int err) {
remove dep on libfuse Summary: This serves a few purposes: 1. We can avoid some conditional code inside eden if we know that we have a specific fuse_kernel.h header implementation. 2. We don't have to figure out a way to propagate the kernel capabilities through the graceful restart process. 3. libfuse3 removed the channel/session hooks that we've been using thus far to interject ourselves for mounting and graceful restarting, so we were already effectively the walking dead here. 4. We're now able to take advtange of the latest aspects of the fuse kernel interface without being tied to the implementation of libfuse2 or libfuse3. We're interested in the readdirplus functionality and will look at enabling that in a future diff. This may make some things slightly harder for the more immediate macOS port but I belive that we're in a much better place overall. This diff is relatively mechanical and sadly is (unavoidably) large. The main aspects of this diff are: 1. The `fuse_ino_t` type was provided by libfuse so we needed to replace it with our own definition. This has decent penetration throughout the codebase. 2. The confusing `fuse_file_info` type that was multi-purpose and had fields that were sometimes *in* parameters and sometimes *out* parameters has been removed and replaced with a simpler *flags* parameter that corresponds to the `open(2)` flags parameter. The *out* portions are subsumed by existing file handle metadata methods. 3. The fuse parameters returned from variations of the `LOOKUP` opcode now return the fuse kernel type for this directly. I suspect that we may need to introduce a compatibility type when we revisit the macOS port, but this at least makes this diff slightly simpler. You'll notice that some field and symbol name prefixes vary as a result of this. 4. Similarly for `setattr`, libfuse separated the kernel data into two parameters that were a little awkward to use; we're now just passing the kernel data through and this, IMO, makes the interface slightly more understandable. 5. The bulk of the code from `Dispatcher.cpp` that shimmed the libfuse callbacks into the C++ virtual methods has been removed and replaced by a `switch` statement based dispatcher in `FuseChannel`. I'm not married to this being `switch` based and may revise this to be driven by an `unordered_map` of opcode -> dispatcher method defined in `FuseChannel`. Regardless, `Dispatcher.cpp` is now much slimmer and should be easier to replace by rolling it together into `EdenDispatcher.cpp` in the future should we desire to do so. 6. This diff disables dispatching `poll` and `ioctl` calls. We didn't make use of them and their interfaces are a bit fiddly. 7. `INTERRUPT` is also disabled here. I will re-enable it in a follow-up diff where I can also revise how we track outstanding requests for graceful shutdown. 8. I've imported `fuse_kernel.h` from libfuse. This is included under the permissive 2-clause BSD license that it allows for exactly this integration purpose. Reviewed By: simpkins Differential Revision: D6576472 fbshipit-source-id: 7cb088af5e06fe27bf22a1bed295c18c17d8006c
2018-01-03 03:25:03 +03:00
channel_->replyError(stealReq(), err);
}
void RequestData::replyNone() {
remove dep on libfuse Summary: This serves a few purposes: 1. We can avoid some conditional code inside eden if we know that we have a specific fuse_kernel.h header implementation. 2. We don't have to figure out a way to propagate the kernel capabilities through the graceful restart process. 3. libfuse3 removed the channel/session hooks that we've been using thus far to interject ourselves for mounting and graceful restarting, so we were already effectively the walking dead here. 4. We're now able to take advtange of the latest aspects of the fuse kernel interface without being tied to the implementation of libfuse2 or libfuse3. We're interested in the readdirplus functionality and will look at enabling that in a future diff. This may make some things slightly harder for the more immediate macOS port but I belive that we're in a much better place overall. This diff is relatively mechanical and sadly is (unavoidably) large. The main aspects of this diff are: 1. The `fuse_ino_t` type was provided by libfuse so we needed to replace it with our own definition. This has decent penetration throughout the codebase. 2. The confusing `fuse_file_info` type that was multi-purpose and had fields that were sometimes *in* parameters and sometimes *out* parameters has been removed and replaced with a simpler *flags* parameter that corresponds to the `open(2)` flags parameter. The *out* portions are subsumed by existing file handle metadata methods. 3. The fuse parameters returned from variations of the `LOOKUP` opcode now return the fuse kernel type for this directly. I suspect that we may need to introduce a compatibility type when we revisit the macOS port, but this at least makes this diff slightly simpler. You'll notice that some field and symbol name prefixes vary as a result of this. 4. Similarly for `setattr`, libfuse separated the kernel data into two parameters that were a little awkward to use; we're now just passing the kernel data through and this, IMO, makes the interface slightly more understandable. 5. The bulk of the code from `Dispatcher.cpp` that shimmed the libfuse callbacks into the C++ virtual methods has been removed and replaced by a `switch` statement based dispatcher in `FuseChannel`. I'm not married to this being `switch` based and may revise this to be driven by an `unordered_map` of opcode -> dispatcher method defined in `FuseChannel`. Regardless, `Dispatcher.cpp` is now much slimmer and should be easier to replace by rolling it together into `EdenDispatcher.cpp` in the future should we desire to do so. 6. This diff disables dispatching `poll` and `ioctl` calls. We didn't make use of them and their interfaces are a bit fiddly. 7. `INTERRUPT` is also disabled here. I will re-enable it in a follow-up diff where I can also revise how we track outstanding requests for graceful shutdown. 8. I've imported `fuse_kernel.h` from libfuse. This is included under the permissive 2-clause BSD license that it allows for exactly this integration purpose. Reviewed By: simpkins Differential Revision: D6576472 fbshipit-source-id: 7cb088af5e06fe27bf22a1bed295c18c17d8006c
2018-01-03 03:25:03 +03:00
stealReq();
}
void RequestData::systemErrorHandler(
const std::system_error& err,
Notifications* FOLLY_NULLABLE notifications) {
int errnum = EIO;
if (isErrnoError(err)) {
errnum = err.code().value();
}
XLOG(DBG5) << folly::exceptionStr(err);
RequestData::get().replyError(errnum);
if (notifications) {
notifications->showGenericErrorNotification(err);
}
}
void RequestData::genericErrorHandler(
const std::exception& err,
Notifications* FOLLY_NULLABLE notifications) {
XLOG(DBG5) << folly::exceptionStr(err);
RequestData::get().replyError(EIO);
if (notifications) {
notifications->showGenericErrorNotification(err);
}
}
void RequestData::timeoutErrorHandler(
const folly::FutureTimeout& err,
Notifications* FOLLY_NULLABLE notifications) {
XLOG_EVERY_MS(WARN, 1000)
<< "FUSE request timed out: " << folly::exceptionStr(err);
eden: add timeout to FuseChannel requests Summary: The macOS FUSE implementation has the concept of a daemon timeout, which is used to set an upper bound on the length of time that the kernel will wait for a request to be satisfied. If a request takes too long, the kernel will shutdown the fuse device and the user is left with a relatively broken experience in their repo until they restart the eden server. Critically, when it is left in the broken state it is hard for the user to realize that they need to restart the server because the kernel will still respond with EIO to file accesses within the mount; even though it has been stopped, it isn't fully unmounted until the fuse process stops. This diff introduces a self imposed timeout in the fuse processing flow and emits an ETIMEDOUT error when it is reached. The intent is that we'll configure this timeout to be smaller than the macOS daemon_timeout (which will be adjusted in a separate diff) so that we don't trigger the problematic behavior when the kernel decides that we've timedout. This change is made for all fuse implementations, not just macOS, for the sake of consistency: there's value to doing this on Linux as well, to avoid some deadlock like scenarios: this should put an upper bound on blocking in certain situations. I've made the timeout default to 60 seconds, but haven't added any configuration code for this yet; will do that in a follow on diff. Reviewed By: chadaustin Differential Revision: D16917954 fbshipit-source-id: 675539c43cf7f0009fd65d138081b9126464b7e0
2019-08-28 19:40:34 +03:00
RequestData::get().replyError(ETIMEDOUT);
if (notifications) {
notifications->showGenericErrorNotification(err);
}
eden: add timeout to FuseChannel requests Summary: The macOS FUSE implementation has the concept of a daemon timeout, which is used to set an upper bound on the length of time that the kernel will wait for a request to be satisfied. If a request takes too long, the kernel will shutdown the fuse device and the user is left with a relatively broken experience in their repo until they restart the eden server. Critically, when it is left in the broken state it is hard for the user to realize that they need to restart the server because the kernel will still respond with EIO to file accesses within the mount; even though it has been stopped, it isn't fully unmounted until the fuse process stops. This diff introduces a self imposed timeout in the fuse processing flow and emits an ETIMEDOUT error when it is reached. The intent is that we'll configure this timeout to be smaller than the macOS daemon_timeout (which will be adjusted in a separate diff) so that we don't trigger the problematic behavior when the kernel decides that we've timedout. This change is made for all fuse implementations, not just macOS, for the sake of consistency: there's value to doing this on Linux as well, to avoid some deadlock like scenarios: this should put an upper bound on blocking in certain situations. I've made the timeout default to 60 seconds, but haven't added any configuration code for this yet; will do that in a follow on diff. Reviewed By: chadaustin Differential Revision: D16917954 fbshipit-source-id: 675539c43cf7f0009fd65d138081b9126464b7e0
2019-08-28 19:40:34 +03:00
}
} // namespace eden
} // namespace facebook